Concepts & frameworks

From Nancy Mayer's chapter in Harrington's new book: if unsure, put about 80% in an ETF and about 20% in a couple of stocks, enough to stay engaged without uncomfortable risk.

Weekly survey of individual investors; bears at a year high while earnings estimates rise was cited as a reason to buy volatility rather than de-risk.

A payout policy that sets an explicit minimum cash balance (Amerigo: $30M, sized to cover two quarterly dividends plus buyback optionality) and returns everything above it each quarter as a variable 'performance' dividend, alongside a fixed base dividend and anti-dilution buybacks.

Parts and maintenance spending runs inverse to new-car sales: when new cars become unaffordable, owners keep repairing older ones (cars 4-11 years old are the sweet spot), so AutoZone's best comps came in weak car-sales years (2009, 2011, 2020, 2021).

When an agent sits between the shopper and the store, the store loses the direct relationship: click and behaviour data, the chance to show deals and promotions, and the subtle cues humans respond to but agents ignore.

25+ years of paying and raising the dividend. A good template for what to look for, but they usually trade at premium valuations with low yields.

A covered dividend can still fail if earnings are falling in a non-thriving industry (Wendy's, dying department stores). Be skeptical above about 7% (frankly above 5%) and compare with healthy peers' yields (Advance Auto vs O'Reilly/AutoZone).

A tiny initial float props up a new listing's price; as lock-ups release shares in steps, supply rises and, absent fundamental change, the price falls.

Most stocks' behaviour is a combination of three factor influences: rates up or down, dollar up or down, economy slowing or accelerating.

The proverbial Japanese saver who swaps low-yielding yen into higher-yielding foreign assets and gains again when the dollar rises; the flow reverses if Japanese yields rise.

Noble's coinage: stocks priced on stories plus the passive bid rather than fundamentals, leaving no price discovery.

High-fee structured products that cap both upside and downside; if the distribution is a return of principal, the income never grows and market value erodes. Ask before buying.

The 1999-2000 internet traffic forecasts came true (43%/yr for 25 years), yet the builders went bankrupt; usage growth does not guarantee returns on capital.

Marketplace ads (sponsored listings) are worth paying for only because humans look at the top of the page; an agent that scans every listing for the best deal makes paid placement nearly worthless, so sellers pay less. Carlson's read of why Amazon, with $76B of trailing ads, blocked Meta's Muse.

A low forward P/E, fast growth and a deep moat are reason enough to own a stock; when sentiment is also negative, a story change alone (with no EPS revision) can re-rate it quickly. Carlson's explanation of Meta's +33% month after the Muse launch.

Decades of hedge funds losing money shorting JGBs, until rising Japanese yields finally made it work.

1) Forward EPS comfortably covers the dividend (e.g. Kimberly-Clark $5.12 vs $7.45); 2) a long record of paying and raising it; 3) the hard part: whether management and the board are still committed to it.

Current dividend divided by your original purchase price. A 2% yielder growing its dividend 7%/yr yields about 4% on cost after 10 years; long-held portfolios reach 20-25%+.

Software between providers, payers and patients that automates eligibility checks, claim scrubbing, denial appeals and patient billing; mission-critical, integrated with EHRs, hard to switch (Waystar).

Choose the EV/P/E/FCF multiple a stock deserves from its history and comps, step it down as growth matures, apply it to forecasts; funds think in multiples, so the price converges on it — sell as it is reached.

If you can't explain in two minutes to a 13-year-old why you own a stock, you probably shouldn't own it. Slegers applies it to Fairfax.

When a shock lifts a producer's prices, ask the downstream buyers whether they can absorb or pass on the increase; if not, the margin is temporary (Aurelion's Huntsman exit).

Plot each stock with risk on the y-axis and modelled base-case upside on the x-axis; favour the low-risk, high-return corner over the biggest upside number, especially when the valuation model is low-confidence.

Proprietary seed genetics built over decades (Corteva's Pioneer), licensed to 100+ seed companies, give pricing power even when crop prices fall, because 'farmers buy yield, not commodity inputs'; royalty-positive economics decouple cash flow from acreage.

Bracket a company exploring a sale between what a leveraged buyer can finance (peer EBITDA multiple less net debt) and what a synergy-paying strategic can pay; the base case sits between (Waystar: 14x ≈ $31-33 vs 16x ≈ $37).

Price how many years of windfall freight rates the market assumes (about one) against what one more year would pay (debt repaid plus 20-30% one-off dividends, legacy contracts re-pricing); own them as insurance on the conflict lasting, not as an outright long.

A deposit small enough to self-build is fine and one big enough to attract majors is fine; in between, a developer is "squashed like a grape." For porphyry copper, ~1 billion tonnes is the line where majors and governments engage and the first ~20 years of a mine can be designed.

Grade a miner's assets and its jurisdiction separately, then ask what the same deposits would be worth in Canada, the US or Australia; the gap is the jurisdiction discount you must decide you are paid for.

Freedom Mortgage CEO Stan Middleman's forecast, made about 10 years earlier, that home prices would correct into 2028.

Classic momentum ranks 12-month returns excluding the latest month (so a fresh pullback isn't penalized); comparing it with 3-month momentum shows which groups are migrating into or out of the momentum basket — overlay positioning to judge how far migrants can run.

Judge an MBS REIT on the MBS vs 8-10-year Treasury spread, its mortgage-servicing-rights book and prepayment speeds (4-5% now vs the old 6% modeling floor), not on the direction of rates.

Compare a bank's securities-portfolio yield with the market average. The gap times the book size is income a restructuring would unlock (Schwab ~1pt; Bank of America 1.5pt below market on ~$800B).

Sort companies by whether they shrink or grow their share count. Diluters (net equity issuers in an expansion) are the "problem children" in a downturn. Screen dividend payers for a falling share count and an active buyback.

Managed-money gross longs / (gross longs + gross shorts) in futures + options; unlike net length it shows how big the short side is. Paulo reads ~77% in oil vs >85% at real highs as a sign the move is not crowded and shorts are fuel.

What a major can justify paying per in-ground ounce is roughly 10-20% of the gold price after build capital: $500 margins in 2011 supported $50-100/oz; ~$2,000 margins today support $200-400/oz, while juniors still trade at $30-100/oz.

In-ground ounce math only holds if the asset becomes a mine; majors build one or two at a time, so only top-decile grade, scale, infrastructure and low capital get bought - screen out anything a major wouldn't build.

Life companies should fund 10-15-year liabilities with investment-grade assets of matching duration. Private-credit-owned carriers that buy PE portfolio-company debt break that match and lack the capital for the losses (unlike P&C, which insures idiosyncratic risks).

Judge a multi-metal deposit by recoverable value per tonne of rock after metallurgy, converted to gold-equivalent grade (~$200/t per g/t gold), against mining cost and capital - largely agnostic to the metal mix.

Value a stock as if you owned 100% of it: annual cash returned plus growth in per-share earnings power, assuming a constant multiple; Vinall requires the sum to reach 15%.

Largest layer = highest-quality producers held through volatility; smaller layers of developers then explorers on top; trade around the core and size to your sleeping level.

A commodity bull market as building a stadium, playing the game, then tearing the stadium down — every cyclical bull ends; locate the current phase (first half, break, second half, extra time) to set how aggressive to be. Sep 2026: uranium 10-15 minutes into the second half.

Derivative hedges charge an insurance premium every year; gold is inversely correlated in equity crashes yet drifts up as M2 (~7–8%/yr) outgrows gold supply (~1%/yr), so you are paid to hold the hedge.

Search ads monetize intent a user has already declared, which an agent can fulfil without the clicks; feed ads create or shape intent inside entertainment, which agents compress far less.

Score a company on how much value comes from removing tedious steps (vulnerability) and how much scarce execution remains once those steps are automated (resilience); high friction + low scarcity = maximum exposure to agents.

Prefer miners/royalties over bullion only when in-ground ounces (reserves less extraction cost) trade at an unusually wide discount to spot; ground ounces should always carry a risk premium, and when that premium isn't paid, skew back to bullion.

Demand stays loyal to an intermediary because people default to the familiar path, not because it owns an irreplaceable capability; an agent that compares every option every time breaks the habit.

Higher rates gradually lift the investment income on an insurer's float, but P&C investors care most about premium pricing; if rates rise while pricing falls, the stocks usually come under pressure.

Value a large drug developer on the ratio of wins to losses across many scheduled late-stage readouts (and whether misses touch the core franchise) rather than on any single binary trial; Haymaker's AZN case: 'two wins and one loss… that ratio is the investment thesis.'

Hedge a long-held position with large embedded gains by shorting the same shares (in a margin account) instead of selling, avoiding the capital-gains tax until you lift the short; his example: 100 long-held NVIDIA shares hedged with a 50-100 share short.

Size gold as a portfolio hedge between a floor (<5% hedges nothing) and a ceiling (>15% is a directional gold bet); trim back into the band mechanically when a gold run breaches the top, with no price target on gold.

Around $110 Brent the White House talks prices down (one Truth Social post can take $15/bbl off), so downside volatility rises even in a fundamentally tight market

No credit risk in agency MBS. The spread over Treasuries pays for the short prepayment option: upside capped, downside open.

Analysts work backwards — targets rise as a stock rises and fall as it falls (Salesforce chart), so a target is a lagging read of sentiment, not an independent input.

Win $1 / lose $1 is zero convexity; make 2 / lose 1 is positive; lose 3 / make 2 is negative. Negatively convex bonds must pay extra yield over Treasuries.

Average the market in 10-year chunks by year-digit; it bottoms in early October and years ending in 7 start bullish (partly double-counting the 4-year cycle).

When a regime marker moves (a 5% 10-year), don't flip the allocation: raise cash specifically to write more covered calls and tighten stop losses, without turning bearish or going to cash.

ETFs sit between open-end and closed-end funds. When illiquid underlyings are sold below NAV to meet redemptions, the fund is harmed, a risk in a volatility spike.

Overlay today's gold path on the 1976–80 bull market — a 94% correlation coefficient, including the mid-cycle correction, would imply gold more than doubling within two years; a scenario, not a forecast ("not a 2011 bull market").

Before trusting a growth plan, compare old guidance (production, capex) with what was delivered; be wary of "more growth with less capex" decks.

When shorting (e.g. euro assets), pair it with a long sharing the same risk factors but with a structural edge (the UK), not an unrelated bet (platinum vs a short French OAT is two bets, not a hedge).

High-yield bonds trade like stocks and draw on the same liquidity; a bearish divergence in their A-D line warns first, and a deeply oversold oscillator on that line precedes the final price low.

Futures-based 2x is linear and holdable. Daily-percentage-reset ETFs suffer volatility drag (+20% then -20% leaves 96) and are for a day or a week.

In a commodity business, a price that goes down and stays down bankrupts leveraged producers even when the mines are sound. So carry net cash rather than utility-style leverage.

A secular build-out is measured in decades, not years — a 2–3 year hold is short-term against a decade-long $1T spending cycle.

Breadth-momentum tools built from daily advances minus declines, developed by Sherman McClellan and his wife in 1969 by hand on ledgers; a higher low in the oscillator at a price low marks the washout.

Miners in safe jurisdictions (Canada, US, Australia) now outperform; Barrick lagged while Gold Fields was rewarded for rotating away from South Africa/Ghana; safe ground also raises takeover odds.

After a new 3-year high in the NYSE advance-decline line, the worst drawdown over the next 3 months is typically ~10% (average ~4%); exceptions are black swans (COVID, abrupt ends to QE).

Selling one-month calls ~10% out of the money on only about half an income fund's holdings lifts a ~1–1.5% natural yield toward ~5% paid monthly while leaving most of the growth uncapped.

Reality oscillates between "pretty good" and "just okay"; a 30% price drop rarely means the business is worth 30% less.

Chop the index into 4-year chunks starting Nov 1, rebase each to the same start and average them; years 1–2 are sideways, year 3 strong. Judge the fit by the timing of turns (the "dance steps"), not the slope.

EPS changes slowly (8–12%/yr for a quality business); the multiple swings like a pendulum — Rollins traded 31x–89x over ten years on steady ~12% EPS growth.

Divide production by diluted shares over time. Absolute growth bought with new stock is not value creation: 'anybody can issue more shares and get bigger.' Agnico: ~14x absolute vs ~3x per share over 20 years.

Low-coupon mortgages prepay away and new higher-coupon production moves the index toward par, where the embedded option is most convex.

If the underlying assets don't earn the payout, you get your own money back, a self-liquidating trade. Tells: 10%+ yields and a steadily falling price.

Margin debt/GDP peaks recur roughly every seven years and shortly before major stock tops; a record level before the projected peak date is a warning, not a signal.

Build a large mine in phases: spend, get your money back, spend more. It may cost more in total, but it sharply reduces the capital at risk.

Standing limit bids well below market fill far more often when volatility is abnormally high, catching tired or forced sellers.

Of four junior-miner stages (prediscovery, pre-resource, resource builder, pre-producer), he favours the ~18 months from construction decision to first pour — "a 90% chance of a 90% return" per history.

Give a catalyst trade a deadline (Terranova uses 90 days); if the stock hasn't moved by then, exit even without a loss — capital sitting flat is the thesis failing slowly.

When the curve flattens or inverts, the prepayment option gains value and MBS become more negatively convex, so spreads widen. This matters more than the level of volatility.

Every time the Shiller CAPE has been 35 or higher, the forward 10-year real return has been negative (most commonly about −5%/yr).

Value a producer at three commodity prices (e.g. $5/$6/$7 copper), with buyback and capex-taper cases and a fixed discount rate. Read the bear/base/bull spread against spot to judge margin of safety.

A jump in correlation to a benchmark signals the market re-categorizing an asset; old trading ranges become a weak guide.

The tax withheld on dividends depends on where a company is incorporated, not where it lists. Lukacs cites 0-15% on UK companies vs 20% South Africa and 30% US for him, so a UK redomicile is a plus.

Components with multi-year lead times (reactor pressure vessels, turbines) gate a build-out; suppliers with sold-out order books capture pricing power, but the market may already price it (Siemens Energy / GE Vernova at 70–80x).

Hotel REITs are the riskiest property type because every room reprices daily; long-lease sectors reprice slowly.

Chart pattern of three contracting impulses (a volatility squeeze); a trigger candle, rest at the first target, midpoint retest, then a fast move through open space to the full target.

Compare where a bond trades with the spread typical of its rating and the notches below; a multi-notch gap means the market rejects the rating (SpaceX BBB- trading ~3 notches lower).

Buy life insurance and annuities only from mutual companies owned by policyholders; PE-owned insurers load up on sponsor private credit and reinsure offshore.

Peterffy: prices alone are a narrow niche; companies sit inside regional, national, global and social/climate environments, and event contracts price those questions, becoming more accurate than polls or experts as hedgers join.

Private-credit firms and their captive insurers shop across 7–8 small agencies for the highest rating to win better capital treatment and more leverage.

Yield divided by duration ≈ how many points rates can rise in a year before price losses erase the income; Gundlach's model portfolio runs ~6.25% / 2.

Miners with sunk capital in a region must stay on good terms with local communities and governments, which can shut operations down; so safety and sustainability belong in management incentives.

Split REITs into core compounders, income & value, and contrarian/special situations (about 60/20/20). No more than 20-25% in one property group, 8-12% per name, and never pick on yield alone.

When a stock re-rates on big news but the same issuer's bonds sell off, trust the credit (Oracle/CoreWeave 2026, housing 2006).

Cited to reject cash-burning, debt-funded AI infrastructure names.

Most insiders sell on schedules, so executives buying with their own money signals confidence; cited for Uber's C-suite purchases.

A listed company where one holder has >50% of votes (e.g. SoftBank at SB Energy) can skip majority-independent board and independent committee rules, limiting minority shareholders.

Rank two beneficiaries of the same spend by gross margin: high-margin design IP (Marvell) over low-margin EMS (Celestica).

When physical-market insiders (buyers, sellers, traders, utilities) are constructive while investor sentiment reads depression, treat the gap as an opportunity; trim at euphoric readings, add at depression.

Buy deposits with scale, high grade, near-surface geometry near infrastructure in a pro-mining jurisdiction; with few such assets outside majors' hands, hold for the takeover.

Buy and hold companies riding a country's long-run growth (population, consumption, infrastructure) rather than betting against it.

A supplier whose customers can supply themselves is the first capacity cancelled when demand slows; worse when it is debt-funded.

About 90% of single-company risk is diversified by ~7 holdings; beyond that, diversify the factors that can hurt holdings (20 regional banks are one rate bet).

Annualized stock total return minus bond total return over rolling 76-month windows; prior peaks (1939, 1956, 1969, 1981, 1999) preceded poor stock returns, and the 2026 reading (~28%) is the highest on record.

Be patient most of the time, but act fast and big when a rare mispricing appears (e.g. Tata Motors in the 2020 crash).

Start from 'what don't I want to own if I'm wrong?': low valuation relative to growth, high gross and operating margins, positive operating and ideally free cash flow.

AI-lab spend must come out of existing budgets; only these three software categories look immune.

Rick Rieder: once a 3-year-duration, single-A portfolio yields 7%+, endowments and pensions can meet their return budget without relying on earnings growth, so equities get a lower grade.

Skip the quarterly numbers: capital in the ground ÷ useful life + running costs + cost of capital = revenue required per year; compare with actual end-customer revenue (e.g. $3T AI capex needs ~$1T/yr).

For a data-center/property developer: annual return on build cost before debt (~8-11% for data centers) minus borrowing cost (~6-7%); leverage magnifies the spread, and overruns or pricier debt erase it.

Yield curve flattens while yields fall, long end falling most — e.g. fall 2023 when Treasury cut long issuance.

Buy a lower strike, sell an upper strike at the realistic stopping point (not the tail), and expire it just after the political event; aim for ~10× premium if the target is hit.

Anchor to prior cycles (1974–81, 1999–2008); a cycle lasts longer when new supply must come from regions without infrastructure. Schachter dates the current one from 2020 to about 2034, 'the fifth hole of the golf course.'

When oil and yields keep rising because a rising market lets the policymaker hold out and eases financial conditions, strength feeds the squeeze until equities drop (~10% before Trump 'tacos').

When an originator books most profit selling loans above par, compare its average sale price with the disclosed buyer's price; a gap implies an undisclosed buyer paying above market (Carvana 109-110 vs Ally 102-104).

Colin's inventory of the 1970s reset (junk bonds and LBOs, the Bloomberg terminal, NYSE May Day and the London Big Bang, passive investing) as a template for spotting today's financial reinvention.

Track the share of outstanding mortgage balances above a rate threshold: turnover replaces low-rate pandemic loans even without refinancing, reviving the prepayment option in MBS (ICE: 5%+ loans ~10% of balances end-2022 to 40%+ July 2026).

Platform funds (Citadel, Millennium, Point72) run beta- and factor-matched long/short books at ~6-9% vol, levered ~5:1; they take 3-5% spreads and cover losing shorts within weeks, crowding fundamental short sellers.

Homeowners can prepay, so MBS lives shorten when rates fall (principal reinvested lower) and lengthen when rates rise (price falls harder) - more yield than Treasurys, but you lose more when rates rise and make less when they fall.

Waves of IPOs and issuance (1929, 2000) absorb the buying power propping up prices; watch issuance against mechanical passive inflows.

If private secondary shares trade far below the proposed IPO valuation weeks before listing, insiders who could wait are choosing to sell.

A fund wants to write a ~$50M cheque without crossing the insider-ownership threshold, which sets a minimum market cap; companies below it are shut out of generalist flows.

Classify a position as a trade or investment; exit when the view is fully priced even if the thesis is unchanged, rotating to a ranked 'plan B' idea.

The Fed controls demand (QE); the Treasury controls supply and maturity mix at quarterly refundings — the Oct-2023 tilt to bills drove a 100bp 10-year rally and a bull flattener.

How far a bond's price can fall before its income is wiped out; 60bp in 2021 vs ~500bp at a 5% 10-year.

A block model estimates metal grade between drill holes; smearing over-weights a few metres of grade across a large rock volume so an uneconomic deposit looks economic on paper. Overlaying the model on what is actually mined exposed the Asanko Gold fraud.

Cost of equity from the current 10-year yield plus beta × equity risk premium (4% here), free cash flow to the firm to terminal value less net debt, with an added 2% small-cap penalty to the discount rate rather than an ad-hoc haircut.

Bank regulators' 300% supervisory benchmark for commercial real estate concentration; used to screen regional banks exposed to SOFR-priced refinancing stress.

Subtract a neutral value for the operating business from the listed company's market cap and divide by its private stake to find the valuation the market assigns the private company; exit when it reaches the most optimistic credible mark.

A take-private by a controlling owner needs a majority of the independent (non-controller) shares; activists need only enough of that minority to block — but the protection does not cover the controller's other actions such as acquisitions.

A fiscal dominance, B cut, C grow / run it hot, D default via debasement, E political realignment and war — a staged map of how a Treasury supply-demand imbalance resolves.

Since the Bre-X fraud, Canadian-listed miners must log and disclose their drill holes (the NI 43-101 regime), so a resource can be independently modelled from public data.

ROIC = NOPAT / invested capital (all capital; can mislead with lots of cash or goodwill); ROE = net income / equity (can be inflated by leverage); ROCE = EBIT / capital employed (the capital actually in use; pre-tax, good for comparing across sectors and countries).

The S&P 600 requires positive earnings for inclusion (~20% unprofitable members) while the Russell 2000 does not (40-45% unprofitable, 30-45% floating-rate debt), making the Russell far more rate-sensitive.

5-yr revenue CAGR >5%, 5-yr diluted EPS CAGR >7%, 5-yr gross margin >40%, 5-yr net margin >10%, 3-yr FCF margin >10%, CapEx/revenue <5%, 3-yr ROIC >15%, 3-yr ROCE >20%; banks and China excluded. Cuts 50,000+ listed companies to fewer than 100.

Skip what the market already prices well (high-grade open pits, flashy drill holes); buy decent assets with modest capex and a credible permit path, after ruling out kill factors such as a unique salmon fishery or a jurisdiction where the mine can be expropriated.

Procyclical producers reinvest peak cash flow at peak prices into higher break-evens, creating the next glut. Favor capital-light royalties over producers.

Authorized participants create ETF shares to lend to short sellers (price down + shares up, ARKK 2022); price up + shares down + short interest up means longs redeeming while shorts borrow from a shrinking float — a squeeze setup.

Revenue a retailer books from gift cards expected never to be redeemed; a non-recurring-quality contributor to a margin beat (flagged in Chewy's Q2 alongside tariff refunds and rebate timing).

A gold rally has historically been followed ~20 months later by rising oil; direction, not magnitude.

Buy a lower-strike call, sell two at the target, buy one higher: low cost, capped loss, peak payoff at a chosen price by a date. CPM used one on gold before the 2016 election (125% in a month) and again for the 2026 midterms (peaks $5,000 / $5,400).

Tangible asset, Oligopoly, Low incremental capital intensity, Long-duration cash flow. Davolos uses it to explain Horizon Kinetics' real-asset capital-light thesis.

A rally that speculators distrust — light positioning, shrunken open interest, crowd fading it — which tends to keep climbing.

Dominating a small category can be immaterial to the company: ~a quarter of foldables (~2% of smartphones) is ~10M Duos against 200M+ iPhones a year.

Companies stop repurchasing shares in the window before earnings; when mega-cap tech buybacks are the main support under the index, entering the blackout removes that bid.

Values a stock by its future dividends; Green says some of the 10-25 largest US stocks come out at ~1/15th of price, a gauge of overvaluation rather than a forecast.

Inflation linkage and a clean credit record don't protect price: TIPS were crushed, Austria's century bond fell ~80%, and Argentina's defaulted century bond did better on its coupon.

A 3x ETF must buy about 60% of its equity after a 10% rise (and sell after a fall), creating self-reinforcing flow; with volatility drag, ±10% at 3x loses about 8%.

The feared mismatch (liquid wrapper, illiquid credit) instead accelerated electronic, portfolio and systematic bond trading, making credit more liquid and more equity-like.

Trading businesses borrow against liquid inventories and receivables, so headline leverage misleads; Glencore's ~$45bn gross debt is ~$10bn net.

A bond coupon comes on top of getting principal back; many 'income' products (covered-call funds, Saylor's 'bitcoin yield') hand back your own capital — 'dressing up a fox to look like a goose.'

Value a premium product on (new price - the substitute the buyer would have bought) x units, not units x price. Apple's iPhone Duo: ~$20B gross but ~$7B incremental ($700 over a $1,299 Pro Max x 10M).

When index buyers must buy a new listing before insiders can sell (CRSP SPAC fast-track: 5 days vs a 20-day lockup, closed Sept 2022), price must rise; Green says fast-track IPOs like SpaceX revive it.

$1 of flow into equities moved aggregate market cap by ~$5 (1992-2019), not the ~1 cent efficient-market theory implies; Green estimates ~$22 today and near $100 for the largest stocks.

Investing as guessing what other people will want to buy; Green calls the ESG wave a variant.

When the seller is forced or indifferent (e.g. Nebius shares landing with ex-Russian-exposure holders) and you hold a materially different view, the setup can be life-changing.

The share of insurance premiums paid straight back out as medical claims - the single ratio that governs a health insurer's earnings, and the one whose repair ('under control') Simpson required before re-entering UNH.

Brokers like Robinhood sell customer orders to market makers such as Citadel in exchange for zero commissions; 'if you're not paying for the product you are the product.'

A drop from ~70% to the low 30s shows a correction well advanced beneath a still-strong cap-weighted index — near levels seen at intermediate-term bottoms.

Markets tend to bottom just before the November midterm election as political uncertainty clears, rally into roughly November of the cycle's third year, then wobble in year four.

Signed-but-unbilled future revenue - Oracle's rose $30B to roughly $700B. The usable test is fractional: ask what share of the backlog must convert for the valuation to work, not whether all of it will.

His 'out there' view: one-day liquidity is bad for managers, investors and markets, and semi-liquid wrappers on illiquid private loans are 'dumb.'

Which sector dominates new lows (consumer discretionary, 26% of members) versus new highs (energy) identifies the macro shock hitting the economy.

Green's version of Grossman-Stiglitz: uninformed traders are random noise traders or predictable 'sunshine' traders (passive/401k); informed traders either correct prices or facilitate the crowd (market makers buying order flow).

Re-enter a stock you were stopped out of on the turn in the fundamental series, not the price: 'the earnings are closer to a trough than a peak,' evidenced by analysts guiding up rather than down. Accepts missing the low by design.

Teams are trophy assets priced at multiples of revenue (Angels ~10x ~$400M), not earnings; apply the multiple, add real estate, subtract net debt to get private-market value per share.

A physical trust like SPUT issues new units only when it benefits existing holders and turns each raise into physical purchases, capped by an annual limit (9M lb); the raise-and-buy pace doubles as a demand gauge.

For a bank interest is income and deposits are raw material, so the standard tests invert; Nubank scored on LDR 58%, CET1 20% vs ~8.75% required, 90+ NPLs 6.8%, ROE 31.6%.

Fundamentals (inflation, expectations, growth) set yields long run; true-or-false narratives can divorce yields from them short run, then yields revert.

Regression of starting Shiller CAPE against the next decade's real stock returns (1965-2015); at a CAPE of 42 the 10-year real return has never been positive, typically -5% to -9% a year.

Lower is better; Nubank's branchless model runs ~27% vs 40-50% at traditional banks (58.2% in Q2'22 to 27.7% in Q3'25).

When seasonality is in effect, gold typically makes its summer low between mid-July and mid-August. It is a prior, not a rule, and needs chart confirmation.

Pull the insider transaction history before the management meeting so the meeting cannot talk you out of it, and treat a sustained pattern of selling — not any single filing — as disqualifying. "This is something we do at Timelo quite a bit. We follow insider buying and selling very closely." It is why he passed on Propel and is avoiding Champion Iron.

The worst trailing 10-yr bond-return decades have preceded strong forward returns: the mirror of high CAPE predicting low equity returns.

Take profits into fevered rallies only from the investment positions. The insurance holdings stay untouched whatever the price does.

At ~5%, long bonds pay you to wait for a stock-market break (which brings cuts and bond gains), unlike a put option that bleeds premium.

An over-funded pension locks in future payouts by selling equities and buying long bonds/TIPS matching its liabilities; US corporate plans over 100% funded create a slow structural bid for Treasuries.

Divide price by the 200-week average; ~40% over it flags exposure to plain reversion, and two assets at the same premium (copper and the S&P) are one crowded trade.

Keep the equity book net long on fundamental ideas, then buy index puts sized so total net exposure falls below zero in a selloff. Cost is a small, known premium every month; the payoff is convex and grows the faster the market falls. Tardif buys puts into rallies and sells them into declines, the reverse of instinct.

Shorting delivers cash to the account which earns the prevailing short rate, so a short's real carry is cash rate minus dividend yield minus borrow fee — not just the dividend. Canada at ~2.5% cash vs a ~2.5% dividend is a wash; in the US at ~5%, "the stock stays flat, your net return is actually positive 5%."

An asset (gold, copper, Bitcoin) whose rolling 100-day correlation with the S&P sits near record highs will fall with stocks rather than hedge them; high correlation in a rising market is a warning, since correlations go to one in down markets.

A negotiated acquisition struck below what holders believe the assets are worth — the opposite of a takeover premium. Applied to the H&R REIT break-up, with the recurring Canadian pattern noted, plus the conflict when management or a founder's family buys assets from the company it runs.

Before buying an instrument as protection, chart its spread to what it hedges; the 30-year TIPS/nominal spread was flat for 5 years, so long TIPS took the same rate rise and hedge nothing.

Buy an actual 5-7-yr bond and hold to maturity: worst case you earn the yield; if yields collapse and stocks fall, sell at a gain and buy stocks on sale.

Cited as the reason he left trading: with VAR limits and risk-capital guidelines you are forced into a short lens, and "it's really hard to participate in a trend if you're going to get stopped out three times before you arrive." The point is not that the view is wrong but that position size and stop discipline can make a correct long-horizon view unmonetisable.

Asset volatility ÷ S&P volatility; a haven running at 2× stock-market volatility is an oxymoron, and gold vol surging vs a record-calm S&P preceded 2007-08.

A put caps the loss at the premium and cannot be squeezed out before the thesis plays out, unlike a short with unbounded loss. Young stopped shorting after an early short squeeze and expresses bearish views this way - accepting the premium as the cost of surviving to be right.

The objective switch to make when the outcomes ahead are binary and un-handicappable: stop scoring the portfolio on return and score it on drawdown, for a defined period. "It's not how much you're going to make, it's how much you don't lose."

With a 2% haircut a $1,000 bond supports $980 of borrowing that can be re-pledged, implying ~50× leverage on collateral; ~77% of global lending is collateralized.

Sell into consensus rather than into a number: backwardation says the market disbelieves the tightness (thesis unfinished), while a flip to contango marks the crowd turning bullish one to three years out - Young's stated trigger to return capital.

Decompose a 'price has been high enough' claim into consecutive runs above the threshold. Ninety non-consecutive days above $75 sanctions nothing; large projects need the price to hold, not to visit.

Before the return test, ask whether you could defend the investment to clients after a seizure. Count prior expropriations as a base rate and check who was left in charge - removing a leader is not a regime change.

State the honest base case, then take exposure that pays if the opposite happens and is still defensible if it does not - so arguing against a scenario costs you nothing. Young doubts the Venezuela rig migration yet profits if it occurs.

Count the specialist funds and analysts covering a sector and track the trend: ~150 oil-and-gas public-equity firms fifteen years ago to fewer than five now. The capital-markets exit and the physical underinvestment are the same signal.

Enter credit at an extreme wide spread, expect compression, and pre-commit the exit to a spread level rather than a date: when junk yields near Treasuries you are no longer paid for the credit risk, so you hold Treasury-like yield with equity-like downside.

Look past the executive team to the register: a family or strategic holder with a stake large enough that their own money disciplines decisions (Lundin at International Petroleum and 20% of Talon, Beaty at Equinox, Goodman at Dundee). Its limit: supervision does not transfer across a change of business model.

Pick a fixed date - 'your birthday or your cat's birthday' - and never optimize the interval: 'let's say I figured out you had to rebalance it once every 267 days. Well, that makes things really complicated.' Skipping it breaks the portfolio; gold's 60% 2025 move is the cautionary case.

When lenders aggressively buy credit default swaps on their own marquee clients, treat it as the lenders' own verdict on the loan book - the insider signal that the underwriting assumption is slipping.

A collapsed price is an asset to a company retiring its own stock: at a high cash-flow yield each dollar of repurchase cancels more shares, so the same buyback becomes more accretive as the stock falls. Measure the actual annual share-count reduction (not the authorization), net it against stock-based comp, check cash and debt capacity — and separate real debt from lease obligations. A flat business shrinking its float 10% a year still compounds EPS at 10%.

"When people buy financial goods, they seem to want to pay more. When people buy physical goods, they seem to want to pay less." Rule's test for whether you are genuinely a net buyer: a falling price should make you acquisitive, not anxious.

Cash judged on yield always loses. Priced instead as an option to buy something cheaper later - its value peaks exactly when everything else is marked down - plus it removes forced selling for life events and dampens portfolio volatility.

Buying a poor business far below the liquidation value of its own assets, so the asset base is the floor and business quality is a bonus. Slegers reconstructs Buffett's Berkshire entry as the canonical case: $14.86 paid against $32.30 of assets (0.46x book), and $20.3 a share of cumulative textile operating profit returned over 1965-1974 — more than the purchase price, from a business Buffett called his dumbest ever.

A long bond at 88 on a 6-1/8 coupon goes to 120-130 if the 30-year returns to 3-3.5%, while one more punch takes it only to ~85 - the asymmetry, not the direction, is the entry condition. 'Like a rubber band that's pulled.'

When every member of a group is weak for years - McDonald's, Shake Shack, Wendy's, Domino's, Chipotle - a separate story for each is usually several stories too many; look for a shared demand-side cause and then confirm it by finding the beneficiary whose chart is the mirror image.

When a multi-year theme has a large embedded gain and an intact driver, buy small dated protection on an uncorrelated, crowded part of the market rather than trimming the position - 'now you want to protect those gains.'

Why rising long yields threaten the AI build-out through the funding market rather than the discount rate: past a point the yield becomes unattractive to the buyer, the hyperscalers stop issuing debt, and the capex slows. Converts a macro worry into an observable corporate action - the issuance calendar.

Yield-oriented securities that have historically raised their payouts at least in line with inflation, so the real return is not eroded by CPI increases the way a fixed-coupon bond's is. Hay's category for names like MPLX.

Rule's jurisdiction framework: there is no safe address, so name the specific mechanism of harm (an 80% Polish profit tax, cartel violence, a 13-year California permitting delay), then set it against a prize you have checked for substitutability.

'I hate log charts with a burning passion' - log scaling makes a 50% fall look like 'a little blip... oh, I could ride that out. It's insanity.' Re-plot history linearly, and in your own dollars, before setting a risk budget.

For a razor-and-blades hardware business, the useful KPI is new units installed this period, not the cumulative installed base: cumulative tells you about the past, while each new install begins a stream of single-use consumables and service revenue running for years. Pair it with consumable revenue per unit per year to see how much of future "growth" is already contracted by machines in the field.

Chess has fixed rules; markets do not, 'mostly in the form of correlation.' Gold's current negative correlation to oil 'started when the war started' and will break, leaving positioned investors 'caught totally off sides.'

"Silver or lead" — the Mexican phrase Rule uses for jurisdictions where a non-state actor, not the government, controls the ground and negotiates by force. Cited on Sinaloa, where he says a cartel faction kidnapped and murdered ten of a mining company's employees.

Not "cheap versus its own history" but cheap versus the multiple the market reserves for businesses with no pricing power. A low forward PE built on shrinking EPS, beneath commodity/auto/airline multiples, means the price no longer discounts slower growth but value destruction — the threshold worth stopping at. Cross-check with free cash flow yield so one accounting-sensitive multiple can't mislead.

When a producer's upside comes from a re-rating rather than from operations, the thesis has an end and the position should end with it: 'these are not long-term investments, producers. They are timing mechanisms.'

Restate the entry multiple as time: purchase price divided by annual free cash flow is the number of years to get your money back, after which the business's own cash flow becomes new investable capital every year. Slegers' illustration: a $5m business throwing off $1m recoups in 5 years and turns $54.2m into $171.0m over 25 years — the difference between compounding at 10% and 20%.

Inverse and levered ETFs are rebalanced daily, so decay is a recurring cost: attach an expiry window to the position at purchase (here, September-October), and switch to long-dated options if the view must be held longer.

The default is to pick a return target and accept whatever volatility it requires. Dillian inverts it: 'think about what risks they want and then back out the returns. It should work in the opposite direction' - and score candidate portfolios on Sharpe, not on absolute return.

Banks stretching outside their mandate into speculative lending to keep marquee clients (SpaceX, OpenAI, the Mag 7) is how the exposure gets built before anything defaults.

Haymaker's stated rule, and the one the ACN post-mortem admits was ignored when the stock decisively broke below $250: a decisive break below a level that has held for years carries the same weight as an upside range expansion, and should veto adding as much as it triggers selling.

After trimming at a gain, the level at which repurchasing leaves the taxable holder whole is the real buy price, not the analyst target - he needed sub-$196 on Qualcomm to offset a half-position trim at ~$230, with the stock at $176.

When a bear thesis is universally recited, measure the crowd instead of arguing: CFTC futures positioning plus bulls/bears surveys against the asset's own opposite extreme. Put a number on how much is discounted ('80 to 90% priced in') and then hunt the bull case nobody is working on.

A basket Santoli prefers to the Mag 7 as an AI proxy because it includes Micron and AMD in its top 10 and so trades like the theme rather than like the index; his tell is its distance from its record high and time since the last one.

20% each in stocks, bonds, gold, cash and real estate. Explicitly Browne's 25/25/25/25 with real estate added: 'the Sharpe ratio goes way up, the returns go up, the volatility comes down.' Backtested to 1 Jan 2026 - Sharpe 0.6 vs 0.7, standard deviation 8.22% vs 17.04%, worst year -11.8% (2022) vs the S&P's -36.55% (2008), ~9% annualized.

Trigger on the price of insurance, not a forecast: when vol is cheap against a dated catalyst, spend ~1% of the year's gains (1-3% of the book) on protection - long-dated puts for the thesis, short-dated vol only as a rental.

Named as a close cousin of the Awesome Portfolio - 'pretty similar to what I'm doing, but it's a little more Rube Goldbergy. There's a lot more moving parts... he's got the rich man version of it and I have the poor man version.'

Check investment grade (LQD vs the S&P) at the top, CCCs and the loan market at the bottom, and BDCs/private equity in the middle, using relative charts rather than absolute yields. Stress at both ends while equities sit at highs dates the cycle ('late 2006').

McDonald's standing dashboard of 2008-vintage stress gauges - CCC yields, investment-grade behaviour, the loan market, BDC and private-equity charts - used to set the house risk level rather than to pick names.

Haymaker's coinage for a security bought at roughly nine times earnings while yielding roughly 9% — the valuation half of his two-legged entry test (a breakout above a multi-year ceiling plus a cash valuation). Coined for MPLX in February 2024; 93.7% total return since, vs 61% for the S&P 500.

Decompose an incumbent's job into creation vs ongoing maintenance before believing AI will replace it. AI is strong at the first draft and weak at continuously updating, patching and adapting software — so where maintenance dominates the spend, an AI-disruption de-rating is a sentiment overshoot to fade, not a terminal repricing.

1980s-90s pattern: hot economy -> commodity-led inflation -> long yields spike and bond portfolios are destroyed -> the inflation wounds the consumer -> recession arrives fast. So you buy the asset the inflation spike is destroying, because the spike causes the recession that re-rates it.

Trade the direction of change, not the position: fire when two to three vertical specialists who do not know each other move the same way in a short window. Triangulated through ideas dinners, a live chat and bull-vs-bear 'cage matches'.

Lynch's bucket for a formerly great business the market has given up on: they are "some of the best opportunities in the market if you get them right." The whole category reduces to one question — is the impairment temporary (management, marketing, merchandising) or permanent (the customer left, the moat is gone)? Competitors imitating the product is evidence it is still wanted.

Brokers must try admitted carriers first; declined risks go to excess & surplus writers like Kinsale, and that business flows back to standard carriers when the market softens.

A payout structure with a fixed base dividend plus a variable top-up formula-linked to free cash flow (Weyerhaeuser: 3.5% base, variable targeting 75-80% of annual FCF). It converts an earnings inflection directly into cash returned, so the dividend — and the yield support — can rise sharply without a board decision to re-base.

The contrarian signal is not extreme sentiment but uniform sentiment — everyone saying the same version of the same thing at once, across S-1s, IPO pitches and earnings calls. Zeigler's 'biggest contrarian itch'.

Not what most people believe, but what everyone believes everyone else believes. A burst of stories only matters if it converts private opinion into this shared, self-referential state — which is when positioning and prices adjust.

Before arguing against a consensus, state it in the form its best proponents would accept and assume they are as smart as you and better informed. Then list what is genuinely different about the present case and what is the same, and size to both lists rather than picking a verdict.

Land without power infrastructure studies trades at $2,500-$10,000 an acre; power-ready sites have gone for ~$800,000 an acre in Texas and a median of $2.8m in Northern Virginia, against ~$28,000 for large unpowered assemblages. A ~100x spread on physically similar ground, where the variable is deliverable megawatts on a timeline a buyer can underwrite.

Accounting games are rare early in a boom, cluster at the crest and into the descent, then wash out with the credit issues. The appearance of the debate is therefore itself a cycle indicator — 'accounting is not something that people care about till the peak of the cycle'.

Read the stated reason for a guide-down word for word. "Customer spending timing", "lumpiness" and "pull-forward" are calendar language and usually reverse; "pricing pressure", "share loss" and "demand softness" are not. Then check the reason against the segment's structure — a high-margin business selling to a few large clients on short contracts will be lumpy by construction.

Buy a mispriced single stock and short a thematic ETF against it so you own only the relative bet — whether the company outperforms its own theme — rather than the theme's direction. Singh's test for when it applies: the company's results moved up while the price moved down (a beat plus a decline is a flow event; a miss plus a decline is information). Only works where the residual case is genuinely idiosyncratic.

Operating expenses plus commissions as a % of premium; P&C carriers run ~30%, so going direct to cut broker commissions buys little.

A regulated US event-contract exchange whose Fed-decision contracts give a money-weighted, continuously updated probability for the next FOMC outcome. Singh treats it as the clearing probability you are being paid to disagree with rather than as truth — reading 52% hike / 48% hold into the Sept-16 2026 meeting against the sell-side distribution and taking the side with the political constraint.

Two distinct failures routinely conflated in credit headlines: a redemption gate is contractual illiquidity (private credit never promised money on demand — "closer to full reserve banking" than a deposit), while solvency is whether the loans are impaired. "You can have two problems at the same time in different magnitudes."

Insurers know revenue (premium) but not cost of goods (claims); if claims inflate while premium is flat, margins erode, so the pricing cycle runs on its own rhythm, not the economy's.

The upgrade from word clouds and positive/negative word counting: track how a claim is constructed and positioned in a sentence, then score where it is appearing (the Journal, Reddit, cable) to tell a high absolute level from one that is suddenly screaming. Disaggregating the signature — central banks by country rather than as a class — is what made it map to prices.

Measure where a story sits in its life cycle rather than how loud it is: a burst when it arrives, a contested phase while people argue it, and a confirmed phase once it becomes common knowledge. The contested-to-confirmed transition is where the price move lives; once confirmed, the asymmetry inverts to the chance of falsification.

Locate the headline coupon against the risk-free curve to get the spread, then price the undisclosed terms: fees, 'minimum-return protections' (a floor on the lender's return, making prepayment expensive) and a warrant package with no fixed count or strike but a much longer tenor than prior issues. Benchmark against the borrower's own prior paper, and diligence the arranger's committed capital.

Inflate each cost input on its own driver — the FX assumption, statutory duty changes, and peer-observed capital inflation from PEA to pre-feasibility — and credit back genuine improvements, before running the study at spot prices. Then re-derive the economic cut-off grade: a lower cut-off widens the pit before anyone drills a hole. Label the output as your numbers, not the company's.

In a concentrated supply chain, one economic event is booked twice with opposite signs: pull the supplier's days sales outstanding against its customers' days payable outstanding, look for the mirror move, then normalise the customers' free cash flow as if payables were unchanged. Neither side looks wrong alone — the pair is what makes it visible.

Nathan's fade: for a scheduled, heavily-trailed product launch (not earnings), measure the anticipation rather than the product — how far the stock ran in, whether it closed a recent gap — and ask whether the buying is new information or "muscle memory" from an era when the events delivered. The call has an expiry date; it is not a view on the company.

When one company owns a large but non-controlling stake in another (roughly 20-50%), it records its share of that company's profit or loss in its own income statement. Microsoft's 25% of OpenAI is the live case: it is one of the few routes by which a private AI company's economics become publicly visible, and it forces progressively more disclosure over time.

A jump from 14 to 25 would mean "everything's going down at the same time" — index vol is a statement about correlation within the S&P 500. So a floor-level VIX alongside violent sector rotation is a composition artefact, and protection should be priced off the sector rather than the index.

Unlike an equity REIT collecting rent, a timberland REIT's earnings function as an option on single-family construction and lumber prices, because it both grows the wood and mills it. Weyerhaeuser's unit sensitivities: a $10/MBF lumber move adds ~$50M of annual EBITDA, $10/MSF of OSB adds ~$30M, and detached houses use ~3x the lumber per unit of apartments.

Learn a system by taking each control to both extremes: statement by statement (income, cash flow, balance sheet), ask what the most aggressive legal treatment of each line would look like, then read the filings for which direction the knob has moved since last year. Track the trend, not the level — a series of individually defensible choices in one direction is the pattern. Define your own absurdity limit in advance.

Vanguard measured that adding a third party telling clients to stop trading raised returns by about 3% a year. Dillian's rebuttal: an advisor changes conduct but not experience - 'if you take a 50% drawdown, you're still going to be stressed.'

Annual return tables split a crash across two rows and understate it: 2008 shows -38%, but the actual fall from summer 2007 to March 2009 was 57% - and 57% is the number that decides whether someone sells.

The top seven names are 35% of the index, so index buying is proportional top-seven buying and they get bigger. The only large-cap managers beating the index are those overweighting the same seven. Dillian credits Michael Green with the deeper work.

Assume compounded EPS growth and a fair exit P/E, solve the annual return from today's price; rank holdings and direct new money to the top.

Own large land tracts for the options stacked on them (timber, development, solar/wind and hunting leases, mineral rights), bought when the market prices only the depressed current use.

No asset class wins forever and leadership persists for decades: answer which regime is in force (paper assets vs hard assets leading) before choosing a security, and set exposure to that answer. The century's best investors were not the ones who went all in on one side.

Express silver as a % of the gold price instead of the gold/silver ratio: 2% floor, 3% target, 4% ceiling, re-run at a forward gold price.

Divide the S&P 500 by the gold price to strip inflation and currency debasement out of a century-long comparison — you measure stocks against hard money rather than a diluted dollar. Rising = capital into paper assets, risk and speculation; falling = rotation into hard assets and protection. A multi-decade regime gauge, explicitly not a timing signal.

Leg one: the metal rips and miners lag, so multiples stay low. Leg two: miners outperform the metal and generalists pile in (the easy leg). Leg three: a mania with rising risk.

Reinvest first where returns are decent; buy back only when shares are cheap (a company that doesn't buy back is signalling it isn't cheap); dividends with the rest. Measure payout as dividends ÷ free cash flow.

A single result decisive enough to settle an open question for an entire sector, not just the company reporting. Lebenthal names it as the desk's own term for NVIDIA's print — and the diagnostic is the lag: a read-through that arrives a week later and escapes the obvious neighbours (software, platforms) is evidence of a theme being resolved, whereas a simultaneous move is usually just index beta.

A reasonable P/E on over-earning (e.g. ~80% memory operating margins) hides a bubble in the earnings themselves; value on cycle-average margins.

Argued out live on the Adobe buyback. The label is a description, not a verdict; two facts decide it. Funding source: repurchases paid from free cash flow are capital allocation, repurchases paid with new debt are leverage dressed as returns. Opportunity cost: check whether R&D and sales spending held up alongside. Weiss adds the step most often skipped — a 10% cut in share count flatters EPS growth by ~10%, so strip it out and ask what the underlying growth was.

With demand elasticity ~0.4, a 10% price rise costs ~4% of volume, so price increases more than offset a slow unit decline and profits still grow.

When the pricing anchor breaks (gold could be priced off TIPS until the day Russia invaded Ukraine), do not patch it — substitute the two inputs a portfolio actually needs: an expected long-run real return and a correlation. For gold: a 150-year ~0.6%/yr real return lifted to ~1% by BRICS official buying, paired with a zero equity correlation.

When the average analyst target sits essentially at the market price, the consensus model contains no credit for the contested asset — it prices only the business analysts can forecast. Splitting the target list into the cluster near the average and the cluster well above it shows what the high group is adding, and the gap between them is the market's quoted price for that option. Haymaker: the $127.13 INCY consensus 'is essentially at the current price… a Jakafi-plus-Opzelura base case with limited KRAS credit', while Leerink $155 / Canaccord $152 / H.C. Wainwright $150 'are the analysts adding in KRAS optionality'.

The end investor funds real-world liabilities — retirement costs, healthcare costs — so the objective is a real return. Most investors have drifted to nominal, asset-class-relative benchmarks (beat MSCI World, beat the bond aggregate) rather than beating inflation; the repositioning onto real targets has not happened yet.

A country-level consensus label conflates independent assets. Split it: the equity leg and the currency leg can resolve in opposite directions, and each is expressed in a different instrument — an allocation weight for equities, a hedge ratio for the currency, which binds only non-domestic investors.

Test any long-run equity return input by ranking the universe at the START of the period, not the end. Rank markets by 1899 market cap: the US works spectacularly and the UK reasonably well, but the next six or seven went to zero, sometimes more than once. Passive cap-weighted index returns are not as high as the surviving sample implies.

The negative stock/bond correlation that made 60/40 work is a ~20-year anomaly; extend the chart 200 years earlier and the correlation was positive almost all the time. The 100-year average is about +0.2, versus the -0.4 investors grew used to — bonds still help, but they are no longer the no-brainer diversifier.

A portfolio bucket named for the property being bought rather than the asset: exposure that is nobody's liability. Gold dominates it; bitcoin and silver enter as small satellites, silver on structural difference (a proportionally smaller investor base) rather than on conviction.

The label is used because it makes the maths easier and helps people sleep, but a government bond's safety is contingent on political and economic states of the world. Instead of a label, enumerate the roles a holding performs — diversification, liquidity, drawdown mitigation, cash-flow matching — and re-underwrite each one separately.

Issuing ordinary assets — shares, funds — as blockchain tokens so they can be traded in fractions and around the clock. Cited by Simpson as the business line that should decay Robinhood's tight correlation to bitcoin over time: a brokerage service rather than a crypto bet.

Before sizing a position around a binary event, ask what is left if it fails. A multi-franchise platform can absorb a failed programme; a single-asset company cannot, however good the data looks. The test is whether the base business alone justifies today's price — if it does, the catalyst is free optionality; if it doesn't, you are paying for the catalyst whatever the write-up says. Haymaker on INCY: 'if it does not [hold], Incyte absorbs it from a diversified multi-franchise platform rather than a binary single-asset position.'

Where value sits in an irreplaceable, slowly-appreciating asset and earnings are cyclical, price the asset from per-unit transaction comps and treat the trough earnings as a floor case — otherwise the multiple rejects the stock at exactly the point in the cycle you want it. State explicitly that a gross asset value is not an equity NAV.

Cisco was 'the Nvidia of its day' at over 100x earnings and only recently regained its 2000 market cap; the NASDAQ fell 78% from March 2000 to October 2002 while the internet itself kept doubling every year. Being right on the technology gives no protection on the multiple.

After a violent breakout above a multi-year ceiling, judge the retracement by where it terminates, not by how deep it is: a correction that stops at the old resistance level (now support) marks a regime change that stuck, while one that slices back through marks a failed, event-driven spike. Hay applies it to the Bloomberg Spot Commodity Index's 2022 breakout, whose drawdown ended exactly at the ~500 shelf that capped the index in 2008 and 2011.

A CDS is the annual premium to insure a company's debt. Because a lender's upside is capped at repayment while equity's is not, credit investors do more work and move first — so watch the big borrowers' CDS against the North American investment-grade average every morning. It was the GFC's leading tell and is now flagging AI capex financing.

Match funding tenor to asset life: AI compute is obsolete in two or three years, so ten- and thirty-year debt commits the borrower long after the asset stops earning. The dot-com build was all equity; this one is levered, and 'the leverage is what gets people into trouble.' Compute the effective coupon (~6%), not the spread (60-80bp).

Live in one country, hold citizenship in another, bank in a third, invest or operate a business in a fourth, keep assets in a fifth. Polomny quotes it in full then rejects it on implementability for a normal reader — advisor access, administration burden and fraud risk.

A forced liquidation transfers ownership from a seller with no choice to buyers who chose to be there, removing the overhang. The caveat is sequencing: the most leveraged player goes first, not last — Bear Stearns in March 2008 was followed by a 17% rally and then Lehman.

Because nobody can reliably predict the next economic regime, own one asset per regime: 25% broad US stocks (prosperity), 25% long Treasuries (deflation/falling rates), 25% T-bills or cash (recession, tight money, liquidity), 25% gold (inflation, currency depreciation, monetary disorder); rebalance annually or on ~35%/15% tolerance bands. Polomny adopts it as the chassis for the new AIA Permanent Portfolio, with less bond exposure.

Before adopting standard advice, score it against real capital, advisor access, time and administrative tolerance, and name the capital threshold at which it flips from sensible to absurd. Keep the objective, discard the implementation.

On a genuinely new theme, take a starter position the first time you hear it and research afterwards — because the alternative ('I'll get to that later') means you never do, 'and then the stock's up 200% and they miss the whole trade.' Size it as a research option. Produced Kite Pharma (2016) and an early GLP-1 position.

A fund at four times leverage is liquidated on a 25% fall in its holdings regardless of whether the positions were right. Finding the leveraged forced seller dates the bottom better than valuation does: once the mechanical supply is gone, so is the pressure.

A listed company that exists to acquire cash-flowing businesses whole or in part, recycling their cash into more of the same, with growth in net book value per share as the stated goal — buying back its own stock whenever it trades below NAV. Aimia is the worked example; Teledyne the historical one.

Public markets have liquidity so they reprice in weeks; private markets do not, so the symptom is assets held rather than sold ('portfolio companies not being sold for a really long time'). Separate the direction of the call from its timetable, and express it where it can actually reprice.

When gross shorts are cut by risk committees rather than by conviction, the resulting rally carries no bullish capital: 'the only thing that's really affected the price has been what your risk team is allowing you to short.' A price that rose without any bulls still has that entire buyer base ahead of it.

Sector return on capital rose with oil from $20 to $60, then flattened from $60 to $100 as capex and cost inflation ate the price - profitability peaked in 2007 and by 2012 returns at $100 oil were 'zero difference' from $20 oil in 2004. So the commodity call and the equity overweight are separate questions; getting the first right does not carry the second.

In a momentum leader the multiple carries no information; the tell is the growth rate itself stepping down — 70% to 60% to 50%. Not the first disappointing print, the first decelerating one.

Since 1957 September is the only month of the year down on average and the only one more likely down than up; in midterm years the peak-to-trough loss from July 31st to November 9th has averaged 10% since 1990 versus 5% otherwise. Stack independent calendar odds rather than arguing a single factor.

Hold a view with full conviction while it fits the facts, and change it the moment they change. Niles' counter to permanent positioning — offered alongside his own live loser rather than a museum example.

A Lehman-era ritual: pull the top 50 (or all 500) index charts by hand, classify each as topping or basing, roll up to sector level, and read the balance rather than the best name — 'a lot more charts that are rolling over than charts that are basing.'

Ignore headline prices per unit; convert both sides into a multiple of the cash flow acquired or disposed. Buying at 21x while selling at 45x is value creation by arithmetic — conditional on management's forward estimates being honest.

Reduce a hyped revenue line to price per unit and units produced, and track them separately: since end-May AI token prices fell ~50% (open source) while tokens produced rose 2.5x, so revenue still expands. The trade lives only while volume outruns price.

The premise that US equities structurally out-earn and out-return the rest of the world, justifying a permanent home-country overweight. Hay's test: it is falsified not by valuation but by relative performance under maximum support — MSCI US has lagged MSCI All Country ex-US since early 2025 while absorbing record inflows.

Separate a price fall caused by the business from one caused by market plumbing — index additions and deletions, merger-driven reconstitution, forced fund selling, tax-loss selling. Only the second is an opportunity, and only inside a bull framework.

The 2005 housing bust first reached creditors' balance sheets when ABCP came unglued in summer 2007; the funding market breaks before the asset market admits it.

Restates this year's foreign-currency results at last year's exchange rates so growth is comparable without FX swings; used for Amadeus's +5% H1 revenue.

Anchor the three cases to management's free-cash-flow guidance range, starve growth and assume fewer buybacks than history; if even the bear case shows upside, the margin of safety survives an under-delivering management.

Anchor the book in a few large, diversified core holdings (the hub) and arrange individual bets around them (the spokes). Two questions set the weight: conviction in the name, and whether you have any relationship with the company.

When an issuer raises cash by selling a large slab of new shares to one buyer rather than into the open market, the buyer gets a discount to the market price as payment for absorbing the whole line at once. Abel's account is a complete worked example: the call came with "no terms or amount were set", and Berkshire named both variables itself - "10 billion and more" at a "6.5% discount" - which the issuer accepted. The size is what creates the leverage to name the discount, so bidding small forfeits the edge.

Doomberg's three-part test before trusting any market's price. Purpose: what does this market exist to do (oil = ensure steady crude supply to refineries at a spread; equities = assemble risk capital)? Structure: are there forcing functions — physical delivery and recurring expiration — that drag price back to reality? (Equities have neither: 'the IPO stocks are basically immortal.') Participants: professionals hedging real cargoes against their banks, or 'everybody', where crowd psychology is a first-order input. High-graded markets get ranked above satellites, officials and narrative; low-graded ones can 'stay wildly inefficient for a very long period of time'.

Free cash flow divided by total assets minus intangibles; Lukacs's core profitability metric for goodwill-heavy companies, used to rank Scandinavian Tobacco against Altria, BAT, Philip Morris and Imperial Brands.

For closed-end funds and listed holding companies the quoted return is the share/unit price, not net asset value; the gap is the discount or premium moving. In the table Hay reproduces, Pershing Square's NAV was −7.1% against a −0.8% unit return — the discount narrowed while the portfolio lost money.

A test for whether a succession or delegation is real: ask whether the escalation trigger is subject matter or size. Abel had been adding to the same position for months unescalated and called Buffett only when the block was large - 'consistent with how we manage Berkshire, but also the governance around it' - and the call was a consultation, not an approval gate, with the recommendation still originating from him.

Terranova on Broadcom into its print: cheaper (21x forward vs 32x in June, 18x ten-year average) is not cheap, and neither long nor short offers an edge — so the output is no position, paired with a falsifiable test ("if they deliver 200% earnings growth and 85% revenue growth and that's not enough, then that AI momentum trade is going to continue to sit silent") that turns the abstention into information about the whole group.

A central bank can push its currency lower indefinitely but not higher; because its intervention spending is countable, a speculator can size to outlast it.

A passive fund holds no cash by design, so redemptions force proportional selling; because it is cap-weighted, the selling lands hardest on the largest and most-crowded positions. The vehicle that absorbed the inflow is the one with no discretion when it reverses, which turns an ordinary outflow into an outsized decline.

One misstatement from a CEO, IR person or board member can be error; the second is permanent — never own anything that person is involved with again, at any company. Honesty is the one variable you cannot model or diversify away.

Don't reject a company on shares outstanding alone. Ask what proportion is locked up and by whom — insider lock-ups, named institutional anchors, holders who structurally will not sell at this price. A big register with non-sellers is supportive, not a red flag.

Paulo's name (via 'my pal JH') for a crowded position with little left to add and a dated catalyst in front of it. Grade a thesis by the crowd's remaining capacity to act, not by whether they agree: CTAs short bonds 'may not have a lot to sell if bonds continue to deteriorate, but they would have a lot to buy if yields reversed lower.'

Average every index member's own maximum drawdown; the gap to the index's drawdown measures how much correction has been absorbed by rotation. S&P: -25.5% average member vs -9% index; Nasdaq -45% vs -13%.

The right way to rank index movers: price performance alone says nothing about who actually moved the index. On this measure Micron ranked 3rd and Tesla 503rd in the same year.

In calm regimes near-dated implied correlation trades below longer-dated. When 1-month flips above 3-month the market is paying up for correlation now — 'something we see as market stress starts to rupture.' The timing layer on top of the level; it should be corroborated by the VIX surface inverting the same way.

Read a bond sell-off by asking which maturities made new lows. Whole curve = macro repricing. Belly alone (5Y/10Y broken while the 2Y and long bond hold) = a bulge, the signature of positions being run rather than a changed outlook. Test it against a real shock: a $5 oil rally that moves the long bond 2bps says the driver is not macro.

Accumulation (only diehard contrarians buy) → public participation (media, product launches, career-risk-averse analysts writing the trend — by far the longest phase) → distribution/parabolic (contrarians sell to retail). Stöferle calibrates each phase with a 'cocktail-party test' rather than a price level.

Layer characteristic screens (rising forward estimates, margin strength, positive surprises; P/E, price/book, price/sales; free cash flow, interest coverage) over sector calls — factors have shown more consistent out/underperformance than monolithic sectors, and dispersion within sectors is now wide.

Paulo's own fragility gauge, built after Kevin Muir plotted implied and realized correlation separately: take the difference. A tightly compressed spread with both legs at extreme lows means options are priced for the calm that is already happening, leaving no cushion — 'if it's gonna go, it has a lot of room to move.'

A fund being wound up must pay out net asset value on a published date, so the terminal price is known in advance while mandate-constrained holders are forced out into thin final-session liquidity — buy the discount to the pending distribution. Ferg's worked example: iShares' FM, last session 6 Jan 2025, $27.23/share liquidation distribution 9 Jan 2025.

At a rich multiple there are ~20 independent ways to disappoint, each maybe 5% likely; combined they compound to a >60% chance of failure — so a high multiple is a forecast, not a risk to size around, and what looks unpredictable is actually highly predictable.

Two separately budgeted buckets: safety gold is physical metal in a safe jurisdiction outside the banking system, buy-and-hold and never timed; performance gold is mining equities and silver, which stack geological, permitting, ESG, cost and management risk on top of the metal and must be actively timed. Keeping them in separate budgets stops a drawdown in one forcing a sale of the other.

Weight a portfolio by cash actually returned to owners — dividends plus buybacks plus debt paydown — rather than by market capitalisation; cash already out the door is harder to fake than reported earnings, and the tilt avoids cap-weighting's mechanical concentration in the biggest winners.

Report owner's earnings (EPS growth + dividend yield) separately from the multiple effect, stating the current level, the assumed fair level and the convergence period — so the part you do not control can be discounted.

Tracks implied volatility in Treasuries the way the VIX does for equities; a yield rise with a calm MOVE is orderly, a level breach plus a MOVE pickup is what spills into equity volatility.

Sonders' extended mega-cap AI basket, posted daily on her X feed; the rank spread inside it is her dispersion gauge.

Replacement for the classic 60/40 once stocks and bonds correlate positively: 14–18% safety gold, 10% performance gold (miners + silver), 10% commodities, 5% Bitcoin and ~15% fixed income as a stabilizer (EM local-currency and corporate). 2–3% gold 'doesn't move the needle'; 30–40% imports a different risk set. Live outperformance vs the traditional 60/40: >25 percentage points over two years.

"The value of gold is not what it promises, but what it spares its owners" — no duration, credit or liquidity risk, no balance sheet to implode, no cash flows to dry up, no management misallocating capital, no counterparty goodwill required. Invert the usual analysis and price any 'convenient' wrapper by which of those risks it reinstates.

After a run of blowout quarters the live bar is the buy-side whisper number, not published consensus; a print between the two is a beat on paper and a miss in practice.

The dated analog Paulo uses to calibrate today's vol compression — the 2024 episode where a crowded yen carry trade unwound and volatility exploded. He notes the current implied-vs-realized correlation compression was 'even tighter than pre-Yenmaggeddon in July 2024.' Dates the condition, not the outcome.

Before acting on "X moves Y," find the biggest, fastest historical move in X and read Y across it — then express the currently-feared move as a fraction of the one already absorbed. Prins' worked case: 525bp of 2022–23 hikes left gold flat, and it doubled with rates still at the peak, so a 25bp hike cannot be the driver.

Credit and equity price the same balance sheet with different loss functions, so credit notices leverage first — it has no upside to compensate for it. Record-wide default-insurance costs on issuers nobody expects to default are a statement about supply and a widening range of outcomes, which transmits to equity as multiple compression (SOX forward P/E 21-22x to ~15x) even while earnings estimates rise.

Before buying a fallen compounder, list each distinct cause of the earnings fall, label it temporary or structural, and say what specifically reverses it and when the comparison base stops being distorted.

Under Kahn v. M&F Worldwide, a controller's take-private earns deferential business-judgment review only if it is conditioned at the outset on BOTH an independent special committee AND a majority-of-the-minority vote. Without both, the controller is exposed to post-closing fiduciary litigation — which is why raising the bid is usually cheaper than forcing it through. The legal spine of the PRTH thesis.

A company buys back stock by letting holders name a price inside a stated collar, then pays everyone the single lowest price at which it can fill the full amount. The collar turns the trade into a bounded distribution you can price before entering: ABUS's $5.00-$5.75 range gave a −1.8% worst case and a +12.9% best case off a $5.20 entry.

In most tender offers, holders of 99 shares or fewer who tender their entire position are bought first and in full, exempt from proration. It is the rare structural edge available only to small accounts — and it applies per beneficial holder per broker, so the same trade can be replicated across accounts. Singh's ABUS trade is built entirely on it.

Write the target as revenue x margin x multiple, label each link as guidance, consensus or your own assumption, then flex the assumed links. A conclusion needing every optimistic input and carrying no date is not a valuation.

Interim dividends in a merger are sometimes paid on top of the deal price and sometimes deducted from it. Two spreads that look identical can have materially different returns. Two Harbors' $0.20326 stub was explicitly paid 'with the merger consideration and will not reduce or otherwise affect' it — free carry on the final weeks of the spread.

Track the trust's net asset value rather than its price and add only when the shares trade below the metal they hold - a computable edge that exists before the commodity moves. His band: -10% to -15%.

Russian-origin term for whether a state can be relied on to honour a deal across a change of administration; Doomberg applies it to the US, whose short political cycle means promises (the UN-blessed Iran deal, offshore-wind subsidies) may not survive the next president — a discount on any long-lived asset whose returns rest on policy.

The venture-capital rule that "you have to be very careful who your first customers are because early adopters aren't like anybody else" — build for them and you never find the valuable customers. Applied to markets: check which cohort generated the data behind any adoption curve before trusting the capacity, hurdle rates and debt sized off it.

Why bearish sentiment need not produce selling: a long-held winner in a taxable account costs 35%+ of the gain to exit — a certain, immediate bill against a merely possible loss. Concentrated positions are therefore stickier than surveys imply, which slows a decline and deepens it.

Judge a junior explorer on who owns the shares before the geology: management owning 20-50% of stock outright (not 3% plus options), treasury cash against a small share count so no dilution is coming, and long-horizon holders rather than warrant-flippers on the register. "It's who owns your shares, not just how many shares are outstanding."

In cyclicals, a compressed multiple at record earnings is the market forecasting the E will fall, not offering a discount on the P — 'in semiconductors, a low P/E can sometimes signal peak earnings rather than a bargain.'

McCracken's core market model, from two years spent on P versus NP: the market is the hardest system humanity has built because it keeps generating new states - front-running a pattern destroys the pattern. The operational consequence: 'if everyone is doing the same thing... evolutionary systems don't let that happen. You get an extinction event afterwards.' Illustrated with quant funds converging on one strategy and losing 5-10% together, and applied live as the reason to cut gold miners by half when 'literally everyone on Twitter called the bottom.'

At a rich multiple you should not hunt for the catalyst but count the failure paths: twenty roughly independent ways to break, each ≤5% likely, compounds to a greater-than-60% chance of failure over the same window — so what looks unpredictable is highly predictable. Worked example: Nike's P/E going from ~70 to ~20 while every post-hoc explanation blamed product or politics. The mechanism looks like being "pecked to death by ducks."

McCracken's test for turning a shortage into a trade. A sulfur shortage genuinely threatens fertilizer supply - but a fertilizer producer is a buyer of sulfur, so the crisis hits its input costs before its selling price: 'there's probably gonna be a fertilizer crisis and it's hard to make money off it.' Same test rules out commodity chemicals, whose naphtha feedstock is distorted by Hormuz. If no listed company owns the scarce input, pass rather than substitute the nearest vehicle.

Stake or option prospective ground cheaply, do the early cheap work to define drill targets, then bring in a partner who funds the drilling — you keep discovery upside while conserving capital and avoiding dilution. Named by Phillips as one of the two models that made Rick Rule the most money; Latin Metals optioning Cerro Bayo to Daura is the live example.

Pay a miner cash up front for a permanent slice of future production or revenue; the royalty holder's cost is fixed at the deal, so operating-cost blowouts are the miner's problem while the royalty keeps paying. Rule's other top model — Franco-Nevada and Wheaton are the mature end, Empress Royalty the early end.

The computer-science idea McCracken credits for his exits rather than his entries - a path that never revisits a state, which he uses to think about when a commodity move has exhausted its available ground. His worked example: 'I pieced out of oil in full' in December 2023, ahead of the 2024 bear market year.

For a fast-moving loss-maker, drop year-on-year comparisons and compare quarter to prior quarter in absolute dollars: revenue added minus costs added. OpenAI's June quarter — revenue +$1B sequentially, costs +$3B — turns before any ratio or margin does, and converts directly into a funding requirement.

Set the cash cushion at whatever percentage stops you checking prices at midnight while keeping you fully engaged — 20% for one person, 10% for another, both correct. The cash is held as optionality on somebody else's forced selling, and as what lets a core position (a gold sleeve) stay untouched when the front end reprices and metals go heavy.

A long/short book is only hedged if the two legs depend on different propositions. Buying every auto company and shorting every carriage company in 1905 is one trade, not two: "your longs and your shorts are not independent of one another. They move exactly in unison." A temporary sentiment shock reverses both together — and at 4:1 leverage it ends the fund before a correct thesis can pay. Eisman's post-mortem on Situational Awareness.

Pay the seller in retained upside rather than headline price. Best form: a free-cash earnout defined as cash flow minus capex, so every euro the buyer reinvests both grows the asset and shrinks the cheque; triggers set far out of the money and measured annually so the seller cannot "clip the tops" monthly.

Split the valuation into what you can forecast well (rates, decline, capex, opex — the "controllable inputs", modeled well by well) and what you cannot (the commodity price). Do the first at absurd granularity; only sensitise the second. The technical work prevents a technical loss, it does not win the return.

Hedge without paying premium by selling a call to fund a bought put; when the market's skew favours you, the sold call sits further out of the money than the put. A "three-way" adds a sold put below the floor — more coverage per dollar, but downside protection stops at that level.

Subtract new debt times its rate from earnings, divide by the reduced share count; EPS can rise while net income falls (Salesforce).

Assets carried below their estimated intrinsic value, in a vehicle whose shares then trade below that carrying value — two stacked discounts, closed only by an event that puts a market price on the assets.

Price the deal as of an effective date well before close; cash the asset generates in the gap accrues to the buyer and nets against the consideration. Tenaz's 16-month gap (Jan-1-24 to May-2025) paid down most of a €165M price and returned cash at the closing table.

The gate at which a development project stops being an option and becomes a committed build. Used as the discrete catalyst for Century Aluminum's Inola smelter, contingent on resolving the Oklahoma AG lawsuit, finalizing the power-supply agreement, and completing detailed engineering.

Where variable cost is a small share of opex (offshore platforms: transportation under 10% of the total), incremental volume is nearly all margin and unit costs fall mechanically as production grows. Plot revenue per unit against cost per unit — two diverging lines. Higher margins are also a risk measure: a price shock hurts a fat-margin producer far less.

Page-one stories are already priced by an efficient market; the edge is in developments not yet widely known that are heading for the front page.

Compare a holding company's realised (monetised) returns against the unrealised returns on its private book. If realised matches the listed book and both exceed the private marks, the private marks are conservative — and the gap scales with the share of assets not marked to market.

Before buying a managed vehicle, convert its management and performance fees to an annual drag and identify the recipient; if the recipient is listed, the same exposure can be taken on the receiving side of the fee.

In a common-value auction the winner is systematically the bidder who overestimated most, so entry price is driven by the bidder count: five or six bidders and "it's hard to make a good deal"; one or two and you can. Marino deliberately hunts markets with few technically qualified buyers.

BCOM weights two-thirds on futures liquidity / trading volumes and one-third on world production, then applies caps: no sector above 33%, no single commodity above 15%, reset at each annual reconstitution. That is why its energy weight is only ~30% while production-weighted competitors are effectively energy bets wearing a diversified label. The counterintuitive payoff: a capped basket's volatility is similar to broad equities and at times dips below the S&P 500's, even though individual commodities are volatile.

In a shortage-driven capex build-out, buyers allocated a fraction of what they request start ordering more than they need ("you tell me I can only get 500, so next time I say I need 2 million") — so the supplier's backlog counts the same demand twice. An X survey found ~50% of AI-infrastructure buyers admitting to it. "That's how these cycles end."

Stocks normally yield more than government bonds as compensation for risk; when that inverts it "usually doesn't last long in history." It is the arithmetic behind a ~0% projected 10-year return on the index, and it sets up an asset-allocation switch out of stocks into bonds that, once flows start, "happens quick."

An MOU is a statement of intent, not a commitment — "it's not a real commitment until the contracts are written… it can just vaporize overnight." Nvidia's $500B Blackstone financing and OpenAI's Stargate $500B were both MOUs; markets re-rate on the headline number as if it were funded cash.

The recurring-revenue metric hierarchy by how early each turns: net new ARR leads, cRPO (contracted revenue due within 12 months) follows, ending ARR is a smoothed sum, and reported revenue is largely last year's bookings arriving. Judge inflections on the leaders.

Three buckets with different jobs, not one portfolio with a risk dial: savings (gold, wealth itself and liquidity), investment (large miners, accepts company risk), speculation (juniors, requires work plus psychological durability). Every allocation question is answered 'on behalf of whom?' rather than with one winner.

Hyper-growth stocks are priced off the rate of change of growth, so a build-out doesn't have to stop to break them — it only has to slow. "You don't need it to stop. You just need it to slow. It's a second derivative question, and semiconductors are the classic canary in the coal mine."

If a stock doubled on the commodity price and not on anything the business did, it is 'precisely half as attractive as it was before the price doubled' — so sell half and hold the rest at zero cost. Applied hardest to positions added recently.

Over 40 years, non-specialists entering a precious-metals bull show up as silver beginning to outrun gold (perhaps its lower unit price). Rule treats sharp silver outperformance as a warning of overly broad participation, not as a buy.

Classify the shape of a move, not just its size. A steady multi-year trend and a two-month exponential blow-off have different aftermaths: the exponential move is 'typically unsustainable' and is followed by months-to-years of range trading, regardless of how good the story is. Wiederhold reads silver's 2026 spike to $100 off the 1980 spike to $50, and gold's 2.5–3 year runs off data going back to 1960.

A KPI that improves in the same quarter its definition or accounting treatment widens has not been shown to improve — Salesforce expanded Agentforce ARR to include Slackbot and Headless 360; Klarna's raised transaction-margin outlook arrived with a new Fair Financing treatment. Lean on the unedited neighbouring metric until a like-for-like figure appears.

Sizing a capital project by what the buyer loses if it fails outright rather than by its projected return — the variable that decides who is willing to be a first customer. 'Even if a microreactor project fails, the cost of walking away will be in the millions, not the billions'; Segra's Arthur Hyde: 'At a minimum, this reactor is not making me bankrupt.'

Write down the single falsifiable condition that made you buy (Rule bought silver only because it was hated), monitor that condition rather than the price, and force a decision when it lapses. Re-entry requires the original condition returning, not merely a lower price. Sentiment is graded hopeful / bored / afraid / hated — only the last qualifies.

Cash is credited near 100c on the dollar, so spending $100M to create $200M of NPV that the market capitalizes at half NAV leaves the stock flat while consuming the best acreage - whereas buying back stock at half NAV is accretive. Explains why high spot prices produce no supply response.

Companies that spend spare cash retiring their own stock, so each remaining owner's slice grows without the business growing. AutoZone is the canonical case — "the business didn't really grow that much… the stock was phenomenal because they were just gobbling up so many shares." A persistently low price is a gift to a cannibal.

How much extra growth a company needs just to keep per-share figures flat once dilution is netted out. "If you're going to own something for 10 years, 1% dilution adds up quite a bit. 2% dilution is very significant" — hence the ranking: shrinking share count > flat > diluting.

Kirby's story of a client who piggybacked every recommendation and never sold any — one position ended up worth more than the whole account Kirby was managing alongside it. Tolerate the zeros; "the winners more than made up for it." Claude Shannon's never-sell VC portfolio is the same shape.

Dynamic pricing varies with the situation (demand, supply, timing, distance); personalized pricing varies with the person (income, spending habits, device). Collapsing the two turns a normal market-clearing mechanism into a scandal — the distinction at the centre of Carlson's Uber rebuttal.

Buy a listed vehicle whose purpose is to allocate capital, judged on the operator's track record in that deal type and geography, plus the structural assets the wrapper carries (net cash, and especially accumulated tax losses that make future profits worth more inside it). Exit trigger is the jockey leaving, not a bad quarter. Illustrated with Aimia under Rhys Simmerton.

Score the industry, not the company: power of buyers, power of suppliers, threat of new entrants, threat of substitutes, rivalry. The Oak Bloke walks it through on a South African coal miner — new entrants nil (nobody finances a coal mine), substitution very real, and every stakeholder from the state to the export railway taking a cut — to force the weaknesses to be stated out loud before a cheap multiple does the arguing.

At a cycle peak the denominator of P/E is a margin that cannot persist, so the multiple prints cheapest when earnings are least repeatable — 2022's $100 oil more than doubled Diamondback's EPS and 'deflated the P/E to an unsustainably low point' while Price/Sales fell 'to a significantly lesser extent.' Sales are smoother and harder to flatter.

Instead of forecasting a fair value, cut the company's actual EPS growth (with a named cause), haircut the exit multiple, fix a required return, then solve for the purchase price that still delivers it. Carlson uses it to set a hard buy target on all 14 holdings.

The private investor's structural edge is not information but the absence of an investment committee, a quarterly report card and redemption risk: "we're not constrained by time, we don't have an investment board looking over our shoulder like why did you buy that, it's down 30%." It only pays if you deliberately fish where professionals cannot — and only if the thesis is genuinely intact rather than patience becoming denial.

Weights each period by the capital actually invested in it, so deposit timing can't flatter the number. Carlson: gain divided by balance would show ~+50%; the honest time-weighted figure is a 14.8% CAGR, because for a third of the portfolio's life he had under $100k and most deposits landed late.

Why a stock falls on a large convertible bond issue regardless of the business: the buyers are arbitrage funds that purchase the bond and simultaneously short the equity to neutralise their exposure. The selling is a function of issue size, not fundamentals — Nebius's $4.5 billion convert pressured the shares in the same week it cleared the Vineland data-centre approval tied to its $17.4 billion Microsoft contract. Distinguishing this mechanical drag from a thesis change is the point.

Free cash flow after interest, divided by market capitalisation — what the equity earns in cash per dollar paid, with the debt already served. PRTH's $91M of expected 2026 FCF on a ~$460M market cap is a ~20% yield, which is what makes a 5.5x-cash-flow take-private bid identifiable as a lowball. Always paired with net debt to EBITDA (~4x here), since leverage is usually what suppresses the multiple in the first place.

For a low-value, heavy product, freight cost sets the market radius; whoever owns the nearest supply inside it owns the market, and permitting makes it unrepeatable.

Revenue this year from last year's customer cohort divided by their revenue last year; above 100% means existing customers are spending more (Zeta 114 to 120).

Two companies buying each other's products to inflate revenue; the allegation in Zeta's short report that a Deloitte review rejected.

When a holder who already controls a company bids for the minority, no rival bidder is possible, so the only protection for outside shareholders is a committee of independent directors legally obliged to negotiate on their behalf. The tradeable consequence: the bid becomes a structural price floor while the committee — under public pressure from activists with a published valuation — is the option on a bump. Worked example: PRTH at $5.52 against a $6.00-6.15 offer.

A per-day payment the acquirer owes the target's shareholders if a deal has not closed by an agreed date — $7 million a day from Paramount Skydance to WBD holders past October 1. It inverts the usual arbitrage asymmetry: delay becomes a coupon rather than a loss, raising the floor under the target, and it is valuable enough that the acquirer demanded the objecting states post a $1.9 billion bond to cover it.

Model revenue from raising the share of each client's category budget you capture, with zero new customers; Zeta takes ~1.7% of Fortune-100 marketing spend.

Munger's standard — you have to argue the other side better than the other player. Because an unappointed bear pays a social cost for dissenting, assign the opposing case to a named person with explicit licence to attack, have the bull answer each objection individually on the record, and rotate the roles on the next name. The objections the bull can only mitigate rather than rebut become the monitoring list.

Identical rights can trade at very different prices across two listings. Establish the normal direction of the premium for that market pair, then check float size, foreign access and fungibility — if the lines cannot be converted, nothing mechanically closes the gap. Anchor the compression estimate to a long-history comparable (TSMC's US line has averaged ~15%) rather than to zero. CATL's Hong Kong H-shares trade 30-35% above the Shenzhen A-shares, the reverse of the usual mainland premium.

Robin Zeng's 2017 letter to CATL staff at the peak of China's EV subsidy boom — if you stand where the wind blows, even a pig can fly. The screen it implies: date each policy support's removal and measure what happened to share and margin afterwards; if both held once the wind stopped, the advantage is internal rather than political.

The moat is not any single item on the checklist but the loop connecting them: biggest -> cheapest -> most profitable -> biggest R&D budget -> best technology -> more customers -> bigger still, extended upstream into inputs and downstream into the customer's product architecture. A competitor must undo years of compounding simultaneously, not match one attribute.

A special management share typically held by the state or a government fund, giving veto and oversight rights without affecting day-to-day operations — used mostly where content distribution matters. Its presence or absence is a governance detail most Western screens skip; CATL is founder-controlled with no publicly disclosed golden share.

A workaround for a political ban on foreign ownership: the local partner builds, owns and operates the plant while the licensor supplies technology for a royalty (typically 3-4% of revenue), earning capital-light high-margin income. Score it on three axes — economics, severability (what physical asset survives if a regulator cancels it), and leakage (the licence transfers know-how to a future competitor by design).

When operating cash flow runs at a persistent multiple of net income, look for structurally negative working capital: cash collected from customers months before suppliers are paid is an interest-free loan from the supply chain, the same economics as insurance float. Price the unwind, not the level — buybacks and dividends are what the float funds first, and a regulator forcing faster supplier payment is the trigger.

Revenue minus transaction costs in a payments/BNPL business — the money left after funding the purchases and absorbing credit losses. Klarna's grew 42% while GMV growth halved to 15%, the clean example of unit economics improving on a shrinking volume base.

Put unit growth and revenue growth side by side; a wide negative gap is realised price deflation. Check whether it is competitive or contractual — raw-material indexation clauses hand efficiency and commodity savings straight to the customer, so the company runs faster to stand still and is paying its pricing power for utilization. CATL: volume +21.8% against revenue -9.7%.

Margin expansion during a revenue contraction is arithmetic, not superiority — input costs fell faster than selling prices. Treat it as a temporary cost wedge that regresses to the historical band rather than as a new baseline, and refuse to upgrade the quality score for it.

How long an AI compute asset takes to earn back its purchase price, judged against how long it keeps working — the gap is the profit window, and it converts a free-cash-flow burn from an unbounded subsidy into a purchase with a horizon. Alibaba states ~3 years breakeven, trending toward 2.5.

Crypto-native contracts that trade 24/7, offer very high leverage and never settle or expire — identified on Halftime (Oliver Renick) as the real competitive threat to the regulated exchanges (Cboe, CME, ICE), alongside prediction markets.

A cure for both bullish and bearish price disbelief: costs move with the commodity, so returns are far more stable than prices. Oil tripled from 2000 to 2006-07 and IRRs went down; a Calgary upstream asset underwritten on a $110 deck at ~25% IRR still returned ~17% at $40 oil, because steel, copper, labour, the Canadian dollar, food and fertilizer all repriced. "You'll live through it if you let it reprice" — the ones who die are the owners of fixed physical assets that cannot re-cost.

How much of reported profit is backed by cash rather than accounting judgement; one of three inputs (with ROE and leverage) in the S&P 500 Quality index construction.

His third category of serial acquirer: instead of buying other businesses it retires its own stock. Argent halving its share count (95m to 53m) at half of intrinsic value doubled Milkwood's 15% stake to 30% without buying another share.

What overseas buyers actually earn on your bond market after hedging the currency. At -120bp for Japanese buyers of the 10-year Treasury there is no hedged bid, leaving only two resolutions — a much weaker dollar, or much higher US yields. 'Japan's just telling us what's going to happen in the US.'

The Chicago price-theory lesson he still trades on: prices are set at the margin, not by the average. "I don't care about the cost structure of the vast majority of producers — all that matters is right at that margin, and that's where all the action takes place." Hence "macro guys are average guys, micro guys are marginal guys: there's a lot more information content found at the margin than at the average."

Subsidiary managers keep complete control of their own business and are incentivised on its results; head office stays out. The corollary is the failure signal: a new CEO who centralises the autonomous units into one head office to strip out their CFOs and HR 'strangles the business'.

Across all five prior US capital-spending booms back to the 1840 canals, an investor two to three years in did better selling most of the position and buying gold — gold outperformed over the full cycle every time, even where the boom itself ran on for years. Not a shorting rule: manias end when new credit stops arriving, not when valuations get silly.

A company that compounds by making one acquisition after another. Three species: perpetual holding companies (hold forever), thematic acquirers (consolidate a single industry and build expertise in it), and cannibals (buy their own shares). What makes one work: decentralisation, management that is motivated and honest, real cash flow, a long countable runway of targets, a tax loss the market never prices, and no debt.

Money sold out of a small company and put into an index fund is redistributed by market cap, so it lands almost entirely in the largest names. Seventeen years of it produced the $600bn/90% regularity at the top and, at the bottom, thousands of cheap, capital-starved companies whose managers have stopped trying.

The structural supply of cheap acquisitions: owner-managers reaching 65-70 with no successor who must monetise, selling at three or four times EBITDA. The UK pool he counts: 5.5 million small businesses, ~20,000 manufacturers earning over $1m of EBITDA and owned by no larger group.

A tax-deferred in-kind contribution of liquid stocks/ETFs to seed a NEW ETF, receiving fund shares back — no capital gain on the day. Two rules govern eligibility: no contributed position above 25%, top five under 50% (~11-12 names; index ETFs are passed through to their holdings). Basis carries through, so it is a deferral, not a dodge — Faber: 'most likely these funds will grow, they will compound, and the tax liability will actually be bigger. It's just delayed.'

A forced dividend takes cash out of the business, is taxed on receipt, and leaves only the remainder to reinvest — a mandatory taxable event you did not choose. The decision is account-type-first (irrelevant in a sheltered account) and measured after tax; the meaningful comparison is against high-dividend strategies, not a ~1%-yielding index.

Reading a published favor/oppose polling series (77% vs 23% for US nuclear, Bisconti) as a quantitative proxy for permitting risk and the durability of policy support across election cycles.

Total cash returned to owners = dividends + net buybacks + debt paydown, with buybacks counted NET of new issuance (especially stock-based comp). A 5% announced buyback against 7%/yr of SBC is a negative buyback yield; 'the average stock in the US is an issuer.' Faber's argument for why dividend-only screens are incomplete — buybacks have exceeded dividends for 20+ years.

Quarterly SEC disclosure required of managers with over $100M in US equities, filed 45 days after quarter end. Four structural blind spots: omits shorts and cash; partial universe (smaller funds excluded); US long equity only; dated by the 45-day lag. Read for diffusion across a fixed fund sample, not for names to copy.

"You can borrow someone else's stock ideas but you can't borrow their conviction. Do the work so you know when to sell. Do the work so you can hold. Do the work so you can stand alone." The argument for using filings as idea sourcing rather than as a portfolio.

The real-estate analogy behind Strathcona and Greenfire: "you find a good neighborhood, you find a good street, you buy a house, then you buy up the block." Only geographically adjacent acquisitions deliver both overhead elimination and transferable technical know-how — the test being whether the deal also improved the assets you already owned.

"If I invest $100, how long is it going to take me to get that $100 back?" — measured on flat production. Waterous prefers it to DCF in long-reserve industries precisely because 50-year engineering reports make DCF models easy to build and easy to tune; he runs DCF as a secondary check, never as the decision.

Write the numeric business conditions for trimming or exiting before you own the position, phrased as business rather than price triggers, with the override defined in advance.

Score every holding on a fixed criteria list after every earnings report, whether or not anything looks wrong; the signal is the trajectory of the score, which catches slow deterioration you would not otherwise see.

Cap position sizes on what you originally put in, so a winner does not automatically look too big and a loser too small — 'I refuse to punish my winners and reward the losers.'

2% of assets plus 20% of gains. Cited as the reason a copy-the-holdings product flatters the strategy it imitates: the Global X Guru ETF trailed the S&P 500 since 2012 even before the fee layer the underlying funds actually charge.

Require both halves at once: the product is personalised to each customer, and one more personalised customer costs almost nothing. Netflix's recommendations pass; personalisation delivered by people or inventory fails.

The cost of major litigation is management attention, not the verdict. IBM's 1970s antitrust case "lasted 12 years. IBM won, quote unquote, but they really lost because it consumes so much of their time and effort that they missed personal computing." Microsoft in the '90s repeated it; Trennert applies the template to Meta's hundreds of social-media suits.

Strategas' label for a regime where returns come from earnings rather than multiple expansion, so dispersion rises and active stock picking beats passive. Their confirming evidence is the tape in the asset managers themselves, breaking out for the first time in ~15 years of QE-driven passive flows.

Judge a leadership drawdown by what the rest of the index did while it happened. The S&P went from 50% of members above their 200-day at the June 2 high to 75% eight weeks later, even as semis and hyperscalers fell 30-50% — money rotating, not leaving.

For decades Japanese corporates optimised for market share and earned no return on equity. The exchange now enforces the opposite by delisting companies that fail minimum price-to-book or ROE thresholds. Trennert's aside: the old Japan is "a little bit reminiscent of what's happening in tech here in the US."

Eisman's worked example: in 1900, buy every automaker and short every buggy-whip maker. History proves the thesis right, and at 4x leverage one bad headline still wipes you out, "because you're long X and you're short Y, but it's the same thing." His explanation for how a $20B fund died with the VIX below 20 and long rates up only 15-20bp.

The only question that dates an equity cycle: at what yield does the marginal owner prefer bonds? Distinguish the level that causes a correction from the one that ends the cycle, and calibrate against bubble-era analogues — Japan '89 (JGBs 4 to 8), Nasdaq '99 (US 10s 4 to 7), 1987 (long rates 6 to 9 while stocks rose 30%) — not the last cycle's average.

The funding mix, not whether the technology is real, decides what the bust looks like. Equity-financed booms break benignly (dot-com: market halved, economy shrugged); debt-financed ones transmit into the credit system and 'quite often end in tears even when the underlying premise comes true' — canals, railways, housing.

'Whenever I hear people talk about democratization of something I tend to reach for my metaphorical gun — it's usually a code word for jamming something down the necks of retail investors that are not really quite ready to digest.' Retail access typically opens after institutional returns have compressed, which makes it a late-cycle event by construction.

A bidder can require existing owners to reinvest their shares in the new entity rather than cash out, reducing the equity cheque it must write. When those owners refuse, the bid usually dies structurally rather than on price — the Carano family's refusal is why Icahn's $34 Caesars offer never became credible.

Systematically scanning thousands of ISINs and CUSIPs for filings tied to M&A, spin-offs, strategic reviews, buybacks, rights issues, restructurings, liquidations, delistings and litigation, then describing each event, valuing the company quickly, identifying catalysts and monitoring progress. Singh's stated purpose is to fix a coverage gap: "one thing that I want to get better at is not to miss names, especially small cap names."

Junk defaulting is just risk and reward. Catastrophe arrives when an asset treated as money — or as the foundation of a bank's business model — turns out not to be. The collateral ladder ran government bonds → high-grade corporates → asset-backed securities, each 'super solid' until it wasn't. Long calm since the last crisis is the precondition, not the comfort.

A window after a merger is signed in which the target may solicit better offers, usually paired with a lower break fee. When it expires the fee steps up, raising the buyer's cost of walking away — a public commitment signal. At Caesars the extended go-shop expiry doubled the termination fee from $100M to $200M.

A business whose only asset is a run-off book that shrinks every year until nothing is left — priced at a deep discount to book on the assumption it liquidates. The thesis flips when new origination replaces the run-off, as Gator Capital argues is now happening at Navient via private lending.

Since 1990, from end-July through November 9, the S&P 500's median gain in midterm years is +0.9% with gains 56% of the time but a median peak loss of about 6.2%, versus +2.7%, 59% and −3.5% in non-midterm years. A positive median with double the drawdown makes it a hedging problem rather than a selling one.

Find companies whose market classification and balance sheet disagree: read the asset side, assign the sector you would give it with no name attached, identify the concentration and the macro sensitivity the label hides, then quantify downside as the re-rating to the honest peer group. GBFH is "a $1.4 billion bank, valued as a gaming payments platform, but really an undiversified economy hotel monoline."

When a company's net cash exceeds its market capitalisation, you are buying the cash at a discount with the business attached. Singh's gates before it becomes a trade: has the cash burn been unwound (otherwise the cash is melting), express net cash and business value separately per share, and check whether the earnings stream is too concentrated for the market ever to capitalise it. Applied to BASO ($40M cap, $38M net cash) and GRVY ($430M cap, $400M net cash).

Reduce a data-centre compute contract to rent per megawatt per year, apply the NOI margin the structure implies (80-90% when the tenant funds its own equipment), subtract build cost per megawatt and capitalise the result. Riot's Anthropic deal: $2.4M per MW of annual rent, 80-90% NOI, ~$11M/MW of capex, 15x NOI equals roughly $15 per share of contract value.

Interest paid in additional debt instead of cash, used when a borrower cannot service its coupon. Rising PIK is the transmission channel from the private-equity exit bottleneck into BDCs: "BDCs own a lot of these PE companies that will need to refi and will not be able to service interest, which is going to mean more PIK interest."

Payment-in-kind is legitimate for fast-growing borrowers but is widely being used to postpone defaults; separately, the market inherits recovery assumptions (70–80c on the dollar) earned by borrowers who owned physical assets. An asset-light software borrower leaves a creditor 'nothing there' — so the loss math, not the default rate, is where the error sits.

A vendor guarantee of what used equipment will be worth, offered to make third-party lenders comfortable financing its purchase. Nvidia caps it at 25% of any project in the $500B AI-factory financing platform, "designed to complement, not replace, independent underwriting" — its answer to the circular-financing charge.

A promoter pattern in listed shells: engineer or ride a spike in the share price, then issue new stock into the strength — enriching insiders and diluting holders. Singh's stated reason for shorting DFNS, whose operator "has done a squeeze and dilute type of thing with SPACs in the past."

Why rating agencies endure: nobody buys Moody's or S&P for the analysis (investors have automated prospectus-reading for a decade) — they buy a shared shorthand. All three deliberately publish the same letters because standardisation is the product, and the designation is written into US law. 'If they didn't exist we'd have to reinvent them all over again.'

Once a scarce input becomes replicable across many suppliers, economics accrue to the layer that aggregates demand (Netflix, YouTube, Booking Holdings, the app stores) rather than to any individual supplier.

When demand spikes (here ~6x on a Friday), owned capacity must either sit idle most of the day or forfeit the peak, while on-demand capacity absorbs surges at no idle cost — a first-class competitive variable in any spiky-demand industry.

For a local marketplace the unit of competition is density within one city, not global scale: the supply level at which any customer is served promptly and any supplier stays utilized. Below it, metrics break and the operator must resume subsidising supply.

Entering a market by orchestrating assets that already exist (roads, fuelling, privately-owned cars idle ~95% of the day, spare labour) instead of building them — why a marketplace can scale globally with almost no capex.

Signed NDAs are not bidders: an exclusive negotiation is a one-bidder process with no deal tension. The commodity price decides who owns the clock — a higher metal price amortises upfront capex faster and forces the buyer to move; a lower one lets the sole bidder wait out litigation and permitting at the seller's cost.

Prepayments are interest-free, non-dilutive capital and the hardest possible demand signal — the structural opposite of equipment-collateralized debt; Nebius expects >$9B of 2026 prepayments covering 50-60% of its capex.

Price contracted AI-compute capacity per unit of the binding physical constraint — the megawatt — rather than per GPU or per customer; a short-term premium over long-term contracts ($40-50M/MW vs $20-25M/MW at Nebius) is an inverted term structure and a scarcity gauge.

Baseline is systematic saving in gold out of every liquidity event; the single override is a rout of selling, when future contributions are pulled forward by shifting existing dollar savings in 'out of a current paycheck rather than a future paycheck'.

Uniformity, not content, is the signal: "anytime I see something happen and I get basically 30 versions of the same thing said about it, that's usually your first clue someone's attempting to propagandize you." Genuine analysis produces dispersion; a placed narrative produces near-identical copies. Test it against an arithmetic constraint and trade the unwind.

Deal velocity substantially above the sector's implies the acquirer is winning auctions by overpaying — unless it has a durable structural advantage that keeps it out of the auction (community/cultural access, or captive affiliate deal flow).

AI assistants transacting on a consumer's behalf; the argument is that agents remove friction and raise purchase frequency while adopting rather than replacing the consumer's existing payment method — expanding a payment network's volume rather than bypassing it.

List each holding's actual revenue driver rather than its sector label, group by driver and total the weights — ten names with four drivers is a four-position portfolio. Arka Bhattacharjee's untouchable twenty-year list carries ASML, Applied Materials and Schneider Electric on one build-out, with Microsoft and Alphabet as its demand side, and never mentions diversification.

Pricing power has two distinct forms: the customer does not notice the price (Diploma, PPG), or the customer cannot obtain the product any other way. Ferrari sells fewer cars than demand on purpose, so the price can rise as long as the number of very wealthy people grows faster than production. Check the scarcity is chosen and defended, not a capacity constraint the company would relieve if it could.

Hours consumed are not a uniform metric: a live event or a series drives sign-ups and retention far more than its share of watch time suggests, so scoring a subscription video business the way you'd score a social feed misreads it.

A per-user average can fall purely because growth comes from a structurally lighter-usage cohort (markets that watch less TV), while totals keep rising — a mix effect, not deterioration.

Decomposing an index's gain by sector: when 8% of companies and 22% of market cap deliver ~85% of the return, the flat 90% is being sold to fund the leaders — a flow phenomenon that creates a shopping list rather than a verdict on those businesses.

Polomny's own construction: own assets that capital cannot recreate, earning contract-escalated, capex-free revenue, run by management that prices the asset's optionality rather than the operating business.

Move one step up the supply chain and own whoever sells to every participant in a growing theme: Thermo Fisher for biotech, Applied Materials for chipmaking. You give up the top decile of outcomes in exchange for not having to be right about which name delivers it. Check the supplier's own position is uncontested.

Contiguous private surface acreage monetized by hosting private power generation, transmission, renewables, carbon capture and data centers — each a recurring, high-margin royalty-like stream on land the owner never has to develop.

The strongest form of switching cost is one enforced by law rather than preference: once an instrument or reagent is written into a regulator-approved testing protocol, replacing it means revalidating the whole process. Cited as Thermo Fisher's real moat.

Muir's stop rule on a contrarian entry: a position entered on price-action confirmation is falsified by price, so cut it rather than averaging down — the opposite discipline to Haymaker's own dollar-cost-average-into-weakness habit.

Gold uniquely prices calls above puts (the perceived tail is up); when the 1-yr 25-delta call skew collapses from its usual 3-4 vol points to a multi-year low, nobody is paying for right-tail risk — a sentiment reading backed by money at risk rather than a survey.

Only take a view on what the crowd has not found yet; when an asset is on the front page every day you have no edge, and a mania is not to be shorted either — it 'ends when it ends, and not a moment sooner.'

A rare signal — the index going from a 21-day closing low to a 21-day closing high in 4 days — is a coin toss across all 18 historical occurrences, but adding one orthogonal condition (the thrust firing within 2% of a 52-week high) flips it to 9-for-9 higher a year later, mean +13.28%. The method: define the signal mechanically, check the unconditional outcome honestly, then condition on regime rather than on the signal itself.

A fortnightly purchase schedule with a drawdown escalator keyed to the index rather than to your own holdings: S&P 500 down 10% invest 20% more, down 20% invest 50% more, down 30% double. Requires new income arriving. Lump sums are split into 52 fortnightly instalments over two years, with the escalator overriding.

Kris Heyndrikx's philosophy: buy growth stocks and hold them so long that they become value stocks. Amazon and Netflix were expensive growth names in 2016, are growth-at-a-reasonable-price now, and the value stage is where Alphabet sat when the market believed AI would kill search. The return comes from refusing to sell at each transition.

A stock sold off alongside a theme it doesn't actually belong to — the mispricing is a filing error, not a fundamental one. Expedia dumped with the 'SaaS-pocalypse' names: 'it's not SaaS, it's travel.'

Read the whole report including the frightening parts, extract every factual assertion into a list, discard the rhetoric, verify each item independently, then act on the classification rather than the tone. Heyndrikx did this to three Citron reports on Shopify — 'a lot of insinuation, but no substance' — and bought more each time. It is only a real procedure if it can also produce a sell.

Holding deep in-the-money calls instead of shares for the same directional exposure at lower capital outlay, then rolling to a later expiry to extend a profitable long. Identified in the $70M RTX trade by open interest showing one leg closing and one opening.

Kris Heyndrikx's first screen is whether he genuinely wants to know more about what the company does, and it eliminates 90% of ideas. The defence is operational rather than aesthetic: he writes a fresh deep dive on every holding each quarter, and boredom makes a decade-long hold impossible. Only then come 20%+ consistent revenue growth and management quality.

Shortage produces record profits, profits fund a capacity wave, capacity lands as demand cools — 'you're going to blow them up, but they're going to come right back and give it all back one of these days.' Forty years of it, all the way back to Texas Instruments.

Their label (deliberately echoing MBS/ABS, and 'or bullshit') for AI-capex paper once the buildout stopped being self-funded. The dating rule is the funding mix, in order: own operating cash flow with buybacks running, then all of cash flow with buybacks stopped, then cash flow plus debt, then plus equity issuance ('sprinkling some equity in there, like Google'). Reaching the debt rung tethers the equity story to the credit market through the same handful of levered names.

Crowding generated by AI models rather than by analysts: retail and pod traders ask the same models the same framing question ('find me the bottleneck in photonics / power / memory'), get the same names back, and build the same baskets — so 'they think they're having alpha, but actually they're just doing momentum investing in a different word.' The damage function is the leverage attached to the copy: Leopold Aschenbrenner's 4x-levered blow-up was copied wholesale and 'they all blow up together.' Its institutionalisation shows up as broker AI portfolio tools and multi-manager platforms partnering with model providers.

Where opportunistic capital is raised and idle, a lender's refusal to roll produces a transfer rather than a default — the paper moves at 200-300bp wider with covenant blockers on LMEs and EBITDA add-backs. Read the new terms as the real mark; the loss shows up as gradual NAV markdowns and compressing fund returns.

Instead of forecasting whether credit stress happens, find the contractual moment that removes everyone's option to wait — then back up the calendar by the normal lead time. $270B of sponsor-held 2028-29 software maturities means the negotiation starts "in the next two quarters, three quarters."

Schorr's rule: "When everybody agrees on something, trading's not going to be so good. And when we don't agree trading's going to be pretty good." Forecast trading revenue from the dispersion of opinion, not volumes — Q2 equity revenues +68% on volumes up only 9-10%.

A committed dividend forces management to allocate capital with discipline, and forces shareholders to hold long term - Hamlin's two-sided case for dividend growth investing.

Companies with good long-term records sold off 10-40% on a temporary headwind or theme (not an AI play, or feared AI disruption) while earnings estimates barely come down - a hunting ground for dividend growers.

A retail broker's most lucrative revenue line: uninvested customer cash is swept to a bank or money fund and the broker keeps the spread. It only works while the customer never optimizes — "the business model is dependent upon basically the customer being lazy" (Eisman) vs "they don't have to necessarily open up to agents" (Worthington).

Why a marketplace's advantage is local, not global: a ride-hailing or delivery network needs enough supply within each individual city to deliver an acceptable wait time, so a challenger must re-earn density market by market while the incumbent already paid for it everywhere. The reason Carlson thinks Uber can adopt self-driving cars years late and still lead.

Grade a market by its reaction, not by the headline. Write down the textbook response before a scheduled release; if the print lands and the market is muted or moves the wrong way, that is information about positioning and control rather than about the data. A negative payrolls print leaving bonds flat — 'this never happened' — is the tell that the natural buyer is absent, or that someone is holding the market. Scheduled catalysts make the test free and repeatable.

Beyond passive's usual mechanical index bid, issuers now game it deliberately: mark the company up in the private market, list it into the index 'at a stupid valuation, and the passive flows will maintain that stupid valuation.' The result is 'peak passive' — trillion-dollar companies entering the index straight off an IPO — where the support under the price is mechanical rather than anyone's judgement of value, and therefore disappears when flows reverse.

Unlisted assets have no daily price, so apply the peak-to-trough de-rating of the cleanest listed comp to the private cohort — ServiceNow -50% implies PE-owned software is worth about half what was paid. It also exposes the substitution managers make: answering "is the borrower performing?" when asked "is the collateral still worth the loan?"

Markets are drifting toward round-the-clock trading while the traditional back office cannot settle on weekends. A digital dollar that moves 24/7 can become the settlement leg of those trades — a use case that doesn't require winning consumer payments.

A four-rung read on whether capital can still absorb supply, in ECM and DCM alike: big pop and run, then pop and fade, then a weak pop, then breaking issue price on day one. It works because the marginal buyer of new paper is the flipper, who must sell something to make room when too much arrives at once. Companions: tails in the auction, rising new-issue concessions, and the desks' new-issue-versus-secondary tracker — with outstanding bonds trading poorly ahead of announced supply as the leading edge.

If a crypto thesis rests on usage of the underlying chains, then "half the market's Bitcoin" is a contradiction: the largest constituent is the one with the least usage. The view is only expressible as a rotation away from the market-cap-weighted version of your own thesis.

A crypto "exchange" like Coinbase is simultaneously the exchange, the broker, the prime broker and a market maker — roles that traditional finance separates across firms and regulators. One entity therefore sits on every side of the customer's trade.

A high dividend yield that exists because the market expects a cut - e.g. a usual 3-4% payer drifting to 5-7%, or a dividend raised on a boom the business can't sustain (UPS). Check balance sheet and free-cash-flow coverage before buying yield.

When net cash exceeds market capitalisation, the market is valuing the operating business below zero. Singh's three gates before it becomes a trade: the quarterly burn divided into the cash (your clock), a named and dated catalyst converting cash to shareholder cash, and management's history of returning capital rather than acquiring with it. Worked example: Nano Dimension at a −$80M EV.

When a company holds a stake in a private business, an up-round or IPO markup flows through the income statement as profit even though nothing was sold. Alphabet's Q2 EPS of $9.11 was only $2.88 operating; $6.23 was markups on Anthropic and SpaceX. Strip them out at Alphabet and Amazon and S&P Q2 growth falls from 47.4% to 28.0% — and the gains cannot repeat without another raise, creating a dated 2027 earnings cliff.

A software-company health check: revenue growth rate plus free-cash-flow (or operating) margin. Above 40 is considered excellent; Palantir printed 155 in Q2 2026 on 93% revenue growth and a 63% cash margin.

Two companies can pass the same durability test and produce results an order of magnitude apart. Pfizer (founded 1849) returned 9.8% a year since 1990 against Eli Lilly's (founded 1876) +17,000%, on near-identical durability arguments; Disney survived a decade of flat returns. Always pull the ten-year return alongside the multi-decade one.

The price a stock traded at immediately before an activist stake was disclosed. Singh's rule is never to chase the announcement spike but to wait for the shares to fade back to that level, so you own them ahead of what the activist actually delivers (board change, buyback, asset sale) rather than paying for it in advance.

A company with excess cash retiring its own stock year after year: fewer shares outstanding means each remaining share owns more of the same assets and cash flow, so per-share value compounds without growth. Polomny's stated reason for buying Athabasca Oil.

The late-1970s label for government bonds whose yield sits below inflation, so holders are repaid in money worth less than they lent — the investor's-eye view of financial repression.

Charging for a delivered result — a qualified lead, a resolved ticket — instead of per user or per unit of compute. HubSpot cut FY26 guidance as the shift lengthened sales cycles; Figma sells AI credits; Klaviyo bills autonomous resolutions.

A GAAP charge forced by an accounting rule (immediate expensing of acquired in-process R&D and upfront payments, impairments, mark-to-market) is bookkeeping about assets you still own — not evidence the operating business deteriorated. Test it against other lines in the same release: guidance raised alongside a headline loss falsifies the deterioration reading. The optical damage persists as long as the number sits in screeners and trailing multiples, which is why the mispricing survives.

In a layered technology stack, when the price of one layer collapses, demand rises for every other layer — because a complement just got cheaper. Pernas uses it to argue open-weight AI models are bearish only for frontier labs and bullish for chips, clouds and applications.

Portfolio construction that matches position size to payoff shape: core 5–15% (~70% hit rate, return comes from being right often), starter ~3% (provisional, can graduate), speculative 1–3% (little downside protection, played for multiples). Cash is the residual of research output rather than a macro call.

Booking the whole transaction and expensing the supplier's cut, versus booking only the commission — identical cash, very different reported revenue. Uber's UK Mobility switch cut reported Q2 growth by 8 points while gross bookings grew 24%.

Stock-based compensation should be read as an ordinary variable cost, not added back. A company whose growth slowed while SBC stayed at growth-era levels screens badly on GAAP for a mechanical reason; if SBC/revenue can fall quickly, GAAP earnings inflect with no change in the business.

A supply contract obliging the customer to pay whether or not it takes delivery — it converts a cyclical revenue stream into something close to an annuity. Cited as why Micron's $100B of bookings is 'visibility' rather than backlog.

A position requires a specific belief the current price contradicts. Two valid shapes: the market is wrong about a hated name, or the market is right about a loved trend but underestimates its strength — so being with the momentum can still be variant. Conviction scales with the size of the variance.

Keeping a roster of commentators who are reliably wrong and listening to them on purpose — the tell is a bear thesis that keeps changing shape (from 'multiple expansion, too narrow' to 'earnings bubble') while staying negative.

Earnings growth and profit margins are mean-reverting series, not trending ones, so an extreme print is a position on a distribution rather than a trend. Hay: S&P earnings the strongest since 1955, and "past times of earnings spikes have been followed by equally dramatic declines… at least some mean reversion is to be realistically expected" — a base rate that needs no catalyst to justify the direction, only to explain the timing.

A miner is a depleting asset: compare ounces produced against ounces upgraded from resource to reserve through the drill bit, over 20 years. Agnico's positive streak vs Hecla's negative record is the same test at both poles.

The longer a non-perishable thing has survived, the longer it can be expected to survive — applies to ideas, technologies and institutions, not to people. In investing it is a valuation argument: value is all future cash flow discounted to today, so a longer expected corporate life adds terminal value. Set against a base rate of 50% of businesses failing within 5 years and 80% within 10.

Rule's own bar for his account: a minimum of $10bn (preferably $20bn) of in-situ recoverable reserves and resources, plus a one-million-minable-ounce floor for gold — which structurally excludes antimony/vanadium/titanium deposits.

One producer shareholder is seen by rivals as having a head start and suppresses the premium; two create dynamic tension. Majors also don't trade, so the effective float is far smaller than reported.

A junior extending expiring out-of-the-money warrants isn't housekeeping: 'they need the money' — the hoped-for chain is warrant to stock to cash to salary.

A company's self-funded growth rate is the product of its return on invested capital and the share of profits it can redeploy: 25% ROIC × 40% reinvestment = 10%, while 10% ROIC × 65% reinvestment = only 6.5%. Screen on both terms at once; stated threshold is ROIC above 15% with a high organic reinvestment rate.

The necessary companion to 'markets work': a correction that must happen need not happen quickly. Conviction comes from the first, position size and patience from the second.

A product that is a trivial share of the customer's total spend but catastrophic if it fails can be repriced above inflation indefinitely without complaint. Diploma raises prices 3-5% a year because its components are ~0.2% of customer costs; the same logic underlies PPG's coatings.

Rather than paying fees on several of a sponsor's funds, buy the listed sponsor: indirect exposure to all of them plus a dividend. 'Would you rather pay or be paid?' The trade-off is owning an operating business geared to fund flows.

Software health check: revenue growth % + free-cash-flow (or operating) margin %. Above 40 means growth and profitability aren't being traded off against each other — so acceleration bought by burning margin is disqualified. Palantir's Q2 FY26 score of 155 (93% growth + 63% adjusted FCF margin) is the extreme case cited against a ~80x FY26 EBITDA multiple.

TCV (Total Contract Value) is the value of contracts signed during a quarter; RDV (Remaining Deal Value) is contracted revenue not yet recognized. When both compound faster than reported revenue — and especially when RDV jumps sequentially — the guidance raise is already banked before management announces it. Palantir: record $2.13B US Commercial TCV (+153%) and $6.24B RDV (+124% Y/Y, +27% Q/Q) ahead of a ~$500M FY26 raise.

At an 8% discount rate, an eleven-year delay to first production destroys 100% of a project's net present value — which is why permitting record, not geology, dominates the valuation of an undeveloped deposit.

Contracted customer commitments are the only forward number that is a signed obligation rather than a forecast. Backlog growing faster than revenue while the company already delivers at full speed signals demand added faster than it can be consumed — the opposite of a pull-forward.

A seller takes stock or warrants in its buyer as part of a transaction, books the sale as revenue, then marks up the stake and books the gain as income — counting one transaction twice. 'Legal but… extremely unsustainable,' and the reason reported AI-era earnings need discounting.

The remaining usable life on a jet engine; FTAI manufactures it by tearing down older engines and rebuilding them into swappable modules, per Crossroads' pitch.

Decompose a falling stock into earnings decline vs. the multiple the market pays. A 100x P/E "is fine so long as investors are not fearful of anything in the future" — when a structural change (lost regulatory moat, a partner turning competitor) introduces fear, the de-rating is correct and the lower multiple is the new fair one, not a bargain.

The Defense Department lending arm which, together with DOE programs, holds more than $350 billion in lending authority now open to permitted U.S. copper projects — lending authority, not appropriated grants, is the real ceiling on a buildout.

A derivative with no expiry, popular in crypto; GreensKeeper's ICE pitch explains why it does not threaten exchanges whose users need fixed settlement dates, standardised contracts and central clearing.

Lynch compounded ~30% a year for 13 years with only about six of ten picks going up — proof that hit rate matters far less than the size of the winners relative to the losers.

Haymaker's stated house yardstick: use the price-to-sales ratio for cross-cycle valuation because sales are far less volatile than earnings, and because revenue is much harder to inflate with accounting than reported profit. The two metrics disagree exactly when it matters — peak-cycle margins make the P/E look reasonable while the P/S screams.

A deeply discounted rights issue that dilutes non-participants heavily and passes unexercised rights to the backstopping holders (Alluvial on McDermott, templated on Garrett Motion 2021).

A loser can only fall 100% while a winner can rise many multiples, so a well-built portfolio's winners outweigh its losers — and no single position should ever be able to collapse the book. Carlson's own case: his worst-ever pick costs $13,000 against $500,000+ of lifetime gains.

Polomny's written mandate filter: three admissible types — early capital compounders reinvesting above-average returns; blown-up/mismanaged names where an event turns them (new management, business inflection, or a commodity cyclically bottoming); and event-driven specials (spinoffs, bankruptcy re-emergence, legal outcomes, management change). Time is the binding constraint: "any company that does multibag needs years for that to happen."

When conviction is in the theme but not in which operator wins it, buy both — equal dollar amounts at the same moment. Same-day, same-size entry makes later divergence information about the businesses rather than about your timing.

Decompose a crash into the vulnerability (extreme valuation, which is measurable) and the catalyst (a rate spike, which is not predictable). Position on the vulnerability and treat the catalyst as an alert condition — Hay's reading of 1999: 'outrageously high valuations had put the conditions in place for a waterfall decline. The rate surge simply applied the coup de grâce.'

A supplier guarantees or lends the money its customers use to buy its own products, flattering current revenue while transferring the customers' credit risk onto the supplier's balance sheet — the structure behind Nvidia's CDS doubling from 40 to 80 bps.

Selling far-dated, low-strike puts after a distressed name has refinanced and its option volatility has blown out; the premium collected lowers the effective entry to strike minus premium — a way to own a risky equity well below the market price.

Never sell a deep-moat compounder at merely 'overpriced' — only at a valuation 'so extreme that you cannot justify it'. Pabrai's calibration: ~50x trailing earnings is not egregious (Costco never has been); ~250x trailing normalized earnings is.

Fix the required return (10% here) and solve for the price at which the model delivers it. Fair value becomes a pre-made decision threshold rather than a valuation opinion - which is why nearly the whole book screens undervalued after a market fall.

When price is set by margin clerks rather than analysts — a levered fund's book being transferred, retail margin calls, CTA selling — the clearing of that overhang typically marks a near-term low, provided the underlying fundamentals never rolled over.

The upgraded Dhandho structure: value the provable cash-generating core alone; if it's worth 2-3x the price and pays a dividend while you wait, the expansion is a free option — 'it's not even tails I don't lose much. It's heads I win, tails I win.'

Unlike a batter, an investor is never penalised for not swinging — 'I can let 10,000 balls go by' — so wait for the pitch that 'looks like a watermelon' and never invest anywhere until you're '5,000% all in'.

Aggregate free cash flow of your holdings divided by their aggregate market value, recorded every year, so you can ask whether the book as a whole is cheap versus its own history (5.8% now vs a 3.3-5.6% range since 2015).

Only buy what you can explain to a 10-year-old in about four sentences, convincingly. If it needs more, it isn't simple enough for a large, infrequent bet.

2005-2011 owner's earnings +10%/yr vs prices +6%; 2012-2014 +16% vs +28%; across 2005-2014 both compounded at 12%. The convergence is the claim; the seven-year divergence is its cost.

Default-reject any idea whose 3-5 year outcome you can't answer with high confidence, however good the present looks — 'everything goes in the too hard pile; it's only the anomalies that don't.' Pabrai frames using it aggressively as an exercise in humility against ego.

Valuing a thin-margin marketplace on the trajectory of profit per transaction rather than its current level — DoorDash earns under 2% per order today but that figure was minus 38 cents in 2022.

A view that differs from consensus and is backed by repeated evidence rather than contrarian instinct — Gromen's test: write down the consensus playbook, count how many times recent history actually contradicted it, then find the mechanism forcing the repetition.

VIX prices expected index volatility, VIXEQ the average expected volatility of its constituents. A wide gap means stocks are moving a lot but in opposite directions, so the market is trading company by company rather than as one theme.

Leveraged and inverse funds reset their exposure every day, so choppy markets grind the value down regardless of whether the directional view is right — illustrated by an Irish-domiciled 3x SK Hynix product falling 96% in a month.

A chess position where you must move and every available move worsens your position — Gromen's frame for policymakers now that stocks back the Treasury market (a sustained 20% equity fall blows out cap-gains/stock-comp receipts) while the Treasury market backs stocks.

Scan licensing disclosures for upfront fees trivial against the addressable category (EUR 50,000 for a product in one of Canada's largest chronic prescription categories). The fee is the option premium; cross-check management's peak-sales guide against independent prescribing data and carry a range, not a point estimate.

Build the forecast so it works without the disputed variable (public reimbursement, an out-licensing deal, further M&A), and keep every excluded branch on a separate free-option list. Check the exclusion is a stated strategic decision with a price test, not an oversight.

Carlson's reason capex is an opportunity not a cost: the spender buys 'an entire infrastructure layer, of which they'll be the demand aggregators' — everyone wanting the service routes through their products, so the aggregator captures the pricing rather than the supplier.

Equal-weighting a sector basket protects against any single constituent blowing up, but it overweights the small members and therefore amplifies whatever they are disproportionately exposed to. Hay's worked case: KRE's equal weighting "diversifies deposit risk" yet "actually concentrates CRE risk," because smaller banks carry heavier commercial-real-estate books and get the same vote as the giants.

Under equity accounting a parent books its share of a JV's output as a purchase (Cameco: at a 5% discount to spot); the real benefit appears separately in equity earnings and only becomes cash when the JV declares a dividend — typically a full year later, so 2026 output is 2027 cash.

A drug company that never discovers anything — it buys the rights to products already developed and approved elsewhere and sells them through its own salesforce. Removes discovery risk and the patent cliff, leaving only commercial-execution risk; converts an un-analysable scientific bet into an analysable sales bet.

One sell-side analyst and a few thousand shares a day means institutional capital structurally cannot participate — which is why the mispricing persists. Illiquidity is the source of the edge, but it dictates a multi-year horizon, sizing against daily volume rather than conviction, and ignoring price moves on light news.

Selling an internally developed component externally expands the addressable market but erodes the exclusivity that made it valuable — worth doing when the moat is the surrounding system rather than the part. Amazon weighing Trainium sales to firms running their own data centres.

A regulatory classification, not a quality judgement, that determines who can pay: a prescription-classified product is eligible for provincial formulary and drug-plan coverage, while a natural-health-product competitor structurally cannot access it. Rank competitive risk by classification before product quality.

A bank's core profit engine: the spread between what it earns on loans and pays on deposits. "For a bank, everything flows from net interest margin" — Hay treats a confirmed group-wide NIM turn (deposit costs falling while loan yields hold) as the single most important operating signal, and insists on a reported quarter rather than guidance.

The discipline of asking whether an anomalously cheap security is cheap because you are missing something — "a universal constant across all sectors of life."

Add back mechanical distortions (a promotional event moving between quarters, extra weeks, FX, divestitures) before calling a guide a slowdown, then cross-check against the profit guide — Amazon's Q3 +9-12% revenue guide loses ~4 points to a Prime Day shift and 80 bps to FX while operating income is guided +40%.

Before believing an EPS miss, itemize what analyst models didn't contain — legal settlements, severance/restructuring, and changes to the depreciation schedule — add them back, then cross-check the operating lines (revenue, users, margins). Meta's 14% miss = $2.4B legal + $1.18B severance + faster-than-peer depreciation; underneath it was 'in line'. Run it in reverse too: strip equity-stake gains out of Amazon's EPS.

A derivative with no expiration date and no delivery of the underlying — so it cannot hedge a commercial exposure; CME argues it is legally a swap, not a future. Retail-oriented, with 50x+ leverage offshore.

An unsupported assertion that makes an argument seem logical at the start, with everything thereafter resting on a false foundation — e.g. "it's pulled back from its risk-on highs and consensus estimates have bottomed out."

Banks trade on a multiple of tangible book, and that multiple is a function of the return earned on it. "A bank earning ~14-15% ROTCE justifies a materially higher multiple of book than returns stuck near 11%" — so a rising ROTCE makes a re-rating earnings-driven and durable rather than a sentiment-driven momentum blip.

A single-product company carries a 'one-trick pony' discount. Dilution of that concentration by adding new pillars (not by the core shrinking) is a re-rate — and the edge is the reporting lag between the structural change and the quarters in which it prints.

In a shortage, buyers pay up AND order early, so a supplier's volume is partly one-time. When supply catches up, price and volume can fall together — why Carlson keeps only a 'leg' in semis and refuses to concentrate there.

A headline pipeline count is an addressable market, not an order book. Grade it by stage (origination / front-end engineering / early services / long-lead ordering / construction) and watch backlog as the line where conversion actually shows up — Westinghouse: 91 opportunities but only $0.8bn New Plants backlog against $13.2bn in Operating Plants.

Bank cheapness is measured on price-to-tangible-book, not EV/EBITDA or price/sales. Hay anchors the range with three points — the crisis trough (sub-1x, 2023), today (~1.4-1.5x) and the good-times peak (~1.7-2.0x) — converting "cheap" into a measurable distance the earnings recovery has to close.

The discount applied when a security's value realization is too far away or too indeterminate for a short-term, relative-return-based asset-management industry.

Discounts imposed on companies for non-economic reasons — too small, too illiquid, or shrinking their own float via buybacks — because index funds structurally cannot own them.

HK's name for explaining compounding through analogy and charts rather than math — including the fractal drawdown check: magnify any earlier window of a long-term chart and the volatility is identical to today's.

Split a capital budget into short-payback equipment bought close to deployment (servers: break even under three years against contracts of five or more) and long-life structures built ahead of monetization (data centres: ~2 years early, 30+ year life, five or six server generations). Negative free cash flow on contracted demand is a timing mismatch, not speculation.

Before valuing any company, count the durable winners analogous platform categories produced (search → one, e-commerce → one) and use that count as a ceiling on the current theme; if the count exceeds one, find the axis the winners split along (usually customer type, not technology) and treat whatever is left over as the structural loser.

Old NYSE-specialist term for the single large block trade in which a forced seller liquidates the remainder of a position at once. Because it removes the overhang in one go, it typically marks a short-term low in the affected names — 'they have dumped all their listed investments through one enormous trade… so expect semis/AI to cop a bounce.'

Ranking companies by cash generated against the full cost of buying the business (equity plus debt), then filtering for expected growth - the rule behind the VFLO ETF.

A basket built to move as little as possible; bought here on a spread rather than a forecast - high-vol stocks were beating low-vol stocks by the widest margin since 2006.

A share rising with no improvement in the business - Apple +60% from 2022-25 while free cash flow fell, because the market moved from paying 20x cash flow to 40x.

The extra spread an issuer must pay above its own secondary curve to get a new bond sold — 'new issues will always come at a concession to secondary (gotta have an incentive to play).' Its level, tracked weekly, is a regime detector: in credit bull markets new-issue spreads tighten toward secondary; in bear markets secondary widens toward new-issue levels. UBS: 7.0bps in July 2026 vs a 3.1bps YTD average.

Order book divided by deal size on a new bond issue — a measure of book depth that deteriorates before spreads do. Falling from a 4.0x 2026 YTD average to 3.1x in July 2026 is the 'falling off the cliff' tell Paulo's credit contact flagged.

Why the bond market's marginal buyer is not who you think: more AUM is passive in fixed income than in equities, and even nominally active IG managers hug 6.3-7.0yr against a 6.6 index duration because 'anything outside of that is too much career risk.' New issues enter the index only at month end, so the real buyer of new paper is 'active guys and fast money hedge funds' — which is what makes the primary market a clean read on risk appetite.

A physical ETF actually holds the underlying shares; a synthetic one obtains the return through a swap and carries counterparty risk.

Returns come from cash flow growing and from the market paying a higher multiple; the two multiply rather than add - Caterpillar 2019-26: FCF +86%, multiple +186%, price +536%.

Before crediting a huge backlog as visibility, ask for its growth excluding the largest counterparty and check who added commitments sequentially — Microsoft's $678B commercial RPO still grew 25% excluding OpenAI, with all sequential growth from non-frontier customers.

Extending an asset's estimated useful life slows depreciation and lifts reported profit; shifting leases from finance to operating removes commitments from reported capital expenditure entirely. Cash committed is unchanged — comparability is not. Microsoft, effective FY27: 15 → 25 years plus a finance-to-operating lease shift.

Deliberately dropping the requirement for a trigger: 'at the risk of sounding glib, often we'll say and there is no catalyst — but if the stocks are cheap and the businesses are good, that's when we want to be active.' The catalyst is unforecastable and unnecessary to the decision.

A high-return business that can also redeploy its own cash flow at a similar rate for a long time; high ROIC without reinvestment capacity is only a cash-return story.

Meet management, then discount them: 'management teams can get too close to their own story… they'll be the last to know when the bad news finally arrives because they're so close to it and can't see the forest for the trees.' Reassurance is weakest exactly when it matters most.

Buy at a large discount to your own estimate of value so that even a wrong thesis (slower growth, real threats) still leaves an acceptable return; Aitken contrasts it with the buyer at 30x EV/EBITDA who needs the world to come in perfectly just to earn a mediocre return.

When realised returns (15-20%/yr) far exceed what the fundamentals delivered, the gap is multiple expansion — not new wealth but future return consumed early. Lower your forward expectations accordingly: 'we're borrowing from the future.'

Great business + management with skin in the game + the ability to reinvest internally at high rates. Fail any leg and it is not a compounding machine.

Separates the probability question from the sizing question: the downside is capped at -100% of the position, so choose the position size as the loss you would fully accept, then buy on a schedule. Brown's example is a 5% Bitcoin allocation whose single most likely outcome he says is zero — justified only by a possible 1,000-10,000% upside against a fixed 5% max loss.

A dating framework, not a price forecast: on a log chart Bitcoin's major tops sit almost exactly four years apart (Dec-2013, Dec-2017, Nov-2021, Oct-2025) and the bottoms roughly one year after each top (Dec-2018, Dec-2022) — so it projects a window (late 2026 low, ~Sep-2029 top) rather than a level. Held explicitly as a conditional and only acted on when an independent mechanism points at the same window.

Instead of judging a multiple in isolation, back out the implied duration of the cycle each price requires, then look for two names in the same supply chain sitting at opposite ends. 'Micron sells like it's six times earnings… as if the cycle is over,' while Intel at 100x is priced 'as if the cycle is continuing for five more years.' One of the two prices has to move.

Hay's entry discipline: spot the multi-year breakout, measure how extended it is above the breakout level and the 200-day, then wait for the retest rather than chase — "I like to look at breakouts and then buy them when they pull back after they break out. The nirvana."

Two independent discounts multiplied: a physical trust priced ~10% below the spot value of what it holds, where spot itself sits below the long-term contract market where nearly all volume actually trades (uranium ~$77 effective vs $85 spot vs ~$95 contracts) — a ~20% total discount, with contract floors near spot capping the downside.

Inverts the usual reading of ETF flows: large inflows AFTER large gains are the warning sign (gold miners, 2016), while record OUTFLOWS after a profitable year signal investor apathy rather than information — the condition that precedes a re-rating.

Watch the yield gap between BB ("high-grade junk" — under 1% annual defaults, price-volatile but rarely defaulting) and CCC ("right on the door of default"). A widening gap flags stress even when junk-vs-Treasury and investment-grade spreads look benign, and the spread itself can be charted for multi-year breakouts.

A four-variable stress composite — Brent crude, the US 10-year yield, Strait-of-Hormuz vessel transits and the S&P 500 level — used to predict when a president will intervene to push energy prices and yields down. Backtests put the trigger at a combined 2.3-3.4 standard-deviation move (avg 2.9σ); it called a de-escalation window of Jul 26-30 2026 and the halt came Jul 27.

A customer using two services on one platform can be worth a multiple of a single-service customer, and is far cheaper to acquire from the existing base than through advertising — Uber cites ~3x the gross bookings and profits at >50% lower acquisition cost, the arithmetic that justifies buying delivery density rather than building it.

An acquirer's post-synergy multiple versus its headline multiple is the execution risk being asked of the buyer — Uber's ~14x EBITDA for Delivery Hero falls to ~8x 2027 adjusted EBITDA only if $1.2B of run-rate synergies land within 18 months of close. Cost synergies from a tech-stack migration are more credible than revenue synergies.

Shorting a stock you expect to go nowhere purely to raise cash for a high-conviction long: borrow and sell the share, deploy the proceeds into the name you want, and aim to break even (or make a little) on the short while the long carries the return. Distinguished from a directional short, where the profit is the price decline less the borrow fee.

Gold-mining-equity exposure timing model with three modes — Offensive / Neutral / Defensive — combining contrarian market indicators with mining-margin fundamentals; turned Defensive ahead of the Q2 2026 miner drawdown. Polomny calls it "interesting and useful" but publishes a critique: internally consistent backtest, not yet independently verified alpha.

Itemize warranty releases, tariff relief and accounting true-ups inside a margin beat, restate without them, and write down the prediction for the following quarter — App Economy Insights flagged Tesla's Q1 beat as one-time-driven and the reversal (auto GM ex-credits 19%→16%, energy GM 40%→20%) arrived in Q2.

Growth from locations open at least a year, stripping out new openings — "the best indication of growth for a consumer-facing company." Eisman uses it to override a headline revenue beat: Domino's comps of +0.1% (weakest in five quarters) invalidated the print.

Pre-agreed divestitures and a distant target close are a deal's own admission of antitrust risk — Uber pre-sold 14 overlapping Delivery Hero markets to SSW Partners for ~$1.6B and targets a 2H 2027 close: "Brussels won't wave this through."

His name for the rare second entry into a big move: the fundamentals are unchanged or improved, but the first wave of longs was stopped out and won't return — so positioning is clean and fear is high, and the trade works as a classic wall of worry.

Sentiment is largely a function of the tape: when share prices erode, bearish explanations are manufactured after the fact — so audit which cited causes are fundamental and which are flow/rates/beta.

Aggregate every institutional fund's quarterly 13F buys/sells across a sector; a record net inflow led by sticky long-only (not hedge-fund) money is an early re-rating signal.

The % of new well licenses filed "confidential" proxies how much true exploration risk operators are taking; a falling share (to 48%) signals development-over-exploration = lower volatility.

Expected annual return = EPS growth + dividend yield + the annualised effect of the multiple moving to a stated exit level. Compounding Quality's second valuation method, and the one that defines its Fair Value.

Net ETF redemptions translate into a dollar figure of forced, price-insensitive selling; Huhn computed -$41.6M across URA/URNM/URNJ in June 2026.

Split total capital spending using depreciation & amortisation as the maintenance proxy; growth CAPEX = total - D&A. Eli Lilly fails the capital-intensity test at 12.0% of sales and passes at 2.8% once only maintenance is counted.

"Industries operating at low rates ... where you won't see capacity increases for at least a couple years, and where the profit margins will be much higher by then" — quoted by Huhn as the frame for the uranium equity universe.

The scheduled date a drug loses exclusivity and generics enter; Trulicity's is expected around 2027, against a replacement (Donanemab) that is not yet approved.

Hunt for sectors where new supply has structurally stopped while a genuinely new demand class arrives; cross-check with replacement cost. Paired rule: it is far easier to track supply than to forecast demand.

The mining-equity lifecycle — a discovery run, then a long dull drift through construction with no news flow, then a re-rating at production; "the boring part" is the accumulation window.

A temporary dislocation — management turnover, a pricing-model question — in a business that isn't structurally challenged; the discount is the queue of large investors waiting to "meet the new CEO." Corollary: "if you wait for the certainty, you pay a high price for the certainty."

The price only moves when a buyer or a seller gives in and crosses the spread; prices are set by whoever is impatient, not by whoever is right.

Ownership is a stock, price-setting is a flow. Index funds hold over 60% of assets but active managers do only 10% of trades (down from 80% in the 1990s), so the marginal price is set by price-insensitive money.

An insurer's claims and expenses divided by premiums collected; below 100% means the underwriting itself makes money, so the investable float is genuinely free.

The options market's own estimate of how much stocks move together; a low reading (with single-stock vol above index vol) means the market expects a shock absorbed sector-by-sector rather than index-wide — why the VIX can stay low while oil, rates and single-name vol carry a war's risk ("Bayesian updating, not complacency").

When a resource gets cheaper, total demand for it can rise rather than fall: cheaper AI models free up budget that gets reinvested into far more usage, so cheaper compute drives MORE chip/memory demand — a ~32% reinvestment rate is hyperscaler breakeven. Singh also applies it to engineers (cheaper output → hire more).

Ask whether a competitor could get permission for the scarce asset, not just afford it. Copart's salvage yards are protected by zoning and environmental rules, and it owns them while rivals lease.

Outsourced HR, payroll, benefits and retirement administration for small and mid-sized employers; the model behind Paychex and Insperity, and the source of its switching costs.

A PP&E subaccount where not-yet-in-service assets (e.g. GPUs not plugged in, capitalized interest/labor) sit undepreciated, deferring write-downs and flattering earnings; a fast-growing CIP balance flags deferred depreciation.

Track which operating metrics a company stops publishing or downgrades in frequency, especially one previously promoted as its North Star, and ask what outsiders can no longer verify — Netflix retired subscriber counts, moved its engagement report to annual from 2027 and decoupled it from earnings, and has never disclosed churn.

Chanos's bull/bear mental model: in bull markets investors pay a premium for future promises; in bear markets they apply a discount to present reality. A gauge of where sentiment sits in the cycle.

The profit earned on each NEW dollar of capital (incremental operating income / incremental capital), not the flattering blended average — a leading tell that a capital-heavy boom is maturing as it trends down toward the risk-free rate.

Book the break-up fee received but also write down the optionality surrendered, and check the cash-flow statement for the tax drag — Netflix collected $2.8B for walking away from Warner Bros. Discovery, lost its shortcut to the next leg of growth, and saw Q2 FCF fall to $1.5B from $2.3B on the resulting cash taxes.

Sponsors underwrite EV ÷ adjusted free cash flow because debt service is paid in cash — Stripe and Advent's $60.50/share (>$53B) bid for PayPal is ~8x adjusted FCF. An unsolicited first bid is an option-opening event that sets a floor, rarely the clearing price.

Judge a content spend on the metric it actually buys: Netflix's live programming takes just over 5% of content spend and drives ~1% of view hours — indefensible on hours, but it produced six of the ten largest new-member sign-up days in five years. It is buying acquisition, not engagement.

Audit a platform on an independent measure of the finite resource it competes for, compare each name to its own prior peak rather than last year, and classify the share gainers by business model — Nielsen's US TV-time data put Netflix at 7.8% against a 9.0% December peak while YouTube, Prime Video and Tubi all set records.

An AI study of ~650 releases from 40 Athabasca juniors over five years: day-one hype volume is largely holders exiting; the noise is gone by day three and the price settles by day five. 60% of releases quoting counts per second never published the assays.

Rank exposure from least to most speculative: physical vehicle (SPUT) → ETFs → producers (track price) → developers (build risk) → explorers (discovery risk, 100–1000x upside, don't rise with the commodity tide).

Investing needs the vision to see a name, the courage to buy it, and the patience to hold it — and patience is the rarest; sitting on a winner (not the cheap buy) is how you compound after-tax.

A selloff-quality diagnostic: precious-metal prices set by 'paper' (ETF share redemptions, futures positioning) can crater on headlines while 'physical' demand (central-bank buying, bar/coin, industrial offtake) and supply deficits are unchanged — a drop showing only in the paper leg is a discount, not a demand break.

Total US market cap divided by GDP; at 241% vs 160% at the 2000 top — Hickey's headline valuation gauge for the bubble.

Suppliers book capex orders as instant revenue while buyers expense the same spend over 5-6 years — inflating reported earnings early in a buildout and reversing when spending slows.

Substituting the cash your holdings earn on your behalf for their quoted value, divided down to the minute ($65,520/yr = $0.12/minute) so the accrual is visible while the price does nothing. Rests on Graham's voting-machine/weighing-machine distinction; the discipline only counts if it is published in good periods too.

10-yr inflation-adjusted earnings multiple; reported ~40 (matched only in 2000), which Hickey argues is ~67 once the AI earnings distortion is stripped.

A monopoly/near-monopoly stock exchange is a toll booth on all trading in its country — steady, high-margin revenue that rises with volume, so it swings more than the index it lists (e.g. BOLSY's ~25% drop vs the Brazil ETF's ~15%). A deep discount to global peers (typically 20×+ earnings) usually reflects the home market's cheapness, not the company's quality; decompose the discount into 'business' vs 'country,' and find the macro lever (falling real rates) that would re-rate the whole market.

Why a recovering company's trailing/forward P/E is meaningless: as it climbs from operating losses back toward profit, near-term expected earnings (the denominator) are a tiny number near zero, so a normal price divided by it yields an artificially sky-high multiple. Value the multi-year free-cash-flow ramp instead (out-year FCF per share on the full share count, checked against enterprise value), then apply a quality-of-business multiple and haircut for execution risk.

When capital floods into an insurance line, premiums fall (soft market); weaker underwriters lose money and withdraw capacity, and prices recover (hard market). The investable question is whether a structural cost advantage lets your insurer survive the soft leg — Slegers buys Kinsale into one while refusing the memory cycle on the same reasoning.

Always ask who benefits from an assertion or agenda and assume self-interest — usually financial — drives the message; a filter for narratives, analyst takes and official statements.

Pro-wrestling term for staged conflict presented as real; his lens for reading geopolitical 'deals' and market band-aids (the Hormuz MOU / SPR releases) as scripted theater rather than genuine resolution.

AI capex books instantly as the equipment/cloud recipients' revenue and earnings, but the spenders depreciate it over 5-7 years — so reported earnings are optically inflated on both legs of the trade while the cash actually leaves now; a reason to distrust the 'E' in AI-complex P/Es.

His investing framework: read widely (especially history) to build a library of ecosystems and repetitive patterns, then reach for the closest analogy when making stock decisions — "the more I read about many different ecosystems, the more analogies I have to draw upon."

A discipline for sizing: can you model the next ~3 years of a company's cash flows with clarity? If yes it's investable (build a fundamental position); if the cash flows are 'too ambiguous' (a high-flyer already pricing out-year best cases, with competition/efficiency risk), it's only tradeable — hand it to charts/instinct and keep it small.

Think in decades not months: buy hard assets (timber/stumpage compounding 6-8%/yr), inculcate the whole family, stay low-profile — 'planting trees you'll never sit under.'

The classic solvency-vs-timing adage (Keynes-attributed): a mispricing can persist past the point your capital (or, in Harrington's tweak, your sanity) can outlast it — a caution against pressing a fundamentally 'right' call in a speculative tape.

How a cash-rich business should return capital, in order: pay off debt → dividend → special dividend → buybacks (the Amerigo playbook).

Templeton's rule is an absolute condition (maximum pessimism); a wide spread between two styles is only a relative one, and can resolve by the expensive side falling rather than the cheap side rising. Stating which condition you actually have is the discipline.

Include cost of capital and taxation in 'total cost'; when the market price sits below it, the industry must eventually earn its cost of capital 'or your car won't start' — so price rises or supply falls. Inside the sector, buy the reinvestors, not the high-distribution 'cannibalizers.'

Fold 80/20 repeatedly — the good 20%, then ~4%, then in juniors the ~1% of managements who generate most of the value — and demand serial success 'at the task at hand' (a gold winner may fail at oil & gas).

Own companies that farm out projects to partners who pay to drill, so you hold small stakes in many exploration 'lottery tickets' bought with other people's money — the only arithmetically predictable way to reach for a 100-bagger. Base rate: ~1 in 3,000 anomalies becomes a mine, cut toward ~1 in 50 by backing the right scientists.

Rule's summary from grading ~100,000 portfolios: money is made on the delta between price and value (do the work), a 10-bagger is a 5–6-year job (patience), and most 10-baggers inflict a ~50% drawdown you must hold or add through (tenacity).

A US bank holds a foreign company's home-market shares and issues dollar-priced receipts that trade on a US exchange like any American stock — how SK Hynix's Seoul-listed shares reach US investors as Nasdaq 'SKHY' (each Seoul share splitting into ten ADRs). Widens the buyer base to US funds/index trackers; the discount it closes is access, not price.

EBITDA flatters capital-intensive businesses — it strips out real interest cost and understates capex vs depreciation. Reconcile to cash flow, capex and interest; if depreciation is ~half of true capex, earnings/EBITDA are overstated (Enbridge case).

Track adjusted EBIT per head across years; profit-per-employee rising while headcount is flat or falling signals real AI-driven operating leverage.

Distinguishes ETF/futures/quarter-end positioning flows (the 'paper' market, which can crater price in days) from central-bank and physical accumulation (built over years). Quarter-end short-pressing into a close can force a 'reset' selloff that unwinds once the new quarter opens.

A vehicle that issues stock to buy an asset compounds only while it trades above the value of its holdings; once it flips to a discount the mechanism reverses and unwinds — 'roll-ups don't work well in reverse' (MicroStrategy).

Chart one theme's weight (% of the index), not price, over decades vs its history and prior bubble peaks; as it rises, 'everything else' falls by the same amount — a reallocation signal.

A regime where weak economic data pushes stocks UP, because the binding variable is the Fed's reaction function, not growth: softer jobs/inflation means fewer rate hikes. The sign flips depending on regime, so you name it before reacting — and verify the bad number is bad for the right reasons (e.g. unemployment falling only because people left the labor force is still dovish).

A hated commodity doesn't need to return to favor to rally, it only needs to 'lose its hatred to jump'; when nothing is hated, drop to the cheapest tier on price-to-fundamentals.

A systemic-risk pattern the BIS flagged: chipmakers, hyperscalers, AI labs and compute providers fund each other while booking future sales from one another, so the same dollars loop through the ecosystem and are counted as demand — making real end-demand hard to read and inflating the apparent size of the boom (much of it debt- and private-credit-financed).

GMV is the total dollar value of goods sold across a marketplace/commerce platform; the take rate is the percentage the platform keeps. For a business like Shopify, GMV growth is the rawest measure of demand and scales almost directly into high-margin revenue via the take rate — so GMV trend and any take-rate expansion are the two levers that drive the model.

80/20 is a bell curve (a different 20% create 80% of the aggravation); fold it twice and ~1% of junior-mining managements generate ~40% of the value — so back serially successful teams in their own lane.

A regulator's approval to manufacture a replacement part someone else designed. Each approval takes years and money rather than cleverness, so a stock of thousands of them (HEICO holds 20,000+) is a compounding barrier — and it is stable precisely because the customer also saves money versus the original part.

Canned pre-mixed cocktails growing 20-30%/yr in the US; a spirits-category format the traditional distillers ceded to brewers and private firms (e.g. AB InBev's Cutwater) because cans carry lower margins than bottles for a distiller but higher margins than beer for a brewer, who also already own the canning lines.

When the FTSE Russell rebalance mechanically pushes mega-caps across style/size buckets — e.g. the Mag 7 migrating into the Russell 1000 Value index (now ~17% of it) because their market-cap growth raised the bar for 'growth'. Value-mandated managers are forced into high-multiple tech, quietly erasing the growth/value diversification benefit; names can also be dumped purely for index-mechanical reasons.

When the market prices a consumer-staples category at cigarette-like earnings multiples on fears of structural, permanent volume decline. The analytical test is whether the volume loss is truly secular (moderation, GLP-1s, substitution) or partly cyclical/affordability-driven - signalled by cheaper and smaller formats still growing even as totals fall.

Companies that consistently retire their own shares so per-share value compounds even without a bigger business; archetype Henry Singleton's Teledyne (bought back ~90% of shares in the 60s–70s), with AutoZone as a 20–30-year example.

A country's swing from left-wing to right/center-right economic policy can lift growth and re-rate its beaten-down stock market; markets often rally in anticipation of the change (the Latin America case).

Companies that consistently buy back their own stock: a shrinking share count on steady cash flow makes per-share value and EPS compound. Buy below book/NAV, with an aligned, value-driven capital allocator. Teledyne / Henry Singleton is the archetype.

A metric that gauges purely speculative vs commercial commodity activity by adjusting the exchange futures margin (the collateral/"cost" of opening a position) for the option-implied volatility of the commodity — so cross-commodity speculative froth can be compared on a like-for-like basis.

Issued expensive high-multiple stock to buy ~130 businesses cheaply in the 1960s, then reversed to buy back >90% of shares (1972-84) when the P/E fell below 10; a 1966 buyer compounded ~17.9%/yr for 25 years. The template for buyback-driven per-share compounding.

The serial-entrepreneur value-creation model (a la Ross Beaty): make a discovery privately, advance it through permits/studies/tax reform to create value before any metal is mined, then either build it or sell to a major.

A calibration exercise: pick ten individual companies, seal them in a trust you cannot touch for twenty years, dividends reinvested. Removing the ability to correct a mistake changes the objective from maximising return to avoiding permanent loss — and the gap between that list and your real portfolio is the part that depends on you being able to change your mind.

Favor recurring-demand franchises (eyewear, medical implants) where once a customer is in they keep buying, with an aging-population/secular driver growing the base — accept durable GDP-plus growth rather than chasing hyper-growth.

Like the FAA changing rules only after a crash, build an investing checklist from great investors' losses; his has 213 questions — the three retail items: no leverage, durable moat, owner quality.

Buy only what the crowd despises; risk-free bargains cluster in boring, ignored sectors (e.g. regulated power at 3x earnings). 'If something is loved, don't buy it.'

A consensus the market is NOT confirming (a firm bullish view while price stalls) is a signal to get out of the way — 'a lot of people who are going to be wrong.'

Hohn's requirement: score a candidate against a fixed menu — intellectual property, brand, hard assets, long-term contracts, network effects, regulatory/switching costs — and count. One barrier is a thesis with a single point of failure. "Often you would like not just one barrier to entry but maybe five... Big jet engines have many of those."

Hohn's one-line formula: "if you can price 1% above inflation and you have a 20% profit margin, your profits will grow 5% faster than revenue." A price rise costs nothing to deliver, so it lands almost entirely in profit — and the thinner the margin, the bigger the multiplier.

Judge decisions on process, not results: good process + good outcome = deserved success; good + bad = tough beat; bad + good = lucky break; bad + bad = just desserts.

Paulo's trading maxim: the funniest, most absurd outcome is also the most likely one — a prompt to keep an open mind to tail scenarios.

A reflexive false premise — a narrative solving for price rather than reading fundamentals (Paulo's framing of the one-sided oil-bear case).

Incumbent software/medtech priced for AI/upstart disruption is often mispriced because the customer's real cost to switch — compliance plumbing, org-wide integration, surgeon habit — is enormous.

One-sided bearish (or bullish) jubilation on social media — mass dunking on the other side — read as a contrarian tell that the crowd is over-committed.

Chris Hohn's screen, adopted by Compounding Quality: a company controlling essential infrastructure or a platform that charges a small toll each time something passes through — high barriers to entry, structural pricing power, predictable cash flows. The test is technology-agnostic: change what passes over the toll (petrol car to electric, Visa to Mastercard) and the toll survives.

A cheap multiple alone signals nothing — diagnose whether the business is structurally eroding (trap) or has a fixable problem with the moat intact (turnaround); demand a real catalyst, not just a low P/E or strong brand.

When a vendor grants a big customer warrants (discounted equity) to win a deal, accounting treats them as a discount, so part of what the customer pays never shows up as revenue — understating the true size of the relationship and widening the gap as the deal ramps. Seen in the Cerebras–OpenAI >$20B deal.

Flip through ~100 charts across leaders, sectors, factors and assets to get the aggregate message fast and to find the names that contradict your view (as Chris Verrone and Jason Trennert do).

Beta measures how much a stock moves vs the market (Nvidia ~2x, semis high-beta, staples low-beta); high-beta-vs-low-vol at a historical extreme is a froth / positioning warning.

Framing a declining-but-durable business as an ice cube and asking how fast it melts (10 years or 50?) given Starlink, fixed wireless and fiber-to-the-home — the durability of the cash flows sets the value.

Total US stock-market value divided by US GDP, popularized by Buffett in 2001 as an overvaluation gauge. Carlson argues it's outdated — many large US firms now earn over half their revenue abroad and run higher margins, so it overstates today's valuation; Buffett hasn't cited it in 25 years.

Adding a quality income name marked down on a one-off charge (Verizon −6.5%, yield to 6.5%) — the lower price lifts the yield, the charge being cosmetic not a payout threat.

In a futures delivery month (e.g. July for COMEX silver), short positions must be closed, rolled, or covered around first notice; that covering is mechanical buying that can push the underlying price up — the same paper plumbing that sold the metal can re-buy it.

A hybrid security paying a fixed bond-like dividend (here ~7.7% to a 2029 maturity) that can also convert into the common — high yield plus 'tight equity-like' upside if the common is undervalued.

Options-based ETFs offering S&P exposure plus income (covered-call) or downside protection (buffered) — now a structural substitute for owning staples for yield/defense.

For a stock that's run hard, classify the move: structural re-rating or cyclical upswing? If cyclical, the operative question is the inning (early/middle/late) — buy on changed fundamentals + momentum if the cycle has further to run.

When aggregate options-dealer gamma is negative, dealers hedge by buying rallies and selling declines, amplifying volatility in both directions; positive gamma dampens moves. Sign flips can be triggered by an options expiry.

Value a cyclical/chip name on its out-year earnings power rather than the current print — e.g. Marvell $10 EPS by 2027, Broadcom ~$20 — then buy a bigger dip against that number.

Cumulative sector-ETF flows from an anchor date map where investors actually put money; one sector taking nearly all the inflow ('one game') is both the engine and the single point of failure.

Sorting stocks by characteristics rather than size; beware that 'quality' and 'momentum' factor ETFs have morphed into disguised mega-cap-tech bets as the S&P itself became a quality index.

In a heavy investment cycle the risk isn't margin contraction but negative free cash flow and the absence of operating leverage until the spending eases.

End-of-quarter brings volatility and confusion largely from index rebalancing and institutional reallocation (equities to fixed income) — don't extrapolate a few-day equal-weight-vs-cap-weight move into a regime change.

Deliver 2-3x daily exposure but reset every day, so they decay point-to-point from slippage + fees and add market volatility by rebalancing into up days — trading vehicles, not buy-and-hold. A ~$200B category.

2x single-stock products (e.g. 2x Nvidia) that didn't exist before 2022; now hundreds, increasingly on small / quantum names — 'kind of like a lotto ticket' and a sign of froth.

A business that benefits from a megatrend on the demand side while the same trend raises its dominant input cost is internally hedged against itself — e.g. Alcoa is long aluminum demand from electrification but short the power that electricity-intensive smelting consumes, the very AI buildout driving both.

A commodity's quoted price is set largely by paper claims (ETFs, futures) whose turnover dwarfs the metal that actually moves; a paper-led selloff (ETF redemptions, futures liquidation, algos) can crater the price in seconds without any change in physical supply/demand — so decompose a drop into its paper leg and its physical leg before reacting.

Annualize the dominant ETF's daily share turnover into ounces and divide by annual mine supply; the larger the multiple, the more the price is set by claims changing hands than metal moving. Silver: SLV ~23M shares/day (~21M oz) = >5B oz/yr on paper vs ~820M oz/yr mined.

For fast growers with little GAAP profit, compare price-to-sales not P/E: prefer the cheaper executing name (Palo Alto ~22x vs CrowdStrike 35x).

Dealer balance sheets tighten into period-ends on capital constraints, spiking equity-funding costs; the stress typically eases just after the turn — separating that seasonal fade from a structural leverage-demand-vs-paper-supply imbalance is the diagnostic.

Momentum oscillator measuring how fast / overbought-oversold a price is; the signal is divergence — a new price high at a lower RSI is a warning, a new low at a higher RSI flags a possible trend change.

Exchange-listed S&P 500 total-return future whose implied funding rate proxies the equity total-return-swap rate / dealer leverage cost; reading the active-vs-front-month roll spread exposes funding stress the headline ticker hides.

In a consolidating industry, size and scale are the thesis: sub-scale assets (e.g. a spun-off NBCUniversal) become 'takeover bait,' while scaled incumbents re-rate as rivals combine.

Risk-adjusted return (return per unit of volatility); 70% of thematic ETFs have a Sharpe below 1 — high vol, mediocre risk-adjusted returns.

Long-term, locked-in supply/purchase contracts replace volatile spot demand — contracted, durable demand is what re-rates a 'cyclical' (e.g. memory) into a secular grower and justifies an extended cycle.

If you're refusing to buy a name you like only because you want a small (e.g. 5%) dip first, you're implicitly admitting it's a buy at today's price — so buy it (Adam Parker's framing, via Lebenthal on Micron).

Read the direction of the 200-day moving average, not just price above/below it: rising = healthy trend, flattening = trend in question, rolling over = trend changing for the worse.

A low trailing P/E on a cyclical is computed on peak-cycle earnings (record output prices + margins), so it looks cheaper than the business is; cross-check with Price/Sales (less margin-distorted) and re-strike EV/EBITDA on mid-cycle, not the annualized peak quarter.

After a leader goes parabolic, the easy alpha is spent — shift to a supplier or smaller player levered to the same boom but less crowded (e.g. Onto Innovation off Micron's memory move) as a tactical alpha-capture trade.

Treat the momentum factor itself as the lead indicator for the bull trend — the first crack shows up there before the broad index; stay with momentum until it breaks. Note it can distribute into reasonable-valuation names (e.g. insurance), not just high-beta growth.

Narrow-theme funds (nuclear, cybersecurity, space) that blur sector lines and reflect performance-chasing; average ~32% three-year drawdown (cannabis ~-80%) with poor Sharpe ratios.

For a price-levered producer, build a fair-value ladder keyed to mid-cycle / current / bull / Street commodity prices; if fair value at current prices already equals the stock and the Street's higher target needs the commodity to rise, the 'upside' is pure leverage, not a margin of safety.

Grantham defines a bubble as a 2-standard-deviation (two-sigma) move off trend — a ~1-in-44-year event — and a 'super-bubble' as ~3 sigma, claiming 26 identified bubbles fully reverted to trend. Carlson's critique: basic statistics dressed up as predictive market timing.

Refusing the market's blended 5-7x multiple on a multi-business company and instead valuing each part on the cash it actually generates — Harrington's Comcast method.

Non-traded private-credit funds (Apollo Debt Solutions, Ares Strategic Income) cap quarterly investor redemptions (e.g. 5%) when requests spike (14-17%) — a liquidity-mismatch warning sign.

A parent whose stake in one listed subsidiary is worth more than the parent's whole enterprise value (Bit Digital's 70% of White Fiber) trades at a discount that closes as the sub delivers.

1934 Securities Exchange Act rule forcing a >10% holder/insider who buys and sells the same stock within six months to disgorge the profit back to the company (Avis clawed ~$650M from Pentwater).

Value a multi-segment business by valuing each segment on its own method and netting a holdco discount — used on Sea Limited (Garena + Shopee + SeaMoney) for ~105% upside.

Binding multi-year supply contracts where the customer must buy set volumes or pay a penalty — converts a cyclical commodity (Micron memory) into guaranteed, floor-priced revenue.

From the 2007 Berkshire letter: good businesses have a moat but must buy growth with capital (FlightSafety); great ones grow earnings without much new capital (See's Candies — $8m earning $5m in 1972, $40m earning $82m decades later); gruesome ones grow fast, eat capital and earn nothing (airlines). The question is what the owner must put in to get another dollar out.

BTC vs its 200-week MA used as a fast proxy for the global-liquidity tide (highs during easing, rollover at the tightening phase-shift).

New-contract bookings reveal a services company's future revenue before it's recognized; a first Y/Y bookings decline after years of growth is a regime-change signal even when reported revenue and EPS still beat.

Enterprise value divided by free cash flow — a valuation gauge that, paired with the dividend+buyback yield, frames whether a sold-off compounder (e.g. Accenture at ~6x, its lowest ever) has priced in the bad news.

Short-term, liquidity causes the initial move and sentiment amplifies it (the source of volatility); fundamentals dominate only long-term.

Revenue per available lower-berth day after certain costs — the core profitability gauge for cruise operators; Carnival's '12th straight record-net-yield quarter' framing and its full-year net-yield-growth guide drive the stock.

FedEx's program to consolidate overlapping Express and Ground pickup-and-delivery networks into one lower-cost network — an efficiency-over-expansion lever (45% of eligible volume, targeting 65% before peak).

Growth from locations open at least a year, stripping out new-unit expansion — the cleanest read on underlying restaurant health; decomposing the blended figure by brand (LongHorn +9.5% vs Olive Garden +2.4%) reveals who is actually carrying a multi-brand operator.

Eisman's filter: capital intensity is tolerable only with a durable moat (a protected franchise worth handing capital to). Capital intensity + no moats implies future price wars, low returns on capital, and 'is just plain scary' given the trillions spent. Moat test: do customers migrate between providers at will, and can key engineers leave with no non-competes?

A gain that is >90% price (same units, higher ASPs) is a cyclical pricing spike, not durable demand — the Micron bear tell.

Rusche's rule for buying quality on a one-off drawdown: separate a fixable/temporary cause (a single-mine seismic event, a guidance cut) from a structural decline before you sell — or buy.

Over a long horizon a stock's price tracks its earnings-per-share growth (Lynch: "100% correlation"); McDonald's 1985→2026 is the worked example. Short-term detachments (when investors price in a future EPS change) are where the alpha is.

When two unrelated quality names fall in lockstep (Netflix and Spotify charts near-identical), the cause is sector rotation/fund flows, not company fundamentals.

Back out the gold/copper price at which a miner's NAV equals its current share price (levered and unlevered) — that is the commodity price the equity actually discounts; compare it to spot and buy the widest gaps. Singh's miners comp sheet showed implied gold ~$3,200-3,400 vs $4,100+ spot.

How to read a P&C insurer: the 'float' is premium cash held between collection and claims payment, invested for income (its yield tracks the rate environment); the 'combined ratio' (losses + expenses / earned premium) shows underwriting profitability below 100%; 'rate adequacy' means premiums have been repriced to fully match current loss costs — the core of the Travelers thesis.

Aggregate every share you hold into one synthetic company weighted by position size, then sum the earnings your shares actually own (Carlson: ~$53k in 2026 → ~$61.6k in 2027) — a salary-like measure of a portfolio's earning power, separate from price.

Before following a respected investor into a name, learn their time horizon; only commit money you can hold that long (a 20-year thesis needs 20-year money — bailing after a year down 25% is the real mistake).

Borrowed from baseball (a ~.200 batting-average threshold for staying in the majors); used for a price level a stock keeps drifting back to and rarely breaks below — Simpson called ~$300 McDonald's 'Mendoza line.'

A stock/portfolio can fall even as earnings grow when investors assign a lower valuation multiple; Carlson maps his -8% YTD to a ~-20% multiple change on a flat ~16% growth engine — far less alarming than a deteriorating business.

If the multiple shrinks from 25x to 15x with earnings unchanged, the stock falls ~40%; severe quantitative tightening can compress multiples across the broad market.

A royalty/streaming company buys a slice of a mine's output at a reasonable return on the spot price, then benefits as operators find more ounces, expand the mill and extend mine lives — so the group tends to outperform the metal itself over time.

Price makes a lower or equal low while the RSI makes a higher low — a textbook sign of a forming bottom, stronger when it shows across a whole sector rather than one chart.

Instead of buying a theme's obvious stocks, identify who gets paid by the theme (the supplier = second derivative) and the supplier's suppliers / the back-office processes it transforms (third). Harrington's frame for playing AI via energy/power — e.g. Chevron powering Microsoft's data centers — after the 'obvious AI stocks' got crowded.

Multi-year contracts locking both volume AND price, backed by customer deposits, that set a margin floor — the test of whether a commodity-cycle business (e.g. memory) has structurally broken its boom-bust pattern. Verify with floor-case margin vs prior peak and customer concentration.

Feel the fear/frustration of a drawdown but pause and act on fundamentals, not price. Magellan investors lacked it — they sold Lynch's dips and earned far less than his fund.

Leveraged players arbitrage the tiny gap between cash Treasuries and Treasury futures. They have replaced price-insensitive central banks as the marginal buyer of US debt — so when volatility spikes they dump Treasuries to meet margin, pushing yields up at the worst possible moment and turning a calm bond market dysfunctional within days.

High-cost, low-quality producers carry the most operating leverage to a rising commodity, so they tend to outrun the quality names in the final, frothy stage of a bull market.

Sort every dollar by purpose: (1) savings/liquidity (gold + short-term cash), (2) core holdings held 'through hell or high water', (3) growth/investment on 5-6-yr plans, (4) speculation - split into active (capital still at risk) and passive (capital recouped). The bucket chosen up front dictates how you act when the price moves.

Terranova's lens: quant and momentum funds increasingly set short-term direction by building positioning where they can find alpha; front-run the flow into a newly-favored group (e.g. financials) rather than fighting the rotation with a pure fundamental counter — 'momentum is becoming more and more of a powerful force as market structure changes.'

Gains made on price rather than volume have a ceiling — at extreme prices customers find workarounds and use less of the input (the DeepSeek precedent in AI memory), which is eventually how a pricing-led cycle turns. Watch whether margin expansion comes from volume or from ASPs alone.

Write a templated memo per speculative position, revisited quarterly, scoring management against the actions that raise a project's value (certainty/size/grade) not its price. Ask management 'what is the most important unanswered question?' - a blank ('never thought of it like that') is a fail; discard the team.

Diagnostic for a commodity business earning outsized margins: is the pricing structural (contracted, no substitute, supply-locked) or a cyclical squeeze that new supply / demand-destruction reverts? Terranova's frame for memory — an 'irreplaceable resource' (AI accelerators freeze without it) with take-or-pay agreements locking price, vs Lebenthal's 'cyclical, just early in the cycle.'

Brown's screen — act only when sector tailwind, a chart breakout, and the fundamental story confirm simultaneously; any one alone is a trap, the edge is the confluence (his Citizens Financial long: regional-bank sector in favor + breakout + a wealth-management growth story).

After a speculation runs, sell enough stock to recoup original capital plus the capital-gains tax, then keep the remaining zero-cost shares - a paid-for 'library card' with no capital left to lose. Reclassifies an active speculation into a passive free-carry that rides the long-term thesis.

Switching costs (FICO scores embedded in bank loan systems), intangible assets (Hermes prestige), network effects (Visa's two-sided loop), cost advantages (Costco's bulk-buying flywheel) and efficient scale (Union Pacific - a market too small to be worth entering). Rated Wide (20+ years), Narrow (10-20) or None.

Convert a growth rate into absolute money: a $50m business needs $10m of new sales for 20%; Apple needs $80bn, more than iPads and Macs combined. Buffett, 1995: "in the early years we needed only good ideas, but now we need good big ideas."

Provisions letting a fund limit or halt investor withdrawals; rising use across private credit flags illiquidity risk shifting onto holders — 'a reallocation of wealth is underway.'

Isolate the most-concentrated index's market cap vs GDP (per David Stockman) to expose tech-led excess; today's reading sits well above the 2000 dot-com peak.

Loss-harvesting technique: the 30-day wash-sale rule applies to securities, not commodities, so physical metal (and crypto) can be sold at a loss and immediately rebought to book the loss and reset basis — confirm with an accountant.

Schools of cycle analysis that read markets as the cumulative, repeating emotions of the herd — cited to explain why counterintuitive price action 'plays the emotions.'

Invest in simple companies that appear dull, mundane, or out of favor and haven't caught the fancy of Wall Street — the rotation lens behind the Uber/DoorDash buys.

Judge an M&A deal by whether it buys a durable layer customers touch daily (workflow, interface, distribution, data loop) before market structure hardens — vs paying up for growth it couldn't build internally.

The S&P AIR TRF future approximates the cost of funding equity leverage (the total-return-swap rate). Tightness far out the curve, while swap spreads and repo stay calm, signals raw leverage DEMAND rather than a balance-sheet/plumbing break — a late-cycle warning that 'skyrocketing funding ends the party' even when liquidity looks fine.

Scale margin debt to nominal GDP (not raw dollars) to judge leverage; a fresh all-time high that is 'extremely extended' has historically tracked market tops.

As a scaled platform's revenue outgrows its fixed costs, it crosses a point where growth converts to real cash profit (Uber past it, DoorDash approaching).

Costco-style: most economics and loyalty come from a recurring membership (Uber One, DashPass), not the transaction take-rate.

Value a company's saleable or independent assets on their own against the whole enterprise value; if the cash those assets could realize approaches or exceeds the EV, you are buying the operating business (and any call options) for nothing — so you only need the 'free' business to clear a low bar.

Pair a slug of cheap, cash-generating compounders with a small bucket of high-variance bets treated like call options (warrants, LEAPs, pre-revenue). Manage risk through position sizing rather than stop-losses, and be ruthless about never adding to the option sleeve — the going-to-zero probability is real.

When a company must raise capital to fund its next step, a depressed share price forces a bigger, more dilutive raise — a self-reinforcing loop in which a single missed milestone compounds. Map every funding step and the 'timer' on it, or avoid the name entirely.

A repurchase executed with investment banks that retires a large block of stock at once (often debt-funded), with share delivery front-loaded and trued-up later — e.g. Salesforce's March $25B ASB that bought ~103M shares immediately under a $50B authorization.

When market leaders (e.g. Mag-7) underperform, classify it: bullish rotation = laggards (ex-Mag-7, small caps, equal-weight) make new highs as leaders pass the torch (rally broadens); bearish rotation = laggards don't pick up, so the heavyweight leaders drag the whole index lower. The ex-Mag-7 index and equal-weight-vs-cap-weight line are the deciding tells.

Normalize margin debt against the money supply (M2) so it's comparable across eras; read the rate of expansion as a euphoria/leverage gauge. Rapid acceleration ranking near historical extremes flags heightened downside risk, but with a multi-month lag before it resolves.

Capital with no maturity - insurance premiums or evergreen vehicles - that removes the obligation to invest and exit on a fund's calendar. KKR's $219bn via Global Atlantic; only genuinely cheap if the underwriting spread holds (stated at 1.8%).

High valuations can be justified by high return on equity (profitability), but ROE is cyclical, so the justification only holds while profitability holds. Monitor profitability directly (margins/ROE) for the first crack — an AI-capex bust or downturn that hits ROE removes the support and lets valuations mean-revert.

A bond covenant that lets holders sell the note back to the issuer at a premium to par if the company is taken over (e.g. an LBO). LBOs normally hurt bondholders via downgrades and price drops; a CoC "poison put" reverses that — bought below par, the put becomes a gain. Cited as a key safety feature of the Macy's 6.7% '34 note.

Offering price is not the opening price; the first-day pop goes to whoever got an allocation, not most public investors. Measure returns from the first-day close (Rivian +70% to peak then down 80%+).

BB-rated bonds sit at the intersection of investment grade and high yield — the highest tier of junk (~50-60% of the HY market) — yet their historic default rate is under 1% annually since the 1980s, so they pay a high-yield coupon for an asset that defaults far less than B/CCC debt; academic studies show BB delivers corporate credit's best risk-adjusted returns. Haymaker's "favorite slice" of the bond market.

Avoid the IPO hype; wait for the second earnings call; nibble in year one (starter small enough a 50% drawdown won't hurt); anchor to valuation; give it time — multi-baggers rarely require buying on day one.

The total annualized return on a bond held to maturity, a function of the coupon, the current market price, and the maturity date. When a bond trades below par its YTM exceeds the coupon — e.g. the Macy's 6.7% note at 95.474 yields ~7.45% to the 7/15/2034 maturity. Compared against a comparable Treasury, the excess is the credit spread.

In a capital-intensive industry with no pricing power (airlines), don't buy the operator — buy the supplier that sells into it and has real pricing power; compare 10-yr charts of American Airlines vs TransDigm. Eisman applies it to AI: hyperscalers may be becoming 'airlines' while their suppliers (power gen, semis, networking) become 'TransDigm.'

At his brokerage cage, any day buy tickets beat sell tickets 4:1 he'd force a sale next day, and vice-versa — a simple breadth/sentiment-extreme tool.

Bear Traps's score rating how washed-out a name/market is on a category-1-to-5 scale (price/sentiment collapse + trigger) to time deep-washout entries.

Per Peter Cundill / Buffett: there's always something to do within what you can actually price; refuse to value what you can't (he won't price Nvidia).

Hyperbolic up-moves (and down-moves) resolve unpleasantly for whichever side chased them; only act on the very obvious extreme.

New listings are sold at a time and price of the seller's choosing, with underwriters paid on the raise (over $500m of fees on the SpaceX IPO). The test to apply: how many years of generous growth does the price already assume?

Rule's bucket distinction: gold is savings (price-insensitive, never sell), silver is a speculation entered only when it's a hated asset class.

The poker-table framing: 'tourists' (fast/weak money) pile into a hot theme and get flushed by a shock, leaving real-money 'strong hands' — the best buying setups.

Investors crowd into recent winners (recency bias, repackaged as the 'momentum' factor); it works until it doesn't, then ends in sudden drawdowns — the case for buying the orphaned compounder, not the crowd.

Hold gold and Bitcoin as one hard-asset sleeve and rebalance between them on the ratio rather than on either price: McDonald sells gold and buys Bitcoin when the ratio falls into the low teens (it had come down from ~40), and does the reverse at the extremes.

Holding cash as a deliberate position (not a residual) to swing at the next washout — defensive 'medium-grip tires,' explicitly not a bear call.

The behavioral error of selling winners and holding losers — investors took profits in gold (the flower) as it rose, then sold faster once the high mark missed.

Heavy S&P call-option buying forces dealers to delta-hedge by buying the market, mechanically driving prices higher ('the market's running on gamma… when the Hulk runs out of gamma, he's just Bruce Banner').

Buying at any price on the assumption someone will pay more later, with the risk section of the annual report ignored - named as the mechanism behind speculative manias.

Heavy physical assets with low obsolescence, trading below replacement cost, and genuine scarcity — irreplaceable assets the market mis-prices on low cash-flow multiples (6-9x) despite being significant parts of world supply.

Debt held in SPVs, JVs and vendor-financing structures never appears on the borrower's own balance sheet — the Lehman-era blind spot McDonald applies to the ~$500-600B of hyperscaler data-center financing.

An allocator's sector-first discipline — '80% of a stock's move has everything to do with the sector,' so call the sector before selecting any single name; get smarter in the sector over time to find alpha.

A baby bond is a bond sold in small ($25-ish) exchange-listed pieces. To judge a BDC issuer's bond, screen first-lien %, average position size, non-accruals, leverage and coupon coverage, then do the LGD math: at a 50% recovery, defaults must run ~2× the loss you'd tolerate (~30% defaults ≈ 15% losses to impair). Buy the senior bond, not the headline-risk common.

A healthier rally where previously weak/overlooked parts (small caps, equal-weight) play catch-up and pull the index along, vs overheated leaders catching-down and breaking it; confirmed by equal-weight (RSP) and small-caps (IWM) new highs and small caps turning up vs Mag-7.

A sell-off-end signal: the index makes a lower low while a participation gauge (% of stocks above their 50-day average) makes a higher low — fewer stocks confirm the new weakness, so selling pressure is exhausting.

Assets (here listed space proxies) run up ahead of a marquee event (the SpaceX IPO) then unwind once it happens — compounded by rotation out of proxies into the newly listed real thing; can be a near-term headwind for proxies yet a long-term tailwind for the sector.

When a VaR/vol shock forces speculators to degross futures longs and substitute options, the dealers who sold those calls go short gamma; on an up-catalyst they must chase the market higher to re-hedge rising delta/gamma (and vanna) — in an illiquid market a self-reinforcing 'upside crash,' the inverse of a put-driven equity selloff.

A bond backed by private-EQUITY stakes (not interest-bearing loans), so it generates no organic coupon — interest is paid only by selling holdings (Ponzi-like). Levered ~60%, rated by conflicted small agencies, then stuffed into life insurers (often ceded offshore). $30B+ issued in 2026; Singh likens it to 2006–07 subprime structures.

When defensive sectors' (utilities, healthcare, staples) earnings/market-cap share probes the lows it signals late cycle (cf dot-com peak); it spikes in downturns (cf GFC) because defensive earnings 'just plod along.' Doubles as a concentration audit for cap-weight index investors.

A surge in trading of leveraged-long equity ETFs into a dip is read as aggressive 'smart money' dip-buying — supporting evidence a sell-off has run its course, though it can be early (fired late-2021 before the top).

70–80% of a tech stock's value is cash flows 5–10 years out, so a cheap trailing multiple is irrelevant if the future is impaired. If a free competitor will arrive in five years, intrinsic value falls TODAY though present profit is unchanged — and "the stock is down but fundamentals are improving" is the value-trap phrase (newspapers vs internet, department stores vs e-commerce, cable vs streaming).

Post-COVID, retail (now ~32% of equity volume) bids up-vol, so out-of-the-money CALL implied vol is elevated and downside put-skew premium is compressed — a "smirk," not the old left-tailed "smile." Gauged via IV/RV (1.4–1.5× vs a 1.2–1.3× norm) and CBOE SKEW (136 vs 118–125); the "VIX suppression" story is incomplete (0DTE gamma compresses short-end vol).

Identify the single biggest risk a business faces, then judge probabilistically — from its track record — whether it can solve it faster each time (Google vs the AI-kills-search fear).

Darnton's core framework: price tracks FCF/share with a lag, so the job is to anticipate the FCF/share inflection and buy quality names whose cash generation is rising while the price is depressed.

Give a basic tier away to build a huge audience, then convert a slice to paid; judge it on paid conversion, not free-MAU growth — Adobe applying it to Firefly/Express AI tools.

Share of gold miners in uptrends; a reading near zero / below ~20 marks extreme capitulation and historically the end of the decline (Russell / Moriarty).

Index and target-date funds must buy whatever enters the index, so insiders / early holders can offload newly-public stock onto retail 401ks regardless of valuation.

A surging order book (remaining performance obligations) is only value if the capex + financing to fulfill it pencils out — Oracle's $638B RPO with −$24B FCF made the stock 'a bet on execution.'

Get users onto the platform first and monetize them later; size upside from the paid-penetration gap vs a mature peer (Duolingo ~12% vs Spotify ~50%).

Trace a front-page event's non-obvious downstream effects two / three steps out (Hormuz -> sulfur -> sulfuric-acid byproduct windfall) to find the under-followed opportunity.

Deliberately running an 'investment year' that suppresses near-term profit/monetization to maximize long-run scale and dominance; bullish when the moat is intact.

A digital product's moat built from thousands-to-millions of tiny iterations that compound into a low-friction, delightful experience — can't be vibe-coded; how Spotify beat larger incumbents.

Per Bezos, growth driven by word of mouth on near-zero marketing is direct evidence of real product utility and stickiness; a bloated sales force is the inverse tell (Palantir vs Snowflake/Databricks).

Profitability engineered from compulsion - near-miss psychology (an 'almost win' drives more betting) or manufactured scarcity creating FOMO; a litigation / durability risk flag (Kalshi, Hasbro's Magic / D&D).

When a cash-rich 'asset-light' business suddenly issues equity, capex has outrun cash flow - the model has turned asset-intensive; rotate toward AI beneficiaries that fund from cash flow.

Pending index slots for giant private names force passive buys and pre-positioning by active managers 'regardless of merit,' breeding sameness that raises the risk everyone underperforms together.

Strip the one or two mega-growth sectors from headline index EPS growth to reveal the true breadth (Q2 +22.6% headline collapses to single digits ex-energy/tech).

Quantifies how few names own the index vs prior peaks — Nifty-50 (40% in 50 stocks then the 1974 bear), Japan 1989 (44% of the world index), dot-com (41% in tech/telecom) — against today's 'AI Big 10' at 41% of the S&P in just 10 names.

Differentiation measured by how fast users switch providers; frequent switching + price cuts = a commodity, so capex doesn't build a durable franchise.

Long-run valuation gauge (data since the 1880s); cited near all-time highs, matching the dot-com peak, as evidence equities are at a cycle top.

His unwind sequence: optoelectronics and memory crack first, then the GPU names, with power providers last — memory prices historically fall a year before demand does.

Vendors invest in their own customers who then buy the vendor's product (Nvidia→neoclouds→GPUs; OpenAI↔Oracle), inflating apparent demand across the AI ecosystem.

Reading sell-side consensus 3-5yr EPS-growth estimates (aggregated for the S&P 500) as 'Wall Street analyst sentiment' rather than a forecast - it overshoots even the best fundamentals in booms and undershoots even the worst downturns in busts; read contrarily at extremes.

From his convertbond.com days: 'the smartest sellers in the world are chief financial officers.' When converts (bonds with equity inside) surge as stocks rip — as in Q3-Q4 2021 and again now — CFOs are quietly selling equity at a fast rate of change; a 30-40% drawdown followed in 2022.

A dependable, historically validated reflation-vs-disinflation signal; in June 2026 it sits at pandemic-lockdown levels near its record low — no reflation.

Industrial-vs-safe-haven gauge: ~55-65 when global industry booms, 80-90 in secular funks (it visibly tracks China's GDP); Snider uses it as silver's fair-value anchor — ratio back to 80 implies silver ~$50.

'The best trades of our careers' — a sector where late-arriving retail 'tourists' (the Hawaiian-shirt guy buying with both hands) just got knocked out by a shock, leaving only strong hands. His capitulation model measures the flush; entries are scaled in thirds/quarters because 'only monkeys pick bottoms.'

Banks evaluate and structure, write the prospectus, pitch institutions on a road show, set the price off reverse inquiries the night before, then the shares debut — price comes from prior rounds and pitching, not cash flow.

When dealers/levered ETFs are short gamma, they must sell into declines and buy into rallies, amplifying every move (~-$20B per 1% S&P move vs -$10B at year end, per MS QDS). Paired with Paulo's rule that gross leverage = f(funding cost, volatility) — the first begets the second and the feedback loop starts.

When orders exceed shares offered, allocations are rationed pro-rata or by lottery; very hot books tempt issuers to raise the price or release reserve shares — both pressure the post-IPO print.

Paulo Macro's framing: the days ahead of a large options expiry give false directional signals (the market 'cha-cha-cha'); only after exiting the window does the tape pick a real direction that sets the tone into quarter end.

Insiders and pre-IPO investors are contractually barred from selling for ~180 days; euphoria fades into the unlock and hedging front-runs it — also the catalyst for unlock shorts (VCX).

Total equity market cap minus federal debt, divided by GDP; now the highest in 65 years, signaling extreme overvaluation.

Portfolio construction: ~50% core index ETFs plus satellite individual stock picks.

Constellation's P/E is distorted by non-cash goodwill charges; value it on free-cash-flow generation instead.

For a large IPO/secondary to be bought, capital must be sold elsewhere; passive indexes fund it by selling their largest constituents (the mega-caps), so new supply mechanically forces selling of the most-owned names — flow, not fundamentals.

Constellation buys businesses at ~4-5x cash flow and reinvests at a 20-25% rate on top of organic growth — the key to its quality.

Own small-caps that will either acquire or be acquired, since the merged larger entity attracts index inclusion and passive buying.

Sacerdote's thesis that AI workloads turned commodity data-center hardware back into high-IP, high-margin businesses.

In a commodity cycle the 'easy money' is the move from hated to unhated (uranium $20 to $85); the 'sure money' from structural deficit lies ahead.

Targeting markets with a discrete, identifiable supply-demand imbalance rather than headline CPI.

Whale Rock's chip-investing screen: percent of sales that are AI plus market share in that AI category.

A business model where costs don't rise with revenue, so nominal cash-flow growth accrues to shareholders.

Gathering ground-level intel from customers, competitors and employees to evaluate a company.

One of several frameworks (with hash rate, mining economics, the halving cycle) for valuing Bitcoin.

Paying miners a small amount per ounce as working-capital monetization rather than a straight royalty.

Whale Rock's core lens: tech adoption follows an S-curve; buy when barriers fall and growth goes vertical.

For autonomous vehicles, utilization is the entire economics; a demand-aggregator network (e.g. Uber's 202M MAUs / 40M daily trips) raises an AV's utilization ~30% vs a standalone app, making the demand layer the value capture.

A commodity product gains pricing power when packaged with distribution, integration and trust (AWS S3, Spotify, Netflix, Texas Roadhouse).

In a capex boom, capital rotates into whichever supply-chain constraint currently binds (memory → semicap → optical → power → CPUs) — the constraint owner gets the biggest earnings revisions, while the demand source (Nvidia) lags.

Carlson uses Texas Roadhouse's ~17-20% ROIC vs an 8-12% commodity baseline to argue branded service beats commodity economics.

When AVs replace human drivers, the ~70%-of-fare driver compensation can stay with the platform - lifting take rate (e.g. 28% to up to 80%) and adding large incremental revenue (~$20B at 20% AV penetration on $193B bookings).

Recapturing the 200-day moving average after a material (6-month) period beneath it is a lower-noise, later-triggering signal that historically flags new cyclical bull markets - 'nothing good happens below the 200dma.'

When orders booked far exceed what a company shipped that period (e.g. Broadcom's AI bookings ~3x shipments), it signals multi-year demand visibility - read it as a forward demand gauge, not near-term revenue.

A lower low in price against a higher low in the 14-day RSI - an early 'what-if a new bull is starting' prompt that precedes lower-noise confirmations.

Surging cash on both the FMS survey and retail sides signals a bearish crowd and stored buying power; spikes to multi-year/decade highs coincide with selling exhaustion near troughs.

Equity-capital-markets activity gauges the cycle: surging issuance floods stock supply near tops; a collapse toward zero chokes supply and historically signals a bottom is close.

In capital-intensive cyclicals you must buy hated commodities/regions; 'when the PEs are high because there's almost no E' is the time to buy.

Big increases in margin debt mark greed; steep drops mark panic as margin calls force selling - a selling climax that exhausts the seller base and clears the way for a bottom.

Crowded net-short positioning is a contrarian bottom tell - speculators are all-in short near lows - and is latent buying power, since a rebound forces panic short-covering.

A stock falling on great numbers is pricing expectations, not the past quarter: a key forward target reiterated (not raised) after a run-up acts as a disappointment (e.g. Broadcom -15% despite a +48% quarter when the $100B FY27 AI target was held flat).

When equity-capital-markets books dump their most recent issues to 'make room' for new supply into illiquid markets, the buckling of recent hot issues is a leading signal of a market rollover.

When a stock rallies 50-60% into earnings, a great quarter is already priced in, so even a clean beat-and-raise can sell off - a re-rating of expectations, not a business deterioration.

His framing of the current over-concentrated, mega-IPO-flooded US market as the fourth bubble in his career.

Banks holding an index on your behalf via a total-return swap charge a funding fee; a rising fee signals reaching for leverage OR tightening dealer balance sheets. The CME AIR Total Return future (Bloomberg AXW) is the listed proxy - a swap quoted as a rate over EFF/SOFR.

For a subscription business, the fresh annual recurring revenue added this quarter shows momentum better than ending ARR (a large, slow base) - and a record in a seasonally weak quarter is an even stronger signal.

To tell leverage-demand tightness from dealer-capacity tightness: if it were balance-sheet capacity, swap spreads would go much more negative and dealer gross shorts in equity futures would jump (a futures-basis trade). If neither moves, the funding spike is raw speculative leverage demand.

A rates trade betting the yield curve steepens — which Woo says makes sense again as the AI trade gets crowded.

~$4.6T of IPOs vs ~$1.5T of all IPOs since 1792; fast-tracking into the S&P/NASDAQ forces index operators to buy the new floats, pressuring incumbents - particularly the Mag 7.

McDonald's framing of AI capex overspending, compared to the 2010-14 shale boom and the 2000 telecom buildout.

Woo shows a 3-month bond sell-off was ~85% real yields (oil + AI), not inflation expectations.

Forced index inclusion of overpriced IPOs distorts the S&P/Nasdaq and hurts forward returns; 'passive as an aircraft carrier' that can't turn fast on regime change.

Enterprises burning through AI token budgets early (e.g. Uber in 4 months), inflating reported AI growth.

In resource bull markets, sector outperformance (beta) earns enough that stock-picking (alpha) is unnecessary for most investors.

US household equity allocation at a record coincides mechanically with record-high market valuations and signals a stretched regime with muted/fragile forward returns; read the extreme as a strategic risk-vs-return gauge.

A rising premium demanded on long-dated US debt because the market doubts the fiscal math is solvable.

Strategic M&A buys adjacent deposits to leverage existing mills; tactical M&A is growth-for-growth's-sake to win index inclusion.

'Bull markets are born on pessimism, grown on skepticism, mature on optimism, and die on euphoria' - used to flag the one-decision 'never sell' zeitgeist as a late-cycle tell (Nifty-Fifty parallel).

When degrossed traders flip from futures into options, market makers end up short calls; a rally forces them to buy futures into thin liquidity to hedge delta/vanna - a self-reinforcing squeeze (the semiconductor analog applied to a commodity).

When median (not just index-level) short interest hits decade+ highs, a broadening rally (e.g. an equal-weight S&P breakout) can force covering and assist prices higher - read crowded shorts as latent upside fuel.

Expensive valuations are a condition, not a trigger; bull markets typically end on a recession, shock/crisis, or prolonged monetary tightening - monitor those catalysts directly rather than selling on multiples alone.

A credible activist taking a large stake plus a board seat (e.g. Elliott/Jesse Cohn at Synopsys) becomes the forward catalyst - capital-allocation/governance change independent of the current print.

Discount a hyped 'AI-agent pipeline' until it shows in cRPO/ARR; size latent demand via the adoption-vs-governance gap (agents in production vs agents governed) without overpaying (Okta).

A stock falling on a quarter it beat - the forward number (next-Q guide / FY+1 growth / FCF-margin cut) re-rates a high-multiple grower regardless of the in-quarter beat (ZS, ESTC).

DRAM/NAND availability and cost, not demand, can bind AI-hardware revenue - a margin bite for assemblers (Dell/HP), read via 'customers signing multi-year supply deals' / 'no demand pull-forward'.

Why AI re-rates software: usage/consumption models earn more as agents work harder; seat-based (per-user) models may shrink as AI replaces human logins.

Anchor a commodity producer's 'cheapness' to what acquirers pay for whole companies in the sector - a stock below the M&A multiple has a built-in floor (EXE re-rate to ~5.5x still under sector deals).

A US tax-code provision that bars cannabis operators from deducting normal business expenses, strangling their margins. Relief — triggered by a federal rescheduling to Schedule III — is the real catalyst the operators are waiting for, distinct from a mere sentiment trade.

A compensation structure paying management a fixed slice of shares outstanding each year plus a share of any increase in market value. At Perimeter Solutions it cost $435m against $653m of revenue in a year the stock doubled - it rewards multiple expansion, pays most in your best years, and makes reported earnings meaningless until normalised.

Heavy-capex companies historically lag; the hyperscalers' ~$700B AI spend will spike depreciation (short ~4-5 yr chip life), opening a gap between reported earnings and free-cash-flow margins (Buffett's preferred lens).

Value a multi-segment company by pricing each business on its own peer multiple, then summing to a range — used to test the SpaceX IPO ($1.0–1.9T SOTP vs the $1.5–2T pitch).

Pre-revenue moonshots (Mars, orbital data centers, robots) get capitalized into a founder-run valuation today while the same bets barely move a skeptically-priced peer — "capitalized for one CEO and expensed for everyone else."

A pre-committed base rate of failure: one year in three the market falls at least 10%, one stock in three disappoints, and one year in three you underperform the index. Used so that a single bad outcome does not trigger a strategy change.

When a company models a restricted/zeroed market (NVIDIA's China) at zero in guidance while the base business grows without it, any restart is pure incremental upside — a free call option on a policy change.

Track gross-margin direction as the earliest tell of softening demand or competition — if either is biting, margin cracks before revenue or unit growth does (NVIDIA holding 75% through the Blackwell ramp).

For turnarounds/cyclicals, collapsed margins shrink earnings and mechanically inflate the P/E; use Price/Sales (less margin-distorted) to see the true depressed valuation and anchor the re-rating upside.

Forward-curve shapes: storable gold sits in contango, bulky oil/iron ore in backwardation.

Go long the turbines/gas/copper AI consumes; avoid the software AI can replace.

Per Horizon Kinetics/Ferg: nearly every exchange with a 20-yr public record beats its regional index; trading volume rises with nominal growth AND spikes in volatility (2008 trading +25%), so revenue grows in calm and chaos.

Michael Mauboussin's framing that bubbles form when everyone converges on the same belief about a new technology; the bubble then funds the build-out.

Baker's distinction for cyclical businesses like memory: a true capacity cycle (demand structurally outruns supply, e.g. mid-90s) has a far longer up-leg than a normal inventory swing — the one cycle you don't sell.

A top is a process, not an event: like water jets easing under a ping-pong ball, buying pressure relaxes gradually and the overpriced 'ball' descends slowly to its proper level.

Trader Ferg / Haymaker framework: favor businesses built on hard assets with low obsolescence (e.g. exchange operators) — network effects, high switching costs, durable cash flows.

The Lynch/Fidelity principle Baker kept: engage deeply with products as a consumer to find stocks.

Sell losers, ride winners — the Lynch axiom Baker says he's spent a career fighting because he's valuation-sensitive and contrarian.

Baker's framing of investor styles: pick whether you're a high-hit-rate (batting average) or high-payoff (slugging) player and stay consistent — it drives position sizing.

AI-complex value self-validates via marks the players set themselves (revenue circularity → 'Other Income' earnings circularity); a ponzi flywheel that reverses when an external 'one true mark' (a public listing) appears.

His term for assets like landfills: hard, irreplaceable, immune to obsolescence and AI disruption.

Find where supply chains are constrained and own the company that supplies it.

Roughly 140 Japanese firms older than 500 years, all sharing two traits: they hold a lot of cash and avoid debt. Survivability is the precondition for compounding - you cannot benefit from the good decades if you do not reach them.

An institutional index long the hottest momentum and short the weakest (currently long semis, short software); watched for two-sigma extremes.

For a contract/services business, forward bookings and a >1.0 book-to-bill reveal future revenue; record bookings refute an 'AI is automating this away' story (ACN: 1.2x, record $22.1B).

Compare free-cash-flow yield to the company's history and to risk-free rates; a 10%+ FCF yield on a high-ROE, dividend-growing franchise implies mispricing rather than impairment.

When the winner of a boom is unforecastable, buy the input every contestant must purchase rather than the contestants. From the 1848-55 gold rush, where 'most miners never found gold' and the reliable money was made selling picks, shovels, pans and boots. Applied to AI as cloud compute, power and grid capacity, cooling, fibre and storage.

A product whose demand rises as its price rises, because the price itself confers prestige — so normal price-elasticity reasoning is inverted. Slegers calls Hermes 'the ultimate Veblen Good'; sustained by deliberate under-supply, a waitlist, vertical integration and family control, and verifiable in the secondary market (a EUR 6,500 bag fetching EUR 35,000 at auction).

Valuing a self-liquidating cheap stock: when a multi-year capital build ends, declining CapEx drops dollar-for-dollar into free cash flow, which funds large buybacks on a small market cap (Eisman's Charter case — ~70% FCF yield in outer years, retiring up to 50% of shares).

Per 100 mortgage applications (~30% funded), FICO's Score 10 T collects 99c x100 + $65 x30 = $2,049 vs VantageScore's 99c x100 = $99 — a ~20x gap. Illustrates how a regulator-cleared 'good-enough' substitute can break a monopoly's pricing.

The tendency to quietly change your investment thesis as the stock moves — letting the tail wag the dog. Guard against it by re-underwriting against the original written pillars, not the new price.

Read where a turnaround is: Phase 1 fix solvency (raise cash), Phase 2 stabilize (stop comp declines, cut costs), Phase 3 the multiple re-rates. Enter in the '3rd/4th inning' after stabilization is confirmed, not at the bottom.

Two companies compounding earnings at the same average rate are not equally valuable: the one that does it smoothly is worth more than the one that swings +30%, -15%, +20%, because its future is more forecastable. Operationalised as three tests — consistent revenue growth rather than boom-and-bust, earnings that compound smoothly, and a resilient business model.

Buffett's ownership question, run as a written exercise on every position: would you still want to own this if the market closed for ten years and you could not sell? Slegers applies it to all 18 holdings and publishes the answers — 15 Yes and three 'Not sure' — which land on exactly the three names his separate conviction scale rates Medium.

At a new all-time high there are no trapped prior buyers waiting to sell, so 'there is no overhead resistance,' giving such breakouts a high propensity to keep running.

For a metals dealer/trader, an inverted (backwardated) curve drags earnings; a return to a normal curve is a mechanical, non-obvious earnings tailwind ahead of the print.

Every holding is measured against the best available alternative rather than against cash: 'if you are invested in a company where you believe the future expected return equals 8% per year, and you find another one with an expected return of 13% per year, you should consider making the switch.' The sell review is run before the buy list so the comparison has a named loser.

Restating a valuation multiple after deducting stock-based compensation, because share-based pay is a real cost that does not reduce reported earnings. Slegers' worked cases: FICO's 'cheapest in a decade' 23.8x becomes 29.6x once SBC equal to 25% of net income is deducted, and Fortinet's 29.2x becomes 33.6x at 15% — and the entry targets are then set on the adjusted figure.

Daily-reset leverage compounds choppy moves into losses and bleeds negative roll in backwardation, so a -2x oil ETF (SCO) can go nowhere despite a big drop in oil.

Ranking every holding into Very Strong / Strong+ / Strong / Medium ('should we consider selling these stocks?'), with one named cause per demotion — and explicitly not demoting on valuation (Medpace keeps top conviction at a 30.3x forward PE).

The valuation metric Slegers uses for serial software acquirers instead of earnings: the free cash actually left for owners after minority and debt claims, sidestepping the acquisition-amortisation charge that suppresses reported profit. CSU $1,683m (2025) -> a 4.8% forward yield; Topicus €218.7m -> 5.1%.

Profit with acquired-intangible amortisation added back, because 'amortization expenses are required by accounting rules, but they're not a real cash expense.' Used to build a multi-year multiple ladder off management's own targets (KPG: 20.8x forward -> 9.2x on 2029).

The cash a business genuinely generates for its owner rather than reported accounting profit; named in the post as the idea NPATA is 'very similar to.'

Funding long-duration assets (30-yr mortgages) with overnight deposits — blamed for the S&L crisis, SVB and First Republic.

Banks restructuring commercial-real-estate loans borrowers can't afford so the asset need not be written down.

Banks carry underwater long bonds at par because they intend to hold to maturity, hiding mark-to-market losses.

Treating the ~4% real yield given up by holding cash as an 'option premium' paid to preserve liquidity for a credit/liquidity squeeze.

His term for investors (and banks) chasing extra yield in junk-bond ETFs without understanding the credit/liquidity risk.

A listed holding or trust can be worth owning for two unrelated reasons, and each needs its own test. A discount case is tested by a sum-of-the-parts against the price (Brookfield, $42 vs $68). An access case is tested by whether the underlying private assets are worth their carried value and whether the wrapper's fee is a fair toll — Scottish Mortgage is included purely on this basis: "via $SMT, you get exposure to companies you could otherwise never own. Think about companies like SpaceX, Anthropic and ByteDance (TikTok)." Confusing the two lets an access premium be justified with discount language.

Holding companies own stakes in other businesses rather than operating anything, so valuing one means valuing several separate businesses and then netting the parent's share. Most investors will not do that work, which is precisely why the discount persists: "To understand Brookfield, you have to understand each one of these businesses. That takes a lot of time and effort. But it also creates opportunities. Investors often ignore these companies, causing them to trade at discounts." The corollary is that the value driver is the allocator, not the operations — "the most important thing for a holding company isn't running the business. It's making great capital allocation decisions." Worked example: Brookfield at $42 against a $68 intrinsic value, a 38% discount.

A player's optimal strategy doesn't change regardless of others' moves; used to argue Iran's Hormuz-constraint strategy is invariant to Trump, so the market's 'TACO' look-through is a misconception.

Read advertising and promotion spend as a percentage of revenue as the annual maintenance cost of a brand, then ask what happens to volumes if it were halved. Slegers raises it against the strongest brand case in consumer goods: "Year after year, L'Oreal spends roughly 32% of its revenue on advertising and promotion expenses. It makes you wonder what is doing the heavy lifting: the brand or the advertising?" The question is posed and left open — the moat is then defended on scale (barriers to reaching relevant size, unit-cost advantage) rather than on brand.

The bigger a business already is, the harder it becomes to grow it further — so size itself is a forecast constraint, not just a description. Used here as the first-listed risk on L'Oreal: "With around $51.7 billion in annual sales, L'Oreal's size makes rapid growth more challenging… the larger a business, the harder it becomes to grow." It is what turns a 7.8/10 quality verdict into a pass, since the reverse DCF then demands 13.0% annual FCF growth from a company forecast to grow revenue 5.0%.

Build a strategy — or an investment case — on the things that are stable rather than on the things that will change. Quoted in full in the L'Oreal write-up: "I very frequently get the question: 'What's going to change in the next 10 years?' And I almost never get the question: 'What's not going to change in the next 10 years?' I submit to you that that second question is actually the more important of the two, because you can build a business strategy around the things that are stable in time." Used to justify a low disruption risk in a mature end market — and, importantly, as a durability argument rather than a growth one.

Hand every efficiency gain from growing scale back to the customer as a lower price, so volume rises and the cost advantage widens — Costco, Amazon, and Action inside 3i Group.

A stock that beats every self-set guidance metric but falls short of an unofficial 'whisper number' can sell off hard — a sentiment dislocation, not a fundamental break. Check what was actually missed (own guide, published consensus, or whisper) and confirm the franchise metrics (renewals, backlog/RPO) are intact.

"Extreme performance… whether unusually good or bad… tends to drift back toward average over time." Offered as the operational meaning of "when in doubt, zoom out" — extend the chart until the current episode becomes small. The honest limit is documented two weeks earlier in the same archive: the claim "after every bear market comes a great bull market" sits alongside an eight-year crawl back from the oil crisis and three lost decades since 1900, so mean reversion says the level recovers eventually, not that you are still invested when it does.

Risk is packaged and sold along a securitization 'conveyor belt'; when it slows, banks retain risk and a credit crisis follows.

Credit-crisis truth emerges slowly until the narrative shifts fast — the tell that something is breaking.

When a quality franchise's margins compress, diagnose the cause before selling: deliberate management investment (price cuts to take share, capex) is very different from competitive deterioration. The acid test — is market share still rising while margins fall? The market often prices the two identically, which is the opportunity.

The archetype behind TransDigm's 100+ mini-monopolies, and explicitly the same shape as Constellation's vertical market software: a component that is mission-critical, faces almost no competition, and represents a rounding error in the customer's total cost. The canonical example is AmSafe's aircraft seatbelts — a 95%+ market share on roughly $20,000 of seatbelts in a $30-50m aircraft, "less than 0.1% of the aircraft's overall cost."

The FAA programme that lets a company other than the original maker produce an approved aircraft spare part, usually by reverse-engineering it. The single mechanism capable of putting a second supplier into an aftermarket designed to have only one — which is why HEICO is named as TransDigm's key competitive risk, given that most of TransDigm's profit comes from the aftermarket. Note the structural irony: the regulator that built the barrier also runs the route around it.

The original product is sold at very low margin, sometimes at a loss, while the replacements are sold at much higher margin. TransDigm's aftermarket (32% of revenue) is the worked example: aircraft fly 30-50 years, and in ageing, shrinking fleets "new competitors have little incentive to enter the market," so the original maker keeps the whole replacement stream. The right treatment in a valuation is to model the aftermarket as recurring revenue and the original sale as cyclical, not as one blended business.

An annual award from the Singleton Foundation — founded by Cary and Will Singleton, children of Teledyne's Henry Singleton — to "an exceptional CEO with a tremendous track record of creating per-share value." The jury is itself a map of the capital-allocation canon: Will Thorndike (The Outsiders), Todd Combs (ex-Berkshire) and Mark Leonard (Constellation Software). Nick Howley of TransDigm won it in 2022.

"We don't price products based on cost, but on the value we deliver to customers, which depends on the product itself and the switching costs." Executed as 5-6% annual price increases, every year without exception. Its precondition is the three-part pricing-power test — mission-critical, almost no competition, and a trivial share of the customer's total cost — evidenced by an Airbus habit of not bothering to claim refunds on parts under $800 against a TransDigm average price near $1,000.

The organising distinction of Compounding Quality's behavioural framework. Owners "don't care much about fluctuations in the stock price" and care about cash flow and growth prospects; Speculators are the exact inverse. Tested with the family-grocery thought experiment: if you never intend to sell, only this year's profit and the future growth can matter, and caring about today's quoted value is the Greater Fool Theory in disguise. "The Speculator sees the stock price drop and panics… The Owner checks the shelves. Customers are still buying milk."

Change in % = EPS growth + dividend yield. Compounding Quality's replacement for the share price on the monthly dashboard, on the premise that "in the long term, stock prices always follow the evolution of the Owner's Earnings." Disclosed for its own portfolio at 19.7% a year over ten years, and guided to +13% annually over the following three. Caveat worth keeping: buybacks flatter EPS growth, so the metric rewards share-count reduction identically to real growth.

A fixed cooling-off period applied to buys and sells alike: "Never buy or sell in a hurry. If you feel the urge to act, wait 48 hours before making a decision." Triggered by the feeling rather than the size of the trade, and paired with two other pre-commitments — writing the investment thesis before purchase, and agreeing the sell conditions before owning the position.

An owner sells only when the business fundamentals have permanently shifted, when the stock "has become so overvalued it defies all logic," or when a significantly better use of the capital appears. Decided at purchase, when you have no position and no emotional stake — "this keeps you from making up excuses in the heat of the moment." Note what is deliberately absent from the list: a falling price.

A low-cost commodity producer that deliberately leaves output unhedged turns every incremental dollar of price into pure free cash flow — making the equity behave like a leveraged call option on the commodity ahead of a supply shock (downside: full exposure if price reverses).

Maintenance capex replaces existing assets and is a genuine cost of staying in business; growth capex builds new capacity and may create value. The rule of thumb used: "the company's maintenance CAPEX is equal to the company's Depreciation & Amortization." Applied to Microsoft, whose raw CAPEX/Sales of 27.2% and CAPEX/Operating Cash Flow of 51.8% both fail the house test but become 2.8% and 5.3% on maintenance alone. The adjustment is only honest if the growth spending is genuinely optional.

Investing a fixed amount on a fixed schedule so the decision is removed from the moment. "You can't time the market. But you can tame it." Worked through the 2000-2012 lost decade: a $14,400 lump sum became $11,181 by 2012, while $100 a month became $16,351 — an outperformance of 35.9%. Note the mechanism's boundary, which the article does not show: DCA wins in flat-to-falling markets by construction and loses to a lump sum in rising ones.

Scores a market crash on two dimensions instead of one. Pain surface = drawdown (%) x recovery time (years), indexed to the 1929 crash (79% over 4.5 years = 355.5) as 100%. Worked: 2008 = 57 x 5 = 285, so 285/355.5 = 80.2% of 1929's pain; the COVID crash scores just 7.1% because the recovery took months. The finding that justifies it is that the ranking inverts intuition — the 1970s oil crisis (48% over eight years) tops the table despite being far shallower than 1929, the dot-com bust or 2008. "It's not the crash that hurts, it's the crawl back."

Ten years or more in which the market rises but not enough to outpace inflation — a real-return drought rather than necessarily a falling index. Three since 1900: 1914-1945 (two world wars plus the 1929 crash), 1973-1985 (inflation and weak markets) and 2000-2012 (dot-com plus the financial crisis). A meaningful share of any investing lifetime, and the reason nominal charts can hide the problem.

How much cash a business generates each year for every dollar of new capital it puts to work — the metric that matters most for a reinvestment compounder, because reinvestment is the entire engine. Constellation Software's figure over three years is put at 29.1%: "for every $100 you invest, CSU makes $29.1 per year for you." A decaying CROI is the first evidence the acquisition runway is closing, well before growth slows.

Kelly Partners Group's preferred profit measure, and a serial acquirer's version of Buffett's owner earnings: amortisation of acquired intangibles is "required by accounting rules, but they're not a real cash expense," so it is added back. Used to value the company at 21.8x expected 2026 NPATA, falling to 15.0x / 12.0x / 9.6x on management's own 2027-29 targets. The add-back is defensible for acquisition intangibles and not for capitalised development spend that must keep recurring.

One year out of three the stock market falls at least 10%; one stock out of three that you buy will disappoint; one year out of three you will underperform the index. Used twice in March 2026 as the base rate that converts a drawdown from evidence of failure into an expected cost — matched to Buffett's own record of underperforming in 20 of 64 years.

Crashes come from unstable positioning, not euphoria: initial correction + relief bounce → retest with a minor breach/intraday tail → a 1-2 day 'all clear' final bounce → collapse (1987, 1929, 2021 bitcoin, 1997 Hang Seng).

A 20-year, non-cyclical super-cycle: as boomers age out and Gen X/Y/Z come of age, more auto insurance is bought online via price-comparison marketplaces, never through a live agent — a durable tailwind beneath the cyclical underwriting market that powers lead-gen businesses.

In cyclical industries earnings swing too much to trust the PE, so anchor on the steadier price-to-sales; a low price-to-sales with a high PE is the classic buy signal, and the rarer low-price-to-sales-AND-low-PE "double win" usually means the market has abandoned the name.

His proposed allocation (30% stocks, 30% bonds, 40% commodities/commodity equities) replacing the old 60/40.

Selling upside and buying downside protection on a concentrated position (Mark Cuban's Yahoo collar is the classic example).

A grading scale rather than a yes/no test. Level 1 convenience moat: customers stay because it is easy. Level 2 economic moat: customers stay because switching costs money. Level 3 structural moat: customers stay because there is no real alternative. The instruction is a filter — "You should focus on Level 2 and Level 3 Moats" — and it is the sharpest tool the March 2026 run produces for the AI-disruption question, since a convenience moat is exactly what a better interface removes.

McDonald's thesis that capital is rotating from financial assets (paper) into hard assets over many years.

In a bull market a move down lures bears to short, then a sharp reversal traps them — the namesake of McDonald's report.

Munger's term for assets marked at fake levels (vs mark-to-market and mark-to-model) — the core of the private-credit mismarking problem.

Mohnish Pabrai's framing — "I'm a shameless copycat. Everything in my life is cloned … I have no original ideas" — applied to quarterly 13F filings. The method's two corrections, both from the same issue: correct every aggregate "most bought" list for manager size (a manager of billions cannot buy small caps, and must own the mega-caps that dominate the benchmark), and weight a purchase most when the buyer steps outside their own style, as a deep-value investor buying a compounder is making a claim about price. The counterweight arrives three weeks later: "you can copy someone's stock idea, but you can never copy someone's conviction."

A business paid a small, near-mandatory fee on someone else's transaction without carrying its risk or capital — FICO on every loan application, S&P Global on every bond issued and every index fund, MSCI on index-linked assets, Cintas on the physical workplace.

Judge a utility deal by purchase price / the target's regulated rate base (the asset that earns the allowed return); ~1.5-1.7x on a constructive regulator implies accretion. Used on NFG's CenterPoint Ohio deal.

Buy a basket of beaten-down names into the year-end forced-selling crescendo for short-horizon mean reversion, then harvest the winners on the January pop.

The house framework: fifteen pre-published pass/fail tests (business model, management, moat, end market, risks, balance sheet, capital intensity, capital allocation, profitability, SBC, past growth, future growth, valuation, Owner's Earnings, shareholder value) aggregated into a single score out of 10. Tesla scores 6.8/10 against 8.3 (Computer Modelling Group), 8.2 (Eli Lilly), 8.0 (LeMaitre) and 7.8 (HEICO) elsewhere in the archive.

Treating SBC as a real cost by subtracting it from the free-cash-flow base before valuing a company, and measuring it as a percentage of net income (Tesla: 74.5% latest, 67.7% five-year average, against a 10% bar). The reverse-DCF base is built in the open: $4,996m FCF less $2,826m SBC plus $2,379m growth capex = $4,549m.

Splitting reported capital expenditure into what keeps the existing business running and what builds a new one, using depreciation and amortisation as the maintenance proxy, then re-running capital-intensity ratios on the maintenance figure alone. Applied to Tesla, the ratios improve from 9.0%/57.8% to 6.5%/41.7% of sales and operating cash flow — and still fail.

The structural argument against market timing: an exit and a re-entry must both be correct, and "the best days on the stock market usually take place just after the worst ones," so missing the ten best trading days over 27 years leaves a fraction of the return. The rule that replaces the judgement is a fixed monthly contribution.

Quoting free cash flow divided by market value instead of a multiple, so the valuation is directly comparable to a bond or a hurdle rate — and comparing it against the company's own history. Constellation Software at "a FCF Yield of 6.8% (the highest it has ever been)," rising to an expected 9.7% on 2027 numbers.

Smallcap/value/non-US catching up is a very late-stage bull narrative that has historically preceded one last large-cap jam higher (1999-2000; the early-70s Nifty-Fifty echo into 1973-74).

Sell a fixed fraction of a position's value at each run-up rather than all-or-nothing; captures most of buy-and-hold's dollars at far lower round-trip risk — Baruch's 'leave the middle 80% for someone else.' Worked through on a TSM four-bagger.

A company with surplus cash can repay debt, reinvest, acquire, pay dividends or buy back shares. The house preference is reinvestment at high returns on capital — "you want companies to reinvest heavily in their own future growth" — which sits in tension with the same archive counting dividends and buybacks as shareholder return elsewhere.

"Income is an opinion. Cash flow is a fact." The two-number definition used across the archive — operating cash flow less capital expenditure — offered as the reason to screen small caps on profitability before size.

The refinement of the small-cap trade: buy small companies, but only those with positive free cash flow and high returns on capital, because a broad small-cap index "includes a lot of unprofitable companies." Implemented in the Pacer SCOW index as profitable-seven-years-running, then ranked on cash conversion and return on capital, keeping the best 80 of 600.

The house screen stated as a checklist rather than a metaphor: a business outperforms if it is (1) essential, (2) has no replacement, and (3) generates stable and predictable cash flow. The plain-language form of the toll-bridge test.

Greenblatt's mechanical screen ranking companies on earnings yield and return on capital, cited as delivering 33% a year over 1985-2005. Used in the archive as the example of a simple, holdable strategy rather than an optimal one.

Hunt for quality businesses sold off under a popular fear ('AI will disrupt this') whose fundamentals are actually improving while the stock falls — buy the gap between the tape and the financials.

Combine the share-price change and the EPS change over the same window into one statement about the multiple: Kinsale down 15% while EPS grew 15% means the stock became roughly 30% cheaper. A fact about the price paid, independent of any forecast.

The ratio Lynch popularised: a stock's price-to-earnings multiple divided by its expected earnings growth rate, to judge whether it is over- or undervalued relative to its growth. Lets a 30x business growing 30% be compared with a 12x business growing 8%.

Lynch's label — "he never followed just one strategy. Instead, he matched different strategies to different markets" — classifying candidates as fast growers, turnarounds, cyclicals or stalwarts, each with a different question and a different exit.

Software built for one specific industry's workflow (dental offices, marinas, public transit) rather than broad horizontal tools like Excel or Slack — producing low churn, high pricing power and deep customer relationships.

"In the short run the stock market is a voting machine, but in the long run it is a weighing machine" — used to frame buying quality software during the AI narrative rather than after it resolves, since "stock prices eventually follow earnings".

When a sector's index weight has fallen very low, even marginal rotation flows into it produce outsized absolute and relative gains (2000-02 utilities/healthcare/energy/staples).

Bear markets start with a key bubble sector failing while the index flounders and unrelated position unwinds cascade under the covers (the 2018 template).

Using the monthly count of watchlist names undervalued on all three methods at once as a market-breadth signal for how much capital to deploy. February 2026 set the series record at 51: "This number (51) has never been higher."

Wall Street saying: hedge when you can, not when you have to — buy protection while everyone is euphoric and insurance is cheap.

Kovner's setup: when everyone believes one thing but price action won't confirm it, a lot of people are about to be wrong.

The firm's qualitative + return-on-invested-capital checklist (strong balance sheet, insider ownership with recent purchases, would-you-own-the-whole-business test) applied US and internationally.

Track the change in what a business is worth and the change in its share price as two separate annual series; the gap is the improvement in expected return. Worked example: Brown & Brown's intrinsic value +13.6% against a -21.9% price in 2025 made the stock "35% (!) cheaper."

Pairing stocks with a negatively-correlated, positively-carrying hedge asset (classically bonds); Muir argues gold has taken over the bonds role since Liberation Day.

Markets cycle through ~one-year manias (EVs, Bitcoin, gold miners…) that roll from theme to theme — and each bubble is getting more violent.

The Buy-Hold-Sell rating engine: every watchlist name scored on its forward PE against its own five-year average, an earnings-growth model (EPS growth + dividend yield ± multiple change to a fair exit PE), and a reverse DCF solving for the growth the price implies. The rating is the composite, so a Buy often fails one method — and the disagreements are the research prompt.

Technical levels no longer hold but become stop-run exits or size-reload spots in the direction of the primary trend; an overnight 'Holy Grail' flush is the hallmark of forced liquidation and a reentry signal.

A cleaned, standardised number is a claim until traced to its primary filing; provenance, not accuracy claims, is what makes data usable when you must defend a decision. 'Trust is built over me showing you instead of me telling you.'

The metrics that actually decide a business (trips, transaction volume, cards in force, gross bookings) are disclosed but sit outside the three statements - revenue and margins can miss the turn entirely.

Chart one commonly disclosed operating metric across every peer - asking which grows fastest, off which base, at what take rate - to test an industry narrative against the companies' own numbers.

Use a screener to filter out what is not a good use of your attention - whole sectors whose behaviour mismatches your style - before using it to find candidates. The scarce resource is attention, not ideas.

Volume and take rate normally trade off; both rising materially at once is the rare confirmation of pricing power and demand together, and usually precedes margin expansion.

When a company stops disclosing the metric your thesis rests on, treat the non-disclosure itself as the sell trigger - it removes your ability to monitor, and is rarely dropped when it flatters management.

Graham's testimony that price eventually resolves to value through recognition; the host's framing that better tools in more hands speed that recognition and make markets work better.

Compounding Quality's taxonomy for matching a portfolio to a goal: Getting Rich = small quality companies growing very quickly; Staying Rich = established quality stocks still growing attractively; Living Rich = quality companies paying an attractive dividend. The third is defined in cash terms — "build a portfolio that generates enough dividend income to cover your expenses. When you can do that, you've achieved financial freedom." Each maps to a different product (Tiny Titans, the main Portfolio, Compounding Dividends).

A moat created by legislation rather than by economics: US state franchise laws restrict who may open a car dealership in a given territory, so "dealerships are protected by state laws that limit competition, creating a regional monopoly." The earnings quality sits in the protected service and parts department rather than in cyclical vehicle sales. Worth pairing with the standing caveat that a legal moat can be removed by the same legislature that granted it.

A lease under which the tenant, not the landlord, pays the property taxes, insurance and maintenance — so the rent received is close to net cash. Cited as the reason VICI Properties' tenants "paid 100% of their rent throughout the COVID-19 lockdowns," and usually paired with annual rent escalators tied to inflation. Durability comes from the asset being embedded in the tenant's business: "you can't move a casino."

Buying single-stock volatility and selling index volatility; it suppresses index vol by relying on abnormally low S&P constituent correlation, and unwinds violently if correlation rises.

Forcing institutions to value assets at distressed market prices in a panic can push even solvent banks into perceived insolvency; its 2009 suspension (with TARP) helped end the GFC.

Shortselling framework: 'tops are a process, bottoms are an event' — hunt downside first in laggard names sitting on / testing a flattening 200-day MA from below ('shoot them when they're running away').

Companies that spend most of their cash retiring their own shares — for which a falling share price is genuinely good news, since the same buyback budget buys back more (AutoZone, Adobe at ~9%/yr, MSCI).

The sentiment gauge Slegers logs in every monthly Best Buys issue alongside the S&P's move, tracking the arc from Neutral (January 2026) to extremely fearful (April 2026) as a read on the quality of available entry prices.

Sales growth measured across the same set of stores over time, excluding new openings — the retail metric that separates genuine demand from expansion.

Compounding Quality's label for a small, obscure, high-quality compounder with multi-bagger potential — the category behind its closed, waiting-list Tiny Titans service. LeMaitre Vascular is the published worked example: a $1.9bn vascular-device roll-up that turned $10,000 into $161,460 since 2006 while remaining almost uncovered. Useful as a reminder that a Tiny Titans write-up is not a candidate for the main portfolio, even when it scores 8.0/10.

A re-rating thesis: a sector the market prices as low-quality/cyclical is structurally becoming durable infrastructure (e.g. mobility as a capital-light 'urban operating system'); buy the leader before the relabeling is consensus.

The idea that the average of many independent guesses beats almost any individual one — Slegers' own illustration is guessing a cow's weight at a fair: "some guess too high, some guess too low, but when you take the average of all guesses, it is almost exactly correct." It is the stated basis for the annual Partners' stock-pick survey. The unstated limit is the word independent: a readership all reading the same newsletter is correlated, and six of the ten 2026 favourites were the author's own holdings.

Depreciation + depletion & amortization per barrel; want it low. Facilities running above design capacity and waterflood reserves booked 'for free' mean economic earnings exceed reported earnings.

Why disciplined engineering & construction firms avoid fixed-price/turnkey work (which forces the contractor to eat cost overruns) in favor of cost-reimbursable contracts where the client pays actual costs plus a set fee.

GaveKal's term for the deepest, most vicious bear markets — the kind that accompanied past recessions (2008, 1981, 1974, 1970).

Government real-time rails (India UPI, Brazil Pix, US FedNow) route around card networks; assess them by fraud protection + cross-border reach, not just speed/cost.

Finality/efficiency (crypto/stablecoin) removes the float, chargebacks and dispute recourse that actually benefit the consumer; a disruptor optimizing efficiency can be an inferior product.

Unlike a simple two-sided network, each additional cardholder is exponentially more valuable to each merchant and vice-versa, making the network un-replicable from zero.

A network bribes high-end consumers with rewards while charging merchants the interchange fee for access to them.

Define the worst realistic outcome (here a slow drift to a regulated utility, not a blow-up) to judge downside vs upside before sizing a position.

Every transaction carries a lack of trust between the two parties; a payment network's core product is bridging that deficit — Carlson's central Mastercard framing.

When scarce HY cash bonds trade tighter than easily-levered CDS, selling HY CDS (synthetic long) earns an excess return — a levered trade that a 200bps spread widening can wipe out.

In a high-inflation/high-rate regime, industrials, materials and energy dominate the S&P (~50%) over tech.

High-capitulation-volume selling, low relative strength and seller exhaustion signal a bottom.

Debt attached to one specific project or asset, so a failure there risks only that project's equity and not the parent — 94% of Brookfield's debt, against 47% group debt-to-capitalization but only 21% at the parent.

Invest, improve, sell mature assets, reinvest the proceeds — so reported revenue and cash earnings move in opposite directions at different points of the loop and must be read through the cycle rather than by the year.

A non-GAAP cash measure used by asset managers when reported revenue is meaningless — Brookfield consolidates 100% of the revenue of assets it owns 30-60% of while its fee businesses barely register; DE before realizations strips out one-off gains to show the recurring core.

Premiums collected today against claims paid years later form a pool the insurer invests in the meantime — effectively free long-term capital when underwriting is disciplined (Berkshire, Markel, Fairfax, Brookfield Wealth Solutions).

Buy-and-forget investing (stash certificates in a can for decades). Polomny invokes it as the WRONG model for cyclical resources, which must be bought when hated and sold into the spike, not held forever.

Commodity-cycle adage: low prices kill marginal supply and stoke demand, which eventually forces prices back up — the basis for buying a hated resource and patiently waiting for the cycle to turn.

The value of standing, uncut timber: the trees keep growing ~6-8% larger each year, so a timberland asset compounds in volume even when lumber prices are low — a built-in margin of safety while you wait for the cycle.

A pass-through partnership structure common among pipeline/midstream operators: units trade like stock and pay large cash distributions, but because they grew distributions too fast ahead of the 2015 commodity-price plunge, many (even Kinder Morgan) were forced to cut. The risk metric to watch is the cash-flow cushion (distribution coverage), which the article says is larger now than in 2015.

A junior lets a major fund its de-risking studies (drilling/MRE/PEA/PFS) via milestone payments in exchange for a staged project interest — de-risking capital when junior equity markets are shut, but handing the funding partner an M&A ceiling over the asset. The partner eats its sunk cost if it walks.

Private-credit contagion framework: when a private fund is forced into its near-identical listed vehicle at a NAV discount, the public price becomes the private loss — forced price discovery.

A real commodity breakout is confirmed by the whole futures curve — moving to backwardation AND the long-dated price inflating (the 2004 oil template applied to copper).

Once a rare big winner is identified after owning it, defend it from being sold — not selling the ~12 winners is what built Berkshire; mistakes of omission (selling Ferrari early) cost more than the zeros.

Copy a proven business rather than invent — Gates and Walton were 'me too' cloners; a great cloner is 90-95% ahead of the field.

Structure any bet so the downside is near-nonexistent while the upside stays open; the core Dhandho principle.

Spot things that should exist somewhere but don't (a new town with no barber) and fill them cheaply and part-time, at a premium, with minimal risk.

Buffett's rule: imagine only 20 lifetime stock purchases — the scarcity forces thoughtfulness; ~4% of stocks make ~90% of returns.

72 ÷ annual return ≈ years for money to double; shows the runway matters more than the rate ($23 at 7% → ~$23T over 400 years).

Robotti's term for the coming comeback of active management and neglected value stocks: capital has fled to passive/private equity for 15+ years, mis-pricing the abandoned small-caps that are the next decade's opportunity — a rhyme of the 1973-74 Nifty-Fifty unwind.

Wayne Gretzky's maxim, applied to investing: position capital where opportunity is heading (neglected value/active management) rather than where it has already crowded in (passive/private equity).

A business whose assets are worth far more than its stock price, but which the market can't value because the assets aren't yet generating cash; the value is 'chained' until the earnings stream manifests, then the stock re-rates dramatically.

A fixed 15-question worksheet with published numeric thresholds (gross margin >40%, ROIC >15%, ROE >20%, interest coverage >15x, goodwill/assets <20%, CAPEX/sales <5%, net margin >10%, FCF/net income >80%, SBC <10% of net income) rolled into one score out of 10, with valuation answered second-to-last.

Every consolidating oil major has a controlling capital allocator (Waterous, Li Ka-shing, Murray Edwards, Exxon); without a strong one you are a takeout target. A screen for who actually drives strategy.

NOPAT ÷ invested capital; Slegers' capital-allocation gauge, wants >15%. A 20% ROIC means every $100 invested throws off $20 of value — but only compounds if the firm can reinvest at that rate.

Instead of forecasting cash flows, solve for the free-cash-flow growth rate the current stock price already implies, then judge whether that's beatable vs history/guidance. 'Invert, always invert' (Munger).

A Harvard Business Review study found founder-run businesses outperform by ~3.9%/yr on average; Slegers requires a founder still operating with the majority of his wealth in the stock. 'Show me the incentive and I'll show you the outcome.'

Informational and analytical edges are largely gone for retail (online data; quants/PhDs/AI); the one durable edge individuals keep is behavioural — no clients to answer to, so you can hold cash and wait.

A large regular dividend re-creates a guaranteed liability just after deleveraging removed one; a US holder loses ~23.6% of a Canadian dividend to withholding + tax. Prefer buybacks when the stock is cheap — the government is in the dividend business, not the buyback business.

Inflation-adjusted earnings smoothed over a trailing 10 years to correct for booms/busts; used to rank whole country markets by cheapness (US highest; Brazil <8 = among the world's cheapest, ~10 = deeply undervalued).

A shrinking count of publicly listed companies (buybacks, take-privates, delistings); when supply is shrinking in a hated market, returning inflows meet little stock and prices can move violently higher (Brazil lost 13% of listings since 2021).

A decisive break above a horizontal ceiling a market has failed to clear for 3+ years is a leading bullish signal — markets often discount a reform/election outcome before it happens (Argentina's ETF broke out in early 2023, pre-Milei).

A preferred (bond-like) share whose dividend starts fixed (e.g. ~7%) and, on a set reset date, switches to a floating rate of SOFR plus a spread (e.g. +4–5%). In a 'higher for longer' regime the payout resets up toward ~9% instead of leaving you stuck in a low fixed coupon — removing much of the duration risk of a fixed perpetual. ~40% of the SSR preferred book is fixed-to-float.

If every Canadian junior mining stock were merged into one company it would post multi-billion-dollar losses every year (~$9B in 2018, ~$6B in 2019) — so as a group juniors destroy capital; the sector only rewards genuine stock-picking.

Cited by Robotti — having the numbers isn't analysis; you must reason about the underlying economic regime, not extrapolate the recent past.

The biggest companies by market cap rotate every decade (1970s energy, 1980s Japan, 2000s China, 2010s low-rate winners) — ask which companies benefit from THIS economic environment, not the last one.

In science/engineering knowledge compounds; in finance it runs in cycles, so what worked last decade isn't what works next — the antidote to recency bias.

Not the 'revenge' but the metamorphosis — capital-deprived, consolidated, restructured old-economy industries whose macro backdrop flipped from disadvantaged to advantaged; 'a butterfly today, not a caterpillar,' yet still priced as the caterpillar.

No matter what the world does, valuation on entry is the only fully controllable variable — the core discipline behind buying at a fraction of replacement/intrinsic value.

Buy great companies with no intention of selling for decades — named after the can a grandmother used as a piggy bank; the selection test is 'why will this still be relevant in 50 years?' rather than valuation.

Grow by repeatedly buying small businesses in a fragmented industry and leaving them to run independently, recycling their cash into the next deal — Constellation Software, Topicus, Watsco, Asseco, Computer Modelling Group.

Supply forecasts that assume every mine runs at full rated capacity overstate output; mines typically produce no more than ~85% of nameplate, and applying that haircut pulls the Red Book's uranium shortfall from 2032 forward to 2030 even on low-case demand.

Moving a uranium deposit from discovery to production takes 15-20 years, so the incentive price signal must arrive well before a forecast shortfall - the binding constraint is time, not geology.

Enterprise value over EBITDA before exploration expense, the standard E&P multiple; Singh's EXE target uses about 5.5x on conservative oil and Henry Hub decks.

Screen producers whose capex is dropping quarter over quarter while mines are already producing: higher metal prices then flow to FCF at the lowest cost.

Restate a past fuel-price shock in today's fleet fuel economy and dollars. The June-2008 peak equals ~$9.50/gal today, so consumers only feel real pain around $6-7.

Drill a regular grid (holes every 50 m across strike, fences every 100-200 m) over half a big anomaly to map its extent, then overlay IP, soil geochem and ground magnetics on the drilled block to build a signature and rank targets on the undrilled ground.

The US-funded rail line from Zambia/DRC west to Angola's Atlantic port of Lobito, and China's refurbished line east from the same Ndola junction to the Indian Ocean - infrastructure rivalry that pulls both Western and Chinese buyers into the Zambian copper belt.

As a futures contract nears expiry its price must meet the physical/prompt price; a big physical premium means either physical falls or futures grind up (usually some of both), cushioning futures.

If activity rises (road freight +3.5%) while apparent demand for the fuel it uses collapses (diesel −20%), the gap is an unseen stock draw — finite, and it predicts renewed buying

Stripping overburden and saw-cutting samples perpendicular to exposed high-grade veins fixes the shoot orientation before drilling, and the samples can be added to a resource estimate like horizontal drill holes.

Dated Brent is the physical cargo benchmark; trading far above prompt futures ($131.77 vs $107.50) shows how tight the physical market is

Measure a bypass by the change in net exports at its terminal, not its nameplate — East-West's 7 mb/d was only a ~3.5 mb/d swing because it already carried oil

China's fear that in a Taiwan conflict the US could blockade the Strait of Malacca and cut its oil imports — the reason for a decade of Chinese energy-security stockpiling and fuel-switching capacity

Recalculate a hole's nickel to 100% sulfide; if a low-grade disseminated step-out shows the same tenor as the known deposit (e.g. 2–3% Ni), it is the same fertile system and could build a comparable deposit with more drilling.

A borehole electromagnetic survey run down every hole can show strong conductors beyond the end of holes stopped once they left mineralization; massive sulfide is highly conductive, so those become the next targets.

Bruce Flatt's framing of the Nvidia-led $500B plan: GPUs, today treated as consumables, become collateral that private credit can lend against, like real estate, power plants and infrastructure.

Commercials (mostly producers hedging output) have been net short since 2009; a very high net short marks tops, and a low one at high prices means producers expect higher prices.

OPEC's habit of cutting quotas whenever oil fell to a certain level, which put a floor under price for years; now reversed, with targets raised symbolically

The Strategic Petroleum Reserve level (~300M barrels) below which oil can't be pumped out quickly, so usable cushion = current level minus ~300M

Rule of thumb: when metal stocks cover less than six months of annual demand the market is very constrained, whatever the current-year balance says.

Subtract H1 actual from the full-year forecast; a headline surplus built on past investment outflows can hide a forward deficit.

Running legacy high-assay depleted uranium tails back through an enrichment plant to recover fresh feed — at Paducah, 200,000+ t of DOE tails are expected to yield ~70,000 t of fresh feed, supply that competes with mined uranium.

Elevated metal lease rates plus a London OTC curve in backwardation signal scarce spot metal; flat-to-contango signals ample supply.

Uranium enriched to between 5% and 10% U-235 — a tier between standard LEU (under 5%) and HALEU (up to 20%), targeted for higher-burnup and advanced reactor fuels.

A 100% offtake that pays the project the buyer's own average long-term realised price net of selling costs — gives a pre-FID plant a guaranteed buyer and saves it building a sales team; the market-risk pillar of a final investment decision.

Older catalytic converters with lower PGM content signal a stockpile near exhaustion, so recycling supply should normalize.

Laser-based isotope separation invented by Australia's Silex and licensed exclusively to GLE; an alternative to gas centrifuges, being commercialised at the planned Paducah (Kentucky) plant.

After a long bear market, producers price off all-in sustaining cost plus contract security, shareholder returns and a reward for waiting, not "$5 over C1"; feasibility-study costs from before the inflation wave are roughly double today.

The first tranche of capacity serves the best customers at the best prices, so later tranches earn less — from oil & gas / mining, applied to AI compute.

Jet fuel prices doubled while jet demand rose 2%: travel treated as a non-negotiable staple.

His 20-year seasonal calendar, flipped this year to long through Dec 5 because inventories must be rebuilt.

When the largest producer can sell its whole output to eastern buyers, Western utilities compete for the remaining pounds and pay a lasting premium for material from trusted jurisdictions rather than one global clearing price.

Mobile inventory draws, upflexing legacy contracts at old prices and deferred buying slow a price rise, but once used they are gone; with most already pulled, the next supply surprise (e.g. an Arrow delay) hits price directly.

Tonnes of waste rock moved per tonne of ore; lower means cheaper mining (Santo Tomas ~1.38:1 vs a ~3:1 average).

China, the marginal buyer, steps up purchases and restocks below ~$80 Brent and backs off near ~$100+, bracketing crude.

The fee smelters charge miners to process concentrate; when it falls toward zero or goes negative (smelters paying miners), concentrate is physically scarce.

Russian-designed VVER reactors historically ran only on Russian fuel; Western fabricators (Westinghouse, Framatome) now qualify substitute fuel assemblies, and licensing approvals and first loads are the proof points of a real exit.

Regulators' core question for an underground mine is whether excavated rock generates acid; routing development through non-acid-generating host rock and shipping sulfide ore off site simplifies water-quality permitting.

Mine high-grade ore and truck or barge it to someone else's existing, permitted mill and tailings facility instead of building your own; only works at grades high enough to pay the haul (10–12 g/t underground gold in Contango's test) and needs a real mine plan, not a bulk sample.

A widening local-grade premium to Brent shows cargoes not arriving (Forties for Europe; Oman futures for what Asian refiners pay vs European).

Uranium contract price tracks the market inside a floor/ceiling band; utilities now accept $80s floors and $150+ ceilings.

States (e.g. Texas, Wyoming) that took over NRC licensing authority after the 1970s reactor boom outran the NRC; producers deal with a single state agency instead of the NRC.

Reserve categories: PDP (proved developed producing, what banks lend against), 1P (proved), 2P (proved + probable). Mid-cycle, a producer's cash-flow multiple should rise to its PDP reserve-life years; late-cycle, it moves toward 2P years.

Most uranium sells under long-term contracts, so a flat spot price can hide a term price at record highs; the daily spot quote can mislead in both directions.

A trader signs a future delivery contract with a utility, buys physical in spot today and keeps the spread; it props spot while term sits well above it.

Robin Brooks' framework: Persian Gulf exports as a share of pre-war capacity combined with a ~0.15 price elasticity of demand implies the percentage premium over pre-war Brent (15 mb/d of 20 → ~33% → ~$95).

Utilities' uncovered needs rise while producers' uncommitted future output (the empty chairs) is steadily contracted away - late buyers are left without supply.

Uranium mined in the US and free of international peaceful-use safeguard obligations - the only material usable for US defence needs, so a restricted buyer pool can pay a premium over world price.

75-80% of silver is recovered after copper, lead, zinc and gold for $5/oz or less, primary mines average under $20, and scrap flows back at high prices - why sustained extreme silver prices draw supply.

KHNP's 1.2M SWU tender is about 25M lb of uranium equivalent at 0.25% tails; an enrichment-only tender is still primary uranium demand.

More renewables/EVs raise oil demand and vice versa (Chinese solar in Africa lifting diesel use) — EV-vs-ICE and solar-vs-coal are false binaries.

Abundance and availability first, then affordability, with geopolitical security as the overlay and environmental concerns after; CO2 is 'no one's organizing principle'.

Chile copper: grade vs electricity per ton correlated -0.91 over 20 years; lower grades mean more rock processed and more power per ton.

The option to take up to ~30% more pounds under an old part-fixed-price contract. The blended price sits far below market, so utilities top up inventory cheaply and put off new term contracts; the option loses its value once contracts are fully market-priced.

Cameco's Cigar Lake (2035) and McArthur River (2042) mean 36M lb/yr to replace at $5-10bn, which justifies $150 ceilings over a ~$50 all-in cost.

Two-factor case for a commodity bull market: physical underinvestment plus currency debasement.

An enricher long on uranium runs low tails (underfeeding) and needs less feed. Russia buying uranium implies it now runs near Western tails, which adds primary demand.

Multiply the uncovered share for a delivery year by projected burn, and compare with the long-run coverage mean (US 9% covered for 2033 is about 40-45M lb uncovered).

Crops respond to high prices the next season; copper must be discovered, developed and permitted over many years, so deficits persist.

Reading a commodity's floor off the marginal producer's all-in cost rather than off the chart: if spot repeatedly declines to break a level and incumbents' costs sit there, no one sells below it. Curyer on uranium: $85 "would be representative of the current producers cost profile" and "there seems to be no supply out there at $85… we're at a bit of a new floor."

Sprott's framing of uranium, copper, rare earths and battery metals as commodities that matter for energy security and national security, with precious metals as the portfolio cornerstone.

Almost all North American gas storage is one of two formats. A depleted reservoir is an already-produced gas field re-used as a store: huge, cheap, geologically proven, and the low-cost way to hold seasonal volume. A salt cavern is man-made — washed out from surface — so it is smaller, shallower and much more expensive, but delivers far higher injection and withdrawal rates (the 'punchiness' LNG and power customers need). Rockpoint's six facilities are all depleted reservoirs, most 'delta pressured' — good enough quality to hold more gas than was originally produced, where facilities built in the last decade only reach original pressure.

Buy near the 60-month average (≈1,600 in Q4 2022); a ~60% year-end premium matched the 1980 and 2011 peaks.

A storage book earns in two ways. Intrinsic value is the observable summer/winter spread — computable off the forward curve. Insurance (non-intrinsic) value is what a customer pays above that spread for the right to act when something goes wrong: the value of the call and the value of the put. Insurance value expands with realised volatility rather than with the level of the curve, so a flat curve with fat tails is a rising-value regime. The split drives contracting: if insurance value is still below where you think it is going, sign short-dated and re-price later.

Selling future production for volume while leaving the price floating with spot, rather than fixing it — keeping the equity a pure play on the commodity. Curyer: NexGen is "the world's most levered company to the future price uranium and our contracting strategy will maintain that status." The flip side is no hedge if the price falls, so read every offtake for its fixed-versus-market-related split.

Funds sustaining 20–30% of open interest net long, with a record share of exchange inventory pulled into one set of warehouses by tariffs, marks a distorted market that needs only a small trigger.

Suppliers push contracts priced off an index at delivery, bounded by a floor (~$75) and ceiling (~$160), so price reporters have no base price to publish; rising floors signal supplier pricing power.