Actionable insights — Weekly SSR: the capex-discipline split, the cross-asset shock classifier, the forced-liquidation buy, the judgment-vs-EV screen, the put-selling entry

1. Split a de-rating cohort by capex discipline, not by drawdown — the "did the guide go up?" test

1. When a whole cohort is being punished for one behaviour (here: spending free cash flow on AI infrastructure), stop treating it as a cohort. Line the members up on a single question: did this quarter's report raise or hold next year's spending guide?
2. Then check whether the spend is still self-funded: does the company generate positive free cash flow after the capex, and is there a revenue line that proves customers are paying for the capacity (a cloud backlog, a paid-seat count) rather than an internal promise?
3. Own the ones that held the guide and still print cash; avoid the ones that raised the guide and guided revenue down and have no external monetization to point at. Read the accounting choices as a tell either way — a lengthened depreciation life flatters near-term earnings and should be noted, not ignored.

• Capex guide raised vs held; FCF sign after capex; an external revenue proof-point (backlog, paid seats, cloud run-rate); depreciation-life changes; and whether a revenue guide-down lands in the same print as a capex guide-up — that combination is what gets punished.

2. Classify a shock by its cross-asset signature — routable vs "the denominator"

1. Before sizing risk around a scary headline, ask which class of shock it is. A shock the system has an organ for (a war, a supply disruption) gets routed — it shows up in oil vol, freight, single-name and rates vol while broad equity gauges shrug.
2. A shock to the pricing anchor itself (central-bank credibility, currency status) cannot be routed, because it sets the discount rate under every cash flow. It gets priced as a correlation event: stocks and bonds fall together.
3. The tiebreaker is the currency. On an ordinary inflation repricing, long yields up ⇒ dollar up (rate differentials attract capital). If long yields rise and the dollar falls and equities fall in the same session, you are being shown an institution premium, not an inflation premium — that is how emerging markets trade on politics.
4. Confirm with the vol surface: check whether the term structure had already pulled deferred risk forward onto the event date, and whether the skew (tail) moved or only the centre. A tail that never moved plus a centre that rose to meet it tells you which risk the market actually feared.

• Yields up + dollar down + equities down in one session; gold rallying with real rates; a bear steepener where the 2-yr is pinned and the 30-yr does the moving; VIX term-structure flattening into a scheduled date; skew static while spot vol rises. The falsification: restored guidance, a hawk-owned hike, or visible Fed-vs-White-House daylight, followed by the 30-yr retracing and the dollar firming.

3. Do the index arithmetic before judging "the market" — decompose the benchmark by cohort

1. Split the cap-weighted index into the cohort under stress and everything else. Record three numbers for each: index weight, YTD return, and contribution (weight × return).
2. Read the residual: if the non-cohort names are compounding while the index looks mediocre, the headline is being dragged by a concentrated group — the "market" is healthier than the tape implies, and the equal-weighted index should be outperforming.
3. Translate to positioning: express the view through the healthy majority (equal-weight, cash-returning value) rather than fighting the drag, and only re-engage the cohort name-by-name via Insight 1.

• Cohort weight × return = contribution; equal-weight vs cap-weight spread; buyback authorisations shrinking in the cohort; the ROIC arithmetic on each incremental capex dollar; and the open-window buyback calendar (31% of the S&P open, ~53% within a week) as a near-term technical.

4. Buy the crash whose seller is mechanical — identify the leverage, then check the fundamentals never moved

1. When a market or sector gaps down violently, first identify who has to sell: margin-called retail accounts, leveraged single-stock ETPs, a levered fund in liquidation, trend-following CTAs. If the seller is mechanical, price is being set by margin clerks, not by analysts.
2. Then verify the fundamentals independently: are the underlying companies' prices, margins, guidance and inventories still improving through the sell-off? If yes, the move is a positioning reset, not a thesis break.
3. Size into it while the forced selling is still visible, and treat the moment the levered book is transferred (a bulk sale, a bailout buyer) as the clearing event — forced liquidations from excess leverage typically mark near-term bottoms because the overhang is gone.

• Leveraged-ETP AUM unwinding; margin-call counts; a performance letter asking investors for more money (a red flag, not a buy signal); the bulk transfer of a levered book to a bailout buyer; CTA sell estimates; and confirmation that pricing/inventory data never rolled over. Historical precedent to expect: Sowood, Amaranth, LTCM, Archegos — the book gets absorbed at a discount and the squeeze follows.

5. Underwrite a court judgment as an asset — the award-vs-enterprise-value screen

1. Screen for companies that have won a judgment materially larger than their own enterprise value. The asset is legal, not operational, so the market usually ignores it until cash arrives.
2. Underwrite the collectability, not the headline: what is the defendant actually worth, and what revenue has it generated? Haircut the award to a plausible settlement (he uses roughly one third) and re-run the upside on that number.
3. Price the timing and financing risk explicitly: how long can the appeal run, and does the company have a debt maturity before the money lands that could force dilutive issuance? Size the position to that bridge, not to the bull case.

• Award size vs enterprise value (not market cap); the defendant's revenue base and ability to pay; settlement precedent as a fraction of the award; appeal timeline; and any debt maturity inside that timeline.

6. Create the stock below the market — sell long-dated puts when vol blows out after the balance-sheet risk is removed

1. Wait for the sequence: a distressed name refinances (terms out its debt) and raises equity. The dilution crushes the stock and the option vol explodes — but the bankruptcy risk the vol is pricing has just been removed.
2. Go far out on the maturity curve (a year-plus) and well down on the strike, and sell puts there. The premium collected at elevated vol lowers your effective entry to strike minus premium — compute that number and compare it with the spot price; that spread is the whole trade.
3. Name the two binary conditions the thesis needs and confirm at least one is structurally likely (a regulatory change, a refinancing already done). Optionally pair a small common position with the puts so you participate if it just runs.

• A completed refinancing/equity raise as the trigger; implied vol blow-out; the effective entry (strike − premium) vs spot; expiry long enough to cover the catalyst; and a sell-side analyst who cut the target but kept the thesis.

7. Buy the post-earnings gap when the miss is in the lagging metric and the monetization metric is inflecting

1. When a stock gaps down on earnings, separate the metrics into the input the market watches (users, units, seats) and the monetization that actually drives revenue (revenue per user, price, take rate).
2. If the miss is a small shortfall in the input while the monetization line beats by a wide margin — and guidance for next quarter is above consensus — the sell-off is a reaction to a headline number, not to earnings power.
3. Add on the gap in defined increments rather than all at once, and carry an explicit price target from your own model so the add is a valuation decision, not a dip-buying reflex.

• Which metric actually missed and by how little; ARPU/price/take-rate acceleration; next-quarter guidance vs consensus; whether the international line is still growing while the domestic one plateaus; and a pre-existing model so the target isn't reverse-engineered from the drawdown.

8. Value a thin-margin marketplace on the slope of per-unit profit, not the level

1. Decompose one transaction end to end: gross order value → the platform's take → the costs that come out of that take → the residual profit per unit. Do it in dollars, not percentages — the number will look terrible.
2. Then plot that same residual over several years. A business that has gone from a per-unit loss to a per-unit profit is compounding two ways at once (volume and margin), which no single-year multiple captures.
3. Extrapolate conservatively — volume × improved per-unit profit — to get an out-year operating-profit figure, and set a price at which you'd act. Decide upfront how you treat stock-based comp, and be honest that adding it back flatters the number.

• Per-unit profit trend over 3-5 years (not the current margin); take rate vs cost-to-serve; the SBC treatment; order/volume growth runway; and a price level that converts the watch into a buy.

9. Stress-test a "new entrant" scare on capability, cost and share — before selling the incumbent

1. When a headline says a domestic/state-backed rival will break an oligopoly, ask three questions in order: can it make the high-margin product at all (not just the commodity version), what is its cost per unit versus the incumbents, and what share does it actually hold versus the share needed to matter?
2. Convert the threat into a revenue number for the incumbent rather than a narrative — units shipped by the entrant against units shipped by the incumbent, and the percentage of sales genuinely at risk.
3. Cross-check the entrant's own valuation: if it trades at a large multiple of the incumbents it is displacing, the market is pricing a story the operating data doesn't support, and the incumbents are the mispriced side.

• Whether the entrant can make the premium SKU at all; cost per unit vs incumbents; current share vs the share needed to set price; units shipped as a % of the incumbent's volume; and the entrant's multiple relative to the companies it supposedly disrupts.

10. Scale duration to a yield threshold, not a forecast — and never through a leveraged product

1. Set an explicit level on the benchmark yield where the risk/reward flips, and a rule for what each increment costs the market (his: ~2% off the index per 10 bps, accelerating above 5%). That converts a macro view into a mechanical trigger.
2. Start with a small tracking position so the trade stays live and monitored, and pre-commit to scaling it aggressively only when the threshold is hit — the entry is the level, not the narrative.
3. Run the same rule in reverse on the short side: as the yield approaches your threshold, take the short-bond bet off rather than pressing it; a resolved catalyst closes it entirely.
4. Express all of it with unlevered instruments. Daily-resetting leveraged/inverse products decay in choppy markets regardless of whether the underlying view is right.

• The 4.90-5.00% band on the 10-yr as the risk-off trigger; the Japanese 30-yr above 4% (a global carry threat that could force Fed swap lines to the BoJ); WTI below $70 as the long-end release valve; 2s30s steepening; and the decay math on any daily-resetting product before using one.

11. Sell the squeeze you didn't cause — trim into rallies driven by other people's covering

1. After a violent one-day rally, ask what moved it. If the answer is a forced unwind — a levered fund's book being transferred, a crowded pair trade covering — the price is being set by flow, not by a change in the company's prospects.
2. Trim held positions into that flow, prioritising the names that rallied hardest and that sit on the short leg of the unwinding pair (they get bought regardless of fundamentals).
3. Keep the trims partial and log them, so the position survives if the move is real. Pair with a rule for the other direction: buy small in the panic (Insight 4), sell into the mechanical bounce.

• A named forced seller/unwinder behind the rally; which leg of the crowded pair you own; the size of the one-day move versus the fundamental news; and whether the trim is partial (view intact) or a full exit (price did the work).