Actionable insights — Weekly SSR: a stress composite that predicts political intervention, short interest as squeeze fuel, sell-the-news discipline into capex prints, separating operating earnings from markups, the direct-selling-discount screen, and buying yield after the raise

1. Build a multi-factor stress composite to predict the political response, then trade the window — 03:43

1. When a policy-maker has a visible incentive to intervene (an election, an inflation mandate, a poll number), stop forecasting the conflict and start forecasting the intervention. Ask which variables the decision-maker actually watches.
2. Build a small composite from those variables — not a model, a dashboard of four things — and z-score each against its own history. Backtest the combined standard-deviation move at which intervention has historically arrived.
3. When the composite reaches that threshold, you have a dated window, not a direction. Position for the de-escalation trade (short-covering assets: high-beta equity, small caps, duration) and against the escalation trade (crude), and size for the window, not forever.
4. Cross-check the political clock separately: the closer a term-defining election, the stronger the incentive, and the more asymmetric the payoff.

• A decision-maker with an election or approval constraint; the 3-5 observable variables they actually track; a backtested σ threshold rather than a narrative; a dated window to trade; and the "fades overnight on renewed rhetoric" risk that caps how long you hold it.

2. Treat extreme short interest + a positioning washout as fuel, and wait for the spark — 03:43 / 28:23

1. Separate the two halves of a countertrend setup: fuel (how crowded and one-sided the positioning is) and the spark (a headline the crowd can't fade). Fuel alone goes nowhere; a spark into an un-crowded market fizzles.
2. Measure the fuel with hard positioning data, not sentiment surveys: index short interest as a share of free float versus its own multi-decade history, and prime-brokerage flow (how much gross exposure hedge funds have cut, and what share of a prior accumulation they've reversed).
3. Look for the turn in the flow data — the first week the same funds start re-buying after a large de-grossing — as confirmation the forced selling is finished.
4. Then buy the sector that was de-grossed hardest, not the broad index; the squeeze concentrates where the covering has to happen.

• Short interest at a multi-year extreme with the index near highs (the tell-tale of a crowded fundamental short); prime-broker de-grossing at record magnitude; the first week of re-accumulation; and a catalyst the shorts can't argue with.

3. Sell-the-news discipline: when a peer is punished for a disclosure you know is coming, wait for the print — 05:24 / 53:06

1. When the first company in a cohort reports and gets hit for a specific line item, identify whether that line item is structural to the cohort (everyone must disclose the same thing) or company-specific.
2. If it's structural, assume every peer prints the same problem within days. Do not pre-position long into it, however cheap the peer looks — the information is already known and the reaction is the trade.
3. Wait for the sell-the-news drop, then buy the layer of the value chain that benefits from the same disclosure (see insight 4). The bad news for the spender is confirmation for the supplier.
4. Keep the two decisions separate: "is this a good business" and "is this week a good entry" are different questions, and only the second one is being answered.

• The first reporter in a cohort punished for a shared disclosure; whether the item is structural (capex guidance, margin mix) or idiosyncratic; the calendar gap that lets you wait; and the beneficiary layer to buy with the proceeds.

4. Split reported earnings into operating profit vs mark-to-market gains — then go straight to the cash-flow statement and the commitments footnote — 05:24 / 41:42

1. On any record-earnings headline, immediately ask what share came from revaluing assets (equity stakes, investments) rather than selling products. Recompute the growth rate excluding those markups — that's the number the business actually earned.
2. Check whether the marked-up asset can even be monetized (lock-ups, sale restrictions). A gain you can't sell into is an accounting entry, not capital.
3. Go to free cash flow (operating cash flow minus capex), not EPS. Accounting spreads capex over years; cash doesn't. A company can post record EPS while burning cash.
4. Read the purchase-commitments footnote — contracted future obligations for chips, power, buildings. That's the forward capex the guidance hasn't shown you yet, and its rate of change is the real signal.
5. Translate the finding into a structural claim: sustained heavy capex to defend growth means lower margins and lower return on invested capital — a multiple story, not an earnings story.

• The markup share of net income and whether the stake is sellable; FCF versus EPS divergence; the commitments footnote and its 3-month change; guidance raised on capex rather than revenue; and management language conceding lower margins (Google guided exactly that on cloud).

5. Test revenue quality by talking to the customer who pays the bill — 19:12

1. When a segment's revenue grows while its underlying volume shrinks, the growth is coming from price or from extraction. Find a large customer and ask what changed in how they are billed.
2. Catalogue the specific mechanism changes — pricing model, targeting precision, spend caps, refund policy. Mechanism changes are verifiable facts; "monetization improvements" in a press release are not.
3. Ask whether the customer has an alternative. Extraction without substitutes produces good numbers now and a fragile franchise later — so treat it as borrowed revenue and de-rate the terminal value, not the next quarter.
4. Pair the finding with the volume data (is the underlying usage being cannibalized?) to decide whether this is a cycle or a decay.

• Revenue growth on falling volumes; billing-mechanism changes a customer can describe concretely; the absence of a substitute ("they have no choice but to continue to spend"); and whether the extraction is funding a capex programme that needs the optics.

6. The "surging earnings, shrinking multiple" special-situation screen — and name the discount you're being paid to accept — 23:38

1. Screen for the scissors: forward EBITDA estimates rising over the last twelve months while EV/EBITDA falls over the same period. Rising estimates plus a falling multiple is a market refusing to believe the numbers — that's the whole opportunity.
2. Verify the earnings are cash: measure EBITDA-to-free-cash-flow conversion and the free-cash-flow yield on the equity. A high conversion rate turns a cheap multiple into a real return.
3. Use the dividend record as the durability test — consecutive quarters paid without a cut, and whether the payout is covered by that cash conversion, not by borrowing. Check leverage post-deal is falling back toward the pre-deal level.
4. Then write down explicitly why it's cheap and decide if each reason is permanent or temporary: a category discount (the multiple the market applies to all peers), a jurisdiction/currency discount, and a transaction-digestion discount all decay at different speeds. Only the temporary ones pay you.
5. For the acquisition specifically, get the first consolidated month's contribution and management's pro-forma trailing-twelve-month EPS uplift — that's the cleanest early evidence of accretion.

• Rising forward estimates against a falling multiple; cash conversion above 80%; an unbroken dividend record with post-deal leverage normalising; a first-month accretion print; and a written list of why-it's-cheap reasons sorted into permanent vs temporary.

7. Merger-arb: convert the deal into a buy limit, and price the antitrust risk separately — 34:09 / 55:52

1. Start from the announced consideration and the expected close date, then work backwards to the price that pays you an acceptable annualised return for the deal risk. That price — not the current quote — is your standing limit order.
2. Score the antitrust/regulatory risk on its own: overlapping share in a defined market, the regulator's recent behaviour, whether the acquirer is foreign or domestic. Low-overlap consumer deals are the cheap end of the risk curve.
3. Let volatility come to you. A post-announcement drift below the limit is the entry; chasing near the deal price surrenders the whole spread.
4. On a name already owned, treat a widening spread as an invitation to add rather than a warning — as long as your regulatory score hasn't changed.

• A stated buy limit below the deal price rather than a "like it here"; an explicit antitrust score; the expected close quarter for the annualised maths; and a spread that widens without a change in the regulatory facts.

8. Hunt where nobody is looking — the cheap-quality contrarian screen — 41:42 / 01:03:30

1. When attention (and money) concentrates in one theme, invert the screen: within the unfashionable part of a sector, rank names by valuation and by quality of the underlying business.
2. Prefer the ones with no story at all. A cheap, well-run business with no narrative needs only an in-line quarter to re-rate, because expectations have been abandoned — "the lack of an issue is a positive in itself."
3. Set the bar at "beat and raise," not "beat": the raise is what forces analysts to move estimates, which is what moves the multiple.
4. Apply the same test to what you refuse to buy: a beaten-down name whose end-market budget is being structurally redirected elsewhere is cheap for a reason — wait for the deceleration to finish rather than catching it.

• A sector where attention is concentrated elsewhere; a cheap multiple with no narrative attached; a beat and raise as the trigger; and — on the reject side — a structurally redirected end-market budget that makes "cheap" a trap.

9. In income names, buy after the dividend is raised — the raise is the signal, not the yield — 56:37 / 58:49

1. For rate-sensitive income vehicles (mortgage REITs, BDCs), treat the headline yield as unreliable — it's highest exactly when the market expects a cut. Look at the dividend history instead.
2. A raise after several years of flat payments is management telling you the spread they earn has stabilised. That is a far better entry signal than the yield, because it is the one number they can't fudge and won't reverse casually.
3. State the macro condition the whole basket depends on and check it separately (here: has the long yield peaked?). If the condition holds, everything in the basket works — so size the basket, and use the capital-structure ladder to set risk: the baby bond/preferred for income safety, the common for upside.
4. Enter the commons on a price rule (a level, not a feeling), scaling in as they fall rather than sizing all at once.

• A dividend raise after multi-quarter flatness; the single macro condition the basket rides on (the long yield peaking); a capital-structure ladder from baby bond to common; and a numeric entry level for the commons.

10. Decide which side of a capital flow you're on — and stress-test the bear case on the receiving side — 37:12 / 41:42

1. Map a boom as a transfer: who is spending the cash and who is receiving it. Then check the cash-flow statements on both sides — the spenders' FCF collapsing is the same event as the receivers' revenue rising.
2. Default to the receiving side while the spenders are locked in by competitive necessity ("they cannot afford to lose the race"), because their capex is the receivers' revenue and it is contractually committed.
3. Before sizing up, take the strongest bear case against the receivers seriously and work the physical mechanics, not the headline. Cheaper inputs usually mean more total consumption, not less.
4. Confirm with pricing evidence from the receivers' own market — sold-out capacity, forecast price increases, the absence of a cheaper alternative supplier — and with behavioural tells (a powerful customer publicly flattering the supplier).
5. Keep the financing risk, not the demand risk, as the thing you monitor: if the spenders must borrow to keep spending, widening credit spreads are what actually ends the flow.

• Committed, competitively-forced capex on the spending side; sold-out capacity and forecast price rises on the receiving side; a bear case that dissolves when you work the physical mechanics; a powerful buyer publicly praising a supplier; and credit spreads as the true kill-switch.