Ten managers' repeatable methods in one article: how each actually finds ideas — the screens, the diagnostics, the sell disciplines — extracted so they can be rerun on different names.
1. The rotation screen — sell what everyone owns, buy what the stampede left behind
The repeatable method
- Identify the crowding: which stocks do investors own because index weight forces them to ("people aren't investing for the business fundamentals" — Rogers), not because of the fundamentals?
- Screen the sectors the stampede exited — energy, materials, healthcare, consumer — for names with intact earnings power at multi-year-low multiples.
- Demand a mechanism, not just cheapness: buybacks funded by free cash flow (Witmer), an activist (Rogers' MAT), an acquisition wave (Giroux's biotechs), a guidance inflection.
Here: all ten panelists gave a variation of the same advice — flee overowned AI tech, buy the left-behind: BCO at ~5x future earnings, MAT at half of private-market value, oil majors at 8x, OI under $10.
Watch for
- Index-concentration extremes (top-10 = 39.2% of the S&P — Black); earnings growth OUTSIDE the AI complex re-accelerating (Giroux: the deceleration in AI earnings will itself make the rest of the market more attractive).
2. Giroux's capex-boom diagnostic — peak multiples on peak-ish earnings
The repeatable method
- In any capex-driven boom, remember the accounting asymmetry (Jain): the spender books the cost over 5–7 years of depreciation, but the recipient books it as revenue and profit TODAY — so suppliers' earnings overstate the durability of the cycle.
- For each supplier, back out the implied multiple on the boom business alone (Giroux on CAT: a 38x consolidated NTM P/E = ~230x the current data-center power business, ~70x its projected PEAK earnings).
- Check the supply response: when every player is adding capacity into the boom (CAT, Cummins et al.), peak earnings arrive just as capex growth slows — sell the long-cycle suppliers first, since the spend mix shifts short-cycle as budgets tighten.
Here: negative CAT, GEV (50x — "isn't sustainable," Jain), ETN, DELL, SMCI, HPE; the same logic makes the hyperscalers (MSFT, AMZN, GOOGL) winners — they're the spenders whose 2030–32 FCF will "shock" once the spending normalizes.
Watch for
- First hyperscaler capex-growth guide-down; order-book vs shipment divergence at long-cycle suppliers; capacity-addition announcements clustering at the top.
3. Giroux's patent-cliff M&A basket — buy what big pharma MUST buy
The repeatable method
- Size the buyer's need: large pharma faces a $400–500B revenue hole from patent expirations it cannot fill with internal R&D — and its shareholders WANT deals.
- Screen SMID biotechs with Phase 2/3 assets that could plausibly become $2–10B products — the size that moves an acquirer's needle.
- Buy a BASKET (he names seven), because any single trial can fail; the exit is a 50–100% takeout premium, and the fallback is trial success itself (CYTK: $30s → $80s on Phase 3 data, no deal needed).
Here: ASND, MLTX, DYN, PCVX, CYTK, CGON, DNLI — "I would be shocked if all weren't acquired"; validated by AbbVie/Apogee ($11B deal; the BUYER gained $20B of market cap on announcement — deals are win-win, so they'll keep coming).
Watch for
- Acquirer stock reactions to deals (a rising-acquirer tape sustains the wave); Phase 3 readouts in the basket; large-pharma cash-flow strength (funds the deals).
4. Witmer's merger-mechanics screen — buy the acquirer the arbs are shorting
The repeatable method
- When a good operator announces a sound cash-and-stock acquisition and the stock FALLS, diagnose whether the selling is mechanical: (a) income/buyback holders exiting because capital returns pause pre-close; (b) risk arbitrageurs shorting the acquirer against the target (visible in rising short interest); (c) no buyback bid under the stock.
- Model the combined company conservatively — hold both sides' operating income flat, credit only announced cost synergies — and value it on after-tax free cash flow per share at the point the balance sheet is repaired (2–3 years out), at only 10–11x.
- If even the conservative case shows a multiple of today's price, the mechanical selling is your entry. Bonus check: management with engineering + M.B.A. backgrounds (her favorite).
Here: BCO at ~$100 (bought in the past month): flat-lined operating income + $200M synergies still yields $15.50 EPS / $18 FCF by 2029 → $180–250 target at 10–11x FCF, from ~5x today.
Watch for
- Deal close (Q1 2027 synergy start) → buyback resumption → arb shorts covering — three mechanical reversals, each a catalyst; the same setup on the next well-received-by-analysts, sold-off-by-flows acquisition.
5. Rogers' private-market-value discipline — price every business like a buyer would
The repeatable method
- Estimate what an informed private/strategic buyer would pay for the whole company (PMV) — not where the stock trades.
- Buy at wide discounts (his current book: MAT at ~50% of PMV, ZBRA at 40% off, JLL at 35% off, SMG at 20% off) — small/mid-caps left behind by index flows.
- Require a closing mechanism: an activist pushing a sale (Southeastern at Mattel), a friendly regulatory window for consolidation, buybacks, or a new CEO refocusing the core (SMG's Baxter — whom he vetted through multiple in-person meetings before recommending).
Here: MAT $13 vs $25–26 PMV with an activist letter on the table; ZBRA at 14x (from >20x) on a fear — memory-chip costs — management is already mitigating.
Watch for
- Strategic interest surfacing (Hasbro), regulatory posture toward consolidation, PMV-gap closure via buybacks; meet (or at least study) new management before crediting a turnaround.
6. Black's four-factor value screen — with the ex-cash, ex-SBC P/E check
The repeatable method
- Screen for all four together: high return on equity, LOW absolute P/E, strong free-cash-flow generation, sustainable earnings power. Build your own earnings model — don't take the Street's number.
- Compute the honest multiple twice: subtract net cash per share from the price before dividing by EPS (cheaper than it looks), then ADD BACK stock-based compensation to the denominator (more expensive than it looks). Truth is between; prefer companies reporting unadjusted GAAP ("what you see is what you get" — URBN).
- Verify shareholder yield: actual buyback dollars and share-count shrinkage, not authorizations.
Here: EXPE: his $20.04 EPS vs Street's $19.77; 11.7x ex-cash, 14.1x adding back SBC — still cheap for 17–18% growth; $4.56B bought back in 2¼ years. URBN: unadjusted GAAP, 11.5x, 17.7% ROE, 20%+ growth.
Watch for
- Widening gaps between adjusted and GAAP EPS elsewhere (a red flag by this method); the P/E differential vs the sector leader (EXPE vs BKNG) as the mean-reversion trade.
7. Jain's two-year-yield tell — and energy as the hedge that pays you
The repeatable method
- Track the 2-year Treasury as the market's true rate forecast: persistently above 4% while the consensus prices cuts = the bond market calling the equity market's bluff; rising 2s raise the cost of capital across speculative ecosystems first (AI capex chains).
- Cross-check the index's composition: when the S&P's leadership is unusually cyclical (GEV at 50x), the index deserves a LOWER multiple — leadership quality, not just level, sets the fair P/E.
- Hedge the scenario with the sector that CAUSES it: energy rallies on the same oil/inflation impulse that forces hikes — defensive, cheap (8x), yield-paying insurance.
Here: 2-yr >4% since mid-May + BofA forecasting three hikes → he adds the oil majors (XOM, TTE, PBR, BP, 0857.HK) with the SPR at a 43-year low: "if [2-year yields] keep rising, the unraveling will be painful. Energy would be very defensive."
Watch for
- The 2-yr breaking above/below 4%; oil through $70; SPR replenishment announcements; hike pricing converging toward BofA's three.
8. Ellenbogen's cash-flow-duration test — own AI's users, not (only) its builders
The repeatable method
- Split the AI complex in two: buildout suppliers (much nonrecurring revenue + technological-obsolescence risk) vs companies USING AI to gain share and cut costs (longer-duration cash flows).
- When a marquee AI release triggers an indiscriminate software/digital selloff, ask what the model actually threatens: a physical-network moat (DoorDash's driver density: $4/order cheaper than 2020) is not disrupted by a chatbot.
- Anticipate the ownership shift: once the LLMs themselves list (Anthropic's confidential filing), proxy premiums deflate and capital re-sorts by cash-flow duration — position before the discernment arrives.
Here: DASH mislabeled "terminal-value risk" post-Opus-4.5 → re-recommended toward $300; Bending Spoons uses the same fear to buy software assets cheaper; NVDA flagged as the proxy that loses when direct LLM ownership arrives.
Watch for
- The Anthropic IPO date and aftermarket (the re-sorting trigger); which "AI victims" keep posting share gains and unit-cost declines through the panic.
9. The close-out discipline — name what broke, then remove the pick
The repeatable method
- Re-underwrite every recommendation at fixed intervals (the Roundtable cadence): restate the original thesis and test each leg against evidence.
- Close when a thesis LEG breaks, not when the price falls: a competitive-structure misjudgment or a failed trial (Ahlsten's BSX) or a regime change that invalidates the driver (Desai's GLD: a hawkish Fed guts the debasement trade) — say out loud what you got wrong.
- Distinguish broken from delayed: Rossbach keeps LVMH ("delayed, not derailed" — demand intact) while Ahlsten kills BSX (structure wrong). The difference is whether the CAUSE of underperformance attacks the thesis itself.
Here: BSX closed ("we got the industry structure wrong, a trial didn't work"); GLD closed on the Warsh regime change; INDA/INDY held-but-on-watch; LVMUY/NKE/DASH re-affirmed with the thesis restated.
Watch for
- In your own book: for each loser, write the one sentence naming which leg broke — if you can't, it's price volatility, not thesis failure.