How this desk works
11 philosophies. 7 methods. 10 quality sleeves. 3-factor screen.
Quality 80% · price 20%. First the three-factor screen (forward P/E ≤ 20, net margin ≥ 15%, ROIC ≥ 15% — ROE for banks and underwriters). Then the published gates: quality sleeve at least 70 so cheapness cannot carry the 80/20 mix, blended score at least 70, and a quote at least 30% below the desk’s estimate. All of it, every time. An AI analysis tool on a value-investing method — not a broker, not an analyst, not a recommendation to buy or sell. Past results do not guarantee future results.
The three-factor screen is an intersection, not a compromise — a name clears only if it is simultaneously cheap on its own earnings (Yahoo forward P/E at or under 20 — price ÷ Yahoo forward EPS; trailing vs forward differing by more than 20% is flagged, not silent), structurally profitable (net margin at least 15% for an operating company; Yahoo ROA at least 1% for a bank — NI / net interest income is not a manufacturer's 15% test; net income / revenue for an underwriter when Yahoo omits the print; omitted for REITs), and efficient with capital (ROIC at least 15%, or ROE for a bank or underwriter — each reinvested dollar actually earns) — so one outstanding print does not get a name through; it has to show no visible weakness on all three at once. PEG is a Lynch tape print, not a Pass box and not a fourth vote in any average — Buffett does not use it. Forward P/E ≤ 20 is the entry gate on this list on purpose: quality-adjusted earnings power is how a name already on the list is valued, not a back door for a 24× compounder.
Flags are auto-detected: red (accounting, leverage, dilution, heavy stock-based compensation, pipeline / patent book, IP licensing, for-profit education regulation, auto OEM cycle, non-franchise semiconductors, customer-concentrated design houses, Lynch PEG over 2), green (quality core — high ROIC, clean book, no red flags — Lynch PEG at or under 2 — and a ROIC or ROE series that holds at 15% for three or more years), and caution (commodity cycle, a one-year ROIC or ROE spike, an incomplete bank tape without CET1/NPL/NIM, and shareholder-realizable value on a VIE or China-ops receipt). A bank without CET1/NPL/NIM/CRE cannot publish Pass. Named sector flags keep a name Size-limit — still a candidate, smaller sleeve. The published estimate is Base. The house print before the Street rule is Bull only. MOS uses Base. Max buy is 70% of Base (30% MOS). Screening finds candidates — not a buy ticket. A low multiple is not automatically cheap if minority holders do not control the assets, information is hard to verify, or cash may not legally reach this ticker. Method A raises the discount rate on those names into the 11–14% band (a mature US franchise stays around 9%). Method B is a 30/50/20 weighted-scenario tape, not the published IV (30% × 70 + 50% × 120 + 20% × 165 = 114). Method C takes an explicit 25% haircut off the published estimate for jurisdiction, governance, VIE, geopolitics, and weaker transparency. Adjusted value = economic DCF × (1 − haircut). MOS = (adjusted − price) / adjusted. Worked tape: 125 × (1 − 0.25) = 93.75 at 82.25 is 12.3% MOS, not 34.2% without the cut. Method D caps a VIE / ADS line at 5% of a book even if it looks cheap — valuation and size are separate. None of A–D is a Pass box. Do not pay for ruin risk with a low multiple alone. A commodity or DRAM-cycle print stays on Search as a score — it cannot publish Pass, TopPicks, or a first-clear. PEG and the moat series are tape flags, not a Pass box and not a score vote. Sector caution and the public money-flow tape sit under the score. A cluster of one sector on TopPicks is a mix warning, not a buy list.
Four layers of risk
Business, finance, structure, jurisdiction
A name can clear growth, margins, and the published bar and still deserve a lower multiple. Economic value and value for the holder of this ticker are not the same question. How sure is it that future cash will reach, legally and economically, the owner of this security?
- The minority holder does not directly control the economic assets (VIE or offshore wrapper).
- Information is harder to verify than on a domestic 10-K franchise.
- Management controls capital allocation with little minority recourse.
- Legal and commercial rules can change quickly.
- Geopolitical risk and delisting sit outside a DCF.
A. Business risk
The operating company can still be an excellent business. That is not the whole print.
- Competition
- Demand
- Price pressure
- Rising costs
- Technology
- Whether management can execute the strategy
B. Financial risk
Whether reported profit is cash the owner can keep. Klarman, Spier, cash conversion, and SBC already sit under the score.
- Debt
- Liquidity
- Profit converting to cash
- Capital needs
- Dilution from stock-based compensation
- Currency risk
C. Structure and governance
The ADS is not the operating company. A VIE or Cayman/Ireland wrapper can leave the minority holder without control.
- Rights of an investor who buys ADSs
- VIE structure
- Founder control
- Reporting transparency
- Capital-allocation discipline
- Minority-shareholder protection
D. Jurisdiction and geopolitics
Rules, audits, and repatriation sit outside a US cost of capital. A low multiple is not automatically cheap.
- Local and international regulation
- Tariffs and trade restrictions
- US–China relations
- Audit verification
- Delisting
- Capital transfer and repatriation
How risk enters the estimate
Method A — Higher discount rate
A mature US company is discounted around 9%. A name with VIE, ADS, or US–China geopolitical risk uses 11–14%. The published 30% price gate does not move — the estimate gets more conservative.
Method B — Weighted scenarios
30% × bear + 50% × base + 20% × bull. Worked tape: 30% × 70 + 50% × 120 + 20% × 165 = 114. Shown on the tape. The published IV is not this blend.
Method C — Explicit haircut
A separate 25% cut for jurisdiction, governance, VIE, geopolitics, and weaker transparency. Adjusted value = economic DCF × (1 − haircut). MOS = (adjusted − price) / adjusted. Worked tape: 125 × (1 − 0.25) = 93.75 at 82.25 is 12.3% MOS, not 34.2% without the cut. Applied to the published estimate on VIE / China-ops receipts.
Method D — Position limit
Even when the print looks cheap, a VIE / ADS line does not get to dominate a book. Valuation and size are separate decisions. Cap is 5% of the book. Not a Pass box.
Layers A and B already vote as accounting flags when the filings trip them. Layers C and D are a caution on VIE and China-ops receipts. Method A raises the discount rate. Method B is the 30/50/20 scenario tape. Method C takes 25% off the published estimate. Method D caps a VIE line at 5% of a book. None of these is a Pass box, and none of them moves 80/20.
Adjusted value = economic DCF × (1 − haircut). MOS = (adjusted − price) / adjusted. Published MOS uses the adjusted value.
Desk practical formula: adjusted value = economic DCF × (1 − haircut). MOS = (adjusted − price) / adjusted. Worked tape: 125 × (1 − 0.25) = 93.75 at 82.25 is 12.3% MOS, not 34.2% without the cut. Method A sets the 11–14% rate. Method B is the 30/50/20 scenario tape. Method C takes 25% off the estimate for VIE / jurisdiction / opacity. Method D caps that line at 5% of a book even if it looks cheap. Not a fifth Pass box. 80/20 and the 30% price gate do not move.
Institutional rule: do not pay for ruin or permanent-loss risk with a low multiple alone. Measurable risks get scenarios and probabilities. Risks that are hard to measure get a position limit. Weaker transparency demands more MOS — the 25% haircut is how that demand hits the published estimate; the 30% gate does not move. A high-jurisdiction name does not become a central book line just because it looks statistically cheap.
PDD institutional tape: bear 55–70 (central 62.5), base 90–105 (central 97.5), bull 125–150 (central 137.5). At 82.25 the quote sits ~32% above bear — no protection in a severe scenario. Base MOS is 15.6%, short of the 30% gate a high-jurisdiction name needs. Bull MOS is 40.2%, but a committee does not buy on the optimistic case. Disciplined zone on base: 97.5 × 0.75 = 73.1 (25%) and 97.5 × 0.70 = 68.3 (30%). Tape only. Published MOS still uses Method A + C. The 30% gate does not move.
The quantitative tape — forward P/E, PEG, net margin, ROIC, ROE, debt/equity, current ratio, EV/EBITDA, and beta — all appear on the score card. PEG and beta are Lynch / market prints: shown, not voted — Buffett does not use them. EV/EBITDA votes only on capital-heavy names. Current ratio is omitted for banks, underwriters, REITs, and capital-light franchises whose working capital is the product. Not a fifth screen box.
Risk warning. Sick of Gurus is an AI analysis tool. It applies a published value-investing method — quality of the business first, then a margin of safety — to delayed quotes and filed fundamentals. It is not a broker, not a bank, not a hedge fund, not a registered investment adviser, and not an analyst. Scores, intrinsic-value estimates, Guru Take, Top Picks, alerts, Track record, and portfolio snapshots are general educational information. They are not personal advice, they do not take a reader’s personal circumstances, tax, or risk capacity into account, and they are not a recommendation to buy or sell any security. Past results do not guarantee future results. You can lose money, including the entire value of an equity. Figures are as of the timestamp on each print. Quotes are delayed. Fundamentals are the latest filed period. There is no obligation to update a snapshot after it is published. The operator of this desk may hold, or may have held, names that appear on public lists. Personal holdings do not move the published bar.
Fast growth does not get a different bar. A live tape without quality at least 70 and 30% below estimate is agility toward the wrong number. This desk re-scores the same two gates every time you open a name. Delayed Yahoo, not a Bloomberg clock. The decade of filings does not pivot because a quarter was busy.
Philosophies · 11
Warren Buffett
Owner earnings, franchise quality, margin of safety.
Charlie Munger
A few wonderful businesses. Invert. Avoid being stupid.
Philip Fisher
Scuttlebutt. Own a decade, not a quarter.
Benjamin Graham
Bond-yield-corrected formula — a conservative floor, not a franchise vote.
Peter Lynch
PEG=1 growth cross-check. Shown on the tape; not a Pass box. Not allowed to outvote owner earnings on a compounder.
Seth Klarman
High leverage needs extra margin of safety.
Michael Burry
Inventory and receivables outrunning sales.
Ray Dalio
Mix, not a stock-picking cult. 13F on Holdings.
Guy Spier
Cash from operations has to fund the business.
Stanley Druckenmiller
Concentrate when the facts change. Lagged 13F.
Bill Ackman
A short list, still run through our two gates.
Valuation methods · 7
Each name uses the methods that match the business. A bank is not a DCF. A REIT is not Graham. A compounder is not a 1962 P/B rule.
Owner-earnings DCF
Compounders — free cash, split-adjusted shares. Not used for banks, underwriters, or asset managers.
Quality-justified earnings power
A multiple this quality of business can historically support.
Graham formula
Conservative floor. Equal vote only on cyclicals, not on Apple-class franchises.
Lynch PEG=1
Growth cross-check. Withheld when it implies a P/E no wonderful business trades at.
Excess return on book
Banks and underwriters — not a corporate FCF DCF. Residual income is capped at 2× book so an 8% US rate cannot mint a 50% bargain at 2.3× P/B.
FFO / AFFO
REITs — funds from operations, not GAAP depreciation.
Mid-cycle earnings
Cyclicals — a peak year is not compounded forever.
Quality sleeves · 10
These vote inside the 80% quality score. Years of cash is a sleeve, not a fifth screen box.
- Return on capital
- Profitability (ROE, capped when buybacks inflate it)
- Operating margin
- Net margin
- Revenue growth
- Balance sheet
- Net debt / FCF (years of cash)
- Current ratio (voted on capital-heavy names; shown on a compounder)
- Beta (versus the market — shown, not voted)
- Share count (buybacks vs dilution)
Screen · 3 factors
Intersection, not a compromise. All applicable boxes at once.
- Forward P/E ≤ 20
- Net margin ≥ 15% (ROA ≥ 1% for banks — NI/NII is not the test; NI / revenue for underwriters; omitted for REITs)
- ROIC ≥ 15% (NOPAT / (equity + debt − cash); ROE for banks and underwriters; fails on negative book or P/B > 10)
The score
- One 0–100 number: quality 80% (Buffett franchise — ROIC, cash, leverage, share count), value 20% (price versus the estimate). The mix does not move.
- Quality has 10 sleeves. Beta and PEG are shown, not voted — Buffett does not use them. Those sleeves are not a fifth screen box.
- Flags from filings can take up to 20 points off.
- Bands: 90–100 Exceptional · 70–89 Strong · 50–69 Watch · 40–49 Risky · 0–39 Avoid.
The published bar
- Three-factor screen first: Forward P/E ≤ 20; Net margin ≥ 15% (ROA ≥ 1% for banks — NI/NII is not the test; NI / revenue for underwriters; omitted for REITs); ROIC ≥ 15% (NOPAT / (equity + debt − cash); ROE for banks and underwriters; fails on negative book or P/B > 10). Intersection, not compromise.
- Quality sleeve at least 70, and blended score at least 70.
- Price at least 30% below the estimate.
- A Strong print at a fair price still misses the cheapness gate. A cheap print that fails ROIC still misses the screen. PEG is Lynch tape, not a Pass box.
- Guru 13Fs never move the score.
- Bloomberg OpenFIGI and SEC EDGAR are identifiers and filings — they never move the score.
- ECB Frankfurter (and open.er-api.com when ECB has no print) is a USD hint only. London GBp is scaled to pounds so the tape and the score agree.
- GLEIF, Wikidata, Nager.Date holidays, and FRED prints (10-year, 2-year, 2s10s, BAA, HY OAS, unemployment) are identifiers and context — they never move the score.
- SEC companyfacts and Companies House / SEDAR+ / EDINET / ASX / ESMA registers are filings — they never move the score.
- The quote is delayed Yahoo, re-scored every time you open a name. The filings are history on purpose.
Independent reprint — before TopPicks, alerts, or a Pass
Before a name is published, the engine reprints the score a second time — the same pass on Search, TopPicks, Alerts, Ask, and notifications. Independent means that second pass of the same rules — not a human committee. It checks three things: the published gates (three boxes, quality sleeve at least 70 so cheapness cannot carry the 80/20 mix, blended score at least 70, and at least 30% below the estimate), that the MOS and 80/20 numbers match the arithmetic, and that the industry, country, and units model is the right one for this business — including a dated Yahoo quote, an unrounded Street cap, and no quarterly YoY treated as annual growth. A miss on any reprint stays off TopPicks, alerts, and a Pass on watchlist or portfolio.
Gates
Three-factor screen, quality sleeve ≥ 70 (the 80% Buffett sleeve — cheapness cannot carry the bar), blended score ≥ 70, and at least 30% below the estimate. A commodity / DRAM cycle cannot publish Pass. Limited confidence cannot pass.
Arithmetic
MOS reprints from price and estimated value. Cheap / fair / expensive matches that MOS. The 0–100 number reprints from 80% quality and 20% value minus named flag deductions — a reader can reconstruct the headline from the two sleeves plus the flag line.
Sanity
Industry, country, and units: banks and underwriters stay on book (P/B ≤ 1.5 for a 30% MOS, IV ≤ 2× book, local CoE on ADRs) and use ROE ≥ 15% on the capital box — classic ROIC (NOPAT / invested capital, net of cash) is not interpretable when capital is the product; ROIC cannot pass on negative book or P/B above 10 (a stub return is not a 15% engine, and ROA is not a substitute Pass); a published estimate is Base = min(desk, 1.25× the Street target) (Yahoo mean when the median is a rounded dollar) when the desk sat above that cap, so a 30% MOS cannot be manufactured by a high multiple; delayed Yahoo is the live source for US and local listings (Madrid, London) — an empty Alpha Vantage print is not a fail; annual growth is consensus +1y or a multi-year EPS trend, never quarterly YoY; YTD/H1 stub years are dropped before any trend; a name without a dated live quote cannot publish; REITs on FFO; utilities on cash yield; cyclicals on mid-cycle earnings; asset managers get no Graham and no cash DCF; lenders are not Visa; memory is a cycle; mixed-unit P/E or market-cap cannot mint a bargain. A leftover ADR book (tenge vs a dollar quote) is not a 1-cent P/B. Buybacks that take book through zero use ROA on the quality sleeve only — not on the Pass box.
Names that cleared all four boxes
A worksheet of the intersection
These names sat inside every box on the same print. Adobe is not here for a 63% ROE and Century Aluminum is not here for a 6.5× P/E — they are here because none of the four was a miss.
Swipe sideways for Graham MoS →
| Ticker | Fwd P/E | PEG | Net margin | ROE | Cyclical | Growth cap | Graham MoS |
|---|---|---|---|---|---|---|---|
| ADBE | 10.70 | 0.71 | 28.7% | 62.9% | No | 25% | 57% |
| CENX | 6.48 | 0.06 | 22.6% | 51.7% | Yes ⚠ | 12% | 75% |
| MTCH | 10.33 | 0.37 | 20.2% | 49.9% | No | 25% | 75% |
| PDD | 8.04 | 0.89 | 20.4% | 22.8% | No | 25% | 66% |
| HRB | 8.33 | 0.67 | 18.6% | 71.4% | No | 25% | 73% |
| FSLR | 12.79 | 0.55 | 32.5% | 18.5% | Yes ⚠ | 12% | 62% |
| BSX | 14.68 | 0.69 | 17.5% | 15.3% | No | 25% | 62% |
| BKNG* | 19.49 | 0.89 | 25.5% | 226%* | No | 25% | 58% |
| APP* | 19.92 | 0.88 | 64.6% | 204%* | No | 25% | 55% |
| APA | 6.68 | 0.84 | 19.6% | 26.7% | Yes ⚠ | 12% | 63% |
| META | 19.08 | 0.85 | 29.8% | 29.8% | No | 25% | 60% |
| GOOG | 16.37 | 0.91 | 54.8% | 48.7% | No | 25% | 62% |
| SKHY ⚠ | 6.08 | 0.31 | 85.6% | 92.7% | Yes ⚠ | 12% | 70% |
Worksheet snapshot (1 Sep 2026): four-factor pass plus Graham-style MoS ≥ 30%. Not TopPicks. The live desk uses ROIC (ROE for banks and underwriters) and a bond-yield-corrected Graham floor — owner-earnings DCF still votes on a compounder. A Graham worksheet can print 50–85% MoS; that is a starting point, not the public bar. Prints move. Not a recommendation to buy or sell.
* Buybacks and treasury stock can shrink book toward zero, so headline ROE in the hundreds is not a quality signal — read the net margin (and live ROIC) instead.
⚠ Aluminum, solar, oil, and memory use a 12% growth cap in this Graham worksheet (instead of 25%) so a peak year is not compounded forever. SK hynix still prints a large MoS after that cap — treat it as a DRAM-cycle peak, not the cheapest name on the list.
Methodology
How that worksheet is built
- Three-factor screen, all three at once: forward P/E ≤ 20, net margin ≥ 15%, ROIC ≥ 15% (banks and underwriters use ROE — classic ROIC is not the print when capital is the product). PEG is a Lynch tape print — not a Pass box and not a fourth vote. Lynch called 1.0 fair and under 1 cheap; 1.5 is still reasonable growth. Buffett does not use PEG, so a missing or high PEG cannot knock a name off the bar.
- Worksheet intrinsic value is Graham-style: IV = EPS(TTM) × (8.5 + 2 × g%), where g% is implied as Fwd P/E ÷ PEG, then capped.
- Growth cap on that worksheet: 25% normally, 12% when the name is cyclical (aluminum, solar, E&P, memory). The live desk caps Graham at 20% / 12% and still applies the AAA bond-yield correction, so the live Graham floor is more conservative than this snapshot.
- Margin of safety on the worksheet is (IV − price) / IV. Qualification on that sheet is MoS ≥ 30% — the same cheapness number as the public price gate, but versus this Graham print, not versus the live blended estimate.
- On a compounder the live estimate is owner-earnings DCF plus quality-justified earnings power. Graham is shown as a floor and is not allowed to outvote owner earnings. META and GOOG can clear a Graham worksheet and still miss the public cheapness gate.
Checked and excluded
- UBER, CRM, and NVO were checked separately on the 1 Sep 2026 worksheet when PEG was still a box. Live Pass no longer gates on PEG.
- VIST (Vista Energy) was excluded from that worksheet: no PEG print, so implied growth / Graham IV could not run. Live Pass no longer needs a PEG.
- BR was 0/4 on an earlier check.
Live Graham still uses 12% on cyclicals. The worksheet's 25% non-cyclical cap is the snapshot only — the live floor is 20%, then the AAA yield correction.