RuleOne

The Rule #1 method

The Professor synthesised this from all 498 InvestED episodes. Bracketed numbers link to the episode notes. Where Phil and Danielle refined a rule later in the show, the later version wins. The screen, Radar and Holdings apply it; the App lines show where.

Checklist

This checklist is what the RULERS analyst and Radar agents apply, and what you can run on any stock page. It is distilled from METHOD.md, which gives an episode citation for every item. Items marked (auto) are computed by the screen. The rest need your judgement.

Radar

  • Where did the idea come from: an event, a great investor's buy or a product you use? It is a name, not a decision.
  • Any guru 13F position ≥4%, a fresh 13D/G, or large open-market insider buying? (auto: Radar, events column)

Understand (Meaning)

  • Can you describe the business in one sentence: what it sells, to whom, and what problem it solves?
  • Have you read the Business and MD&A sections of the 10-K, now and from ~10 years ago? Can you name the top rivals?
  • What hooks the customer?
  • Is it in your circle (two of three: passion, talent, money)?
  • Have you checked the weather (industry trend, country and political risk, revenue concentration)?

Moat

  • Name the moat type: brand, secret, toll, switching, price or network.
  • Does it pass the replication test (a rival with your market cap can't easily beat it)? Does it pass the disappears test?
  • Is there evidence of pricing power through past inflation?
  • ROIC ≥10% across 10/5/1 years (auto: Big Five), and not falling (auto: "ROIC falling" flag)?
  • Sales, EPS, BVPS and OCF growing ≥10% and moving together (auto: Big Five tests)?

Management

  • Total debt ≤ 3 years of free cash flow (auto: "debt > 3 years of FCF" flag)
  • Cash is real: owner earnings ≥75% of net income (auto: "cash not real" flag)
  • Capital allocation: buybacks and acquisitions done below value? Is ROIC holding up after deals?
  • Incentives (proxy): rewarded for the business, over 5+ years?
  • Candour: is the CEO letter checkable or fog? Did old plans match results?

Margin of safety

  • Ten Cap price (10 × owner earnings − net debt) (auto)
  • Payback Time price (8 years of FCF) (auto)
  • Sticker / MOS price (the optimistic case) (auto)
  • How many of the three agree? Aim for 2–3 of 3 (auto: "methods agree"). Can you explain any wide gap?

Event

  • Is there an event with fear behind it, rather than just a cheap price? (auto: "cheap without an event" flag)
  • Is it one-time and solvable in 1–3 years, without new debt?
  • Is it not a changed story, a miracle-needed problem or a bankruptcy risk?

Reduce basis (tranches only)

  • Tranche prices and sizes written down before the price moves.
  • Dry powder kept for a further ~50% fall.

Story

  • One page: why you own it, the bear case (inverted) and your rebuttal.
  • Measurable sell triggers: what would mean the story changed?
  • End with one verdict: no, yes at $X or too hard.

Markers of a wonderful business

This file is the shared contract between the screener, the agents and the site. Each marker below comes from the InvestED synthesis (METHOD.md). The screener computes it from up to 11 years of SEC filings (ruleone/metrics.py: wonderful_markers). The RULERS analyst confirms it, or overrides it with reasons, and Radar and the Editor watch for it breaking.

Marker score is the share of known markers passed (needs ≥5 known). It is shown on stock pages and added to the rank score. Markers are evidence that a moat existed: numbers look out the back window [023, 110]. They never replace understanding why the moat will last.

Marker Test Why it matters Episodes
roic_consistent ROIC ≥10% in ≥8 of the last 10 years The first proof of a moat: ten years is long enough for rivals to attack 004, 021, 022, 270
roic_not_falling Latest 3-year average ROIC ≥ 75% of the 10-year average Falling ROIC after deals means capital was misallocated and the story may have changed 270, 271, 432
growth_coherent Sales, net income and OCF 10-year growth within 10 points of each other and all positive The big growth rates should move together; tangled lines (GM) mean too hard 019, 020
margin_stable Gross margin (else operating margin) standard deviation ≤ 3 points over 10 years, with the latest not below the median Pricing power: steady margins through inflation and recessions 178, 318, 332, 362–365, 460
fcf_margin Free cash flow ≥ 10% of revenue One of Phil's two numbers tests of a moat (FCF ÷ revenue) 081, 082
cash_real After-tax owner earnings ≥ 75% of net income Earnings can be tweaked; cash can't (Ackman's 75% rule) 271, 273, 274, 476
low_debt Total debt ≤ 2 years of free cash flow Debt turns a scare into bankruptcy; Ackman's test is ~2 years and Phil's is 2–3 075, 089, 260, 428
no_dilution Share count grew ≤ 0.5%/yr over 5 years Dilution transfers value away from owners; buybacks only help below value 100, 226, 227
predictable Revenue rose in ≥8 of the last 10 years "Simple and predictable" (Ackman #1); the ten-cap only needs "it will be bigger" 274, 343, 467
recession_tested Profitable, with ROIC ≥10%, in 2020 Wait for ten years of data and a recession to see the business and management when the tide goes out 390, 442, 443

Interpretation:

  • 9–10 of 10: the numbers of a wonderful business. Now prove the moat and price.
  • 6–8: good, but find which markers fail and why.
  • Below 6: not a Rule #1 business on the numbers. Too hard unless there is a specific reason.
  • Banks and insurers: the OCF-based markers (fcf_margin, cash_real) mean little (float). Use ROE and book value growth instead [411].

Changing a marker: the Professor proposes changes in METHOD.md → Proposed app changes. The Engineer implements them with a test and updates this table.

Lessons

These are lessons learned from applying the method, written by the Professor's weekly review from the scorecard and dossier reviews. Every agent reads this file before judging. Lessons are about process ("we ignored a falling-ROIC flag"), not outcomes ("the stock fell") [294, 319, 339]. Newest first. Each lesson cites its evidence and the METHOD rule it sharpens.

Seeded from the show's own mistakes

  • Selling compounders too early is Phil's most repeated error. Don't trim a rare ~20% compounder just because it doubled. Trim only above Sticker and when the cash has a better home [467, 492, 493]. (METHOD 9)
  • Thinking without finishing costs money. Activision went unbought for want of rules set in advance (about $4M by Phil's account). Write the tranche prices before the price moves [352]. (METHOD 8)
  • Trusting managers on technology you can't judge was Phil's costliest mistake. Get independent expert help, or put it in too-hard [320, 323]. (METHOD 2)
  • A bigger margin of safety doesn't fix a terminal flaw. That is cigar-butt thinking [377]. (METHOD 6)
  • Cheap at a record high is usually a value trap. No event means no sale [240, 324]. (METHOD 7)
  • Debt turns a scare into bankruptcy. Horsehead, Boeing, cruise lines [075, 335, 365, 496]. (METHOD 4)
  • When the story changes, sell, even at a loss. Alibaba (political risk), Bank OZK (lending against what management said) [385, 478, 493]. (METHOD 9)

From the pipeline

(The Professor appends dated lessons here from the weekly review.)

  • 2026-10-09: Show the marker line, don't assume it. The ADBE, KNSL and LULU dossiers have no per-marker line. LULU's FCF margin is about 8% (fails fcf_margin) and its OCF growth lags sales (growth_coherent), and neither is mentioned. For KNSL nobody says the OCF markers are set aside for an insurer. Write ✔/✘/? for each marker, with a confirmation or override, before the price work. (METHOD 2–5, MARKERS)

This is a synthesis of the Professor's notes on every episode of InvestED (2015–2025), organised as the checklist RuleOne applies. Numbers in brackets are episode notes (/learn/NNN/). Where the hosts refined an early rule later in the show, the later version wins and the change is noted. The App lines say what the screener, site and agents do with each rule.

Not investment advice. These are study notes on a published method, and automated numbers must be checked against the filings.

0. The frame

  • Rule #1: don't lose money. Protect the downside first. Being right only 40% of the time works if the misses don't lose capital [002].
  • Price is what you pay, value is what you get. Value is the cash a business will hand its owner [006, 177].
  • Investing vs speculating. An investor values the business and would be content to own all of it. Anything else is speculation [228, 341, 357].
  • Index if you won't do the work. The method is for people who will research businesses [130, 189, 321].

1. Radar: where ideas come from (names, never decisions)

  • Ideas come from events in the news, great investors' buying and products you use [001, 011, 077, 190].
  • Cloning is a legitimate edge [292, 293, 423]:
    • 13Fs show only long US positions, filed up to ~45 days after the quarter (a snapshot lagging ~4 months).
    • 13D/G filings (owners of more than 5%) arrive within days.
    • Fewer than ~20 holdings signals a Rule #1-style investor. Treat a guru position of ≥4% as meaningful.
  • Insider buying counts when it is large against the insider's holding. Token buys are PR [277].
  • Analysts' price targets are leads, not valuations [077]. A target is a 12-month price opinion. Analysts lean optimistic because their banks want the business, so treat their growth rate as a ceiling [093, 122]. Their downgrades produce the fear that makes a price [249], and thin coverage leaves room for mispricing [440]. What the crowd focuses on tells you where good news is already priced in.
  • Radar's gate: the business is in two of your three circles, a guru holds it at 4%+, and you know the industry [253].
  • App: the analyst consensus (ruleone.analysts, refreshed on weekdays): low/mean/high targets, ratings and up/downgrades, read against the Rule #1 prices as agree / contrarian / crowded. Radar treats downgrades and target cuts as fear signals; RULERS uses analyst growth as a ceiling.
  • App: the screen's event columns (drawdown, Form 4 open-market buys, 13D, negative 8-Ks), and the Radar agent (daily news, filings and guru 13F moves for the watch list).

2. Understand (Meaning)

  • Capable of understanding, then actually understand. If you can't explain the business in one sentence, stop [001, 254, 256].

  • Research in order:

    1. The root-canal test.
    2. Use the product.
    3. Name the rivals.
    4. Test the moat.
    5. Then price.

    Read the 10-K Business and MD&A forward over many years, then the competitors' [002, 017, 153, 476].

  • Work in layers [343, 345]:

    1. Munger's four, in your head, in minutes.
    2. Ackman's eight, over about a week: simple and predictable; free cash flow; dominant; barriers to entry; high ROIC; low extrinsic risk; strong balance sheet; excellent management.
    3. Phil's ninth: price.
    4. The long (~90-point) checklist.

  • End every project one of three ways: no, yes at $X or too hard [190, 281]. New industries and startups fail the list [135].

  • The weather is the industry, country, politics and incentives around a business. A country's politics is a checklist risk [452–456, 397].

  • App: the 10-K/EDGAR links on stock pages, the Learn course, and the RULERS dossier (planned).

3. Moat

  • A moat is an intrinsic, durable advantage: brand, secret, toll bridge, switching cost, lowest cost or network effect [003, 080, 088].
  • Tests [088, 269, 376, 432]:
    • Replication: could a rival with your market cap build it?
    • Disappears: would anyone notice if the company vanished?
    • Escape from perfect competition.
  • Pricing power is the payoff. Look for sales after a price rise and steady margins through past inflation [178, 318, 362–365]. A brand isn't automatically a moat [460].
  • Tech split [396]: an ecosystem with switching costs is predictable. A product-cycle business has no moat to count on.
  • The numbers look out the back window. ROIC ≥10% for ten years proves a moat existed. You still need to know why it will last [023, 081, 110].
  • App: the Big Five ROIC windows and the tier. The "ROIC falling" flag catches an eroding moat or poor capital allocation [270, 271].

4. Management

  • Talent is capital allocation: reinvest, acquire, pay out or buy back. Buybacks and acquisitions are good only below intrinsic value [005, 100, 226, 462].
  • Debt is the real danger. Keep total debt within 2–3 years of free cash flow (never measure it against EBITDA) [075, 089, 090, 123, 270].
  • Cash must be real. Owner earnings or FCF should be ≥75% of earnings, with appropriate maintenance capex [271, 273].
  • Incentives: pay should reward a better business over 5+ years, not the share price. Read the proxy [126, 226, 276].
  • Candour: score letters for checkable facts vs fog, and compare old plans with results [162, 171, 176].
  • Key-man risk and banks. Every leader leaves, so judge the bench. Banks build leverage in [411, 443].
  • App: the "debt > 3 years of FCF" flag (total debt, including current), the "cash not real" flag (after-tax owner earnings ÷ net income < 75%) and the Tier A debt test.

5. The numbers (Big Five)

  • ROIC ≥10% (ROE is secondary because debt inflates it) [004, 021, 022].
  • The big four growth rates: sales, EPS, equity (BVPS) and operating cash flow, each ≥10%/yr and moving together. Tangled lines mean too hard [019, 020].
  • Operating cash flow is the canary for accounting fiction [020].
  • App: the Big Five pass rate over 10/5/1-year windows and tiers A/B/C.

6. Price: three methods, triangulated (later refinements win)

  1. Ten Cap first. It needs only "the business will be bigger" [327, 328, 467].
    • Owner earnings = operating cash flow − maintenance capex + tax provision [278].
    • Maintenance capex: read the filings, use depreciation, or default to 50% of capex [184, 186, 278].
    • Ten Cap price = 10 × owner earnings − net debt. Interest is already inside owner earnings, but the principal is not [341, 467].
  2. Payback Time: the price that 8 years of free cash flow repays (6 years is almost always good). Use FCF, not EPS [008, 059, 279, 495].
  3. Sticker Price / MOS (the most speculative, now treated as the optimistic case):
    • EPS grown 10 years at the windage growth rate.
    • Future P/E ≤ 2 × growth.
    • Discount at 15%, then halve for the MOS price.
    • Use the lower of historical and analyst growth [009, 093, 122, 280, 327].
  • Triangulate. All three should be in one ballpark. A wide gap needs an explanation [071, 186, 280].
  • Velocity. A stock bought at half price that snaps back in a year leaves only ~7–9% growth afterwards [328].
  • Dividends aren't an input to value. Values veto; they don't discount [044, 066].
  • App (revised):
    • The Ten Cap uses refined owner earnings: depreciation as maintenance capex (capped at capex), otherwise 50% of capex, plus tax, less net debt.
    • Payback uses FCF.
    • "Methods agree" (0–3) shows how many of the three prices the stock is under. It is added to the rank score.

7. Event: no event, no sale

  • An event is a marvelous business with a one-time, solvable problem, plus fear [025, 217, 249, 255].
  • The six event checks [334, 335]:
    1. You know the event.
    2. It was easy to find (a six-inch bar).
    3. It takes ≥1 year to resolve, or managers see through it.
    4. It resolves in ≤~3 years.
    5. The fix needs no new debt.
    6. You have three reasons it's the one company you'd own for life.
  • Not an event: a problem that "takes a miracle", bankruptcy risk, or a changed story [241, 260, 270].
  • Down is not cheap. Debt turns a scare into bankruptcy [255, 260].
  • Cheap at a record high is usually a value trap [240, 324].
  • System-wide events move price, not value [083, 361].
  • App:
    • The "cheap without an event" flag (a buy signal within 10% of the 52-week high).
    • Drawdown and 8-K events on the screen.
    • The Radar agent classifies news as EVENT / PROBLEM / NOISE using these checks.

8. Reduce basis: tranches, rules in advance

  • Buy in tranches as the price falls, in dollars, not shares. Set the triggers before the price moves [199, 266, 352].
  • Keep dry powder, because a wonderful business can fall another 50% after you buy [314, 390].
  • A paper loss is not a loss while value stays far above cost [272].
  • RuleOne excludes options (owner's decision). Episodes that discuss selling puts are noted but not applied [352, 160, 167].
  • App: the Holdings page tracks tranches and flags "below MOS: room for another tranche if the story holds".

9. Story and selling

  • Write the story, the bear case and measurable sell triggers before buying (a pre-mortem) [213, 337–339, 370].
  • Sell when:
    • the story changes (sell at once, even at a loss) [284, 385, 478, 493], or
    • the price is far above value, or
    • the cash can be redeployed much better [158–160, 301, 392–395].
  • Don't sell a rare ~20% compounder just to buy it back cheaper. Phil's recurring error is selling winners too early [467, 492, 493].
  • An event on a stock you own: ask the buying question again. A 1–3 year fix means hold or add. A changed story means sell [463, 493].
  • App: Holdings dangers (above Sticker, falling ROIC, negative 8-Ks, debt), plus the Radar verdict per holding.

10. Portfolio and temperament

  • Own ~5–10 (up to ~20) businesses you understand. Full positions are ~10% [037, 206, 252].
  • Cash is a position. Judge returns over the whole period, idle years included [072, 243, 331].
  • Market gauges (CAPE, Wilshire/GDP) tell you how hard to hunt, not when [039, 247, 248].
  • Volatility is not risk. Permanent loss is [048, 104, 290, 426].
  • Biases (sunk cost, anchoring, confirmation): use a written checklist, a journal and a partner who argues the other side [108, 115, 213, 294, 299].
  • AI is a research assistant, not a source. Verify every figure against the filings [414, 415, 420–422].

Proposed app changes

The Professor adds rule refinements here (from new episodes, the book or the weekly review), and the Engineer implements each one on a branch and marks it "(implemented on branch …)". Implemented so far: refined owner earnings and Ten Cap, methods agree, the four screen flags, and the ten markers in MARKERS.md.

  • (2026-10-09) Flag a dossier in ruleone/rulers.py or the Editor step when it has no per-marker line. Also write markers (pass/fail/unknown) to the universe CSV, so the Professor can check failing markers without recomputing. Not yet implemented.

How the agents share knowledge

The agents never call each other. They share knowledge through files with fixed formats. This page lists who writes what, who reads it, and how InvestED's Rule #1 method reaches every decision. Every agent reads this map.

Analyst consensus (Yahoo) ──► Radar / RULERS / Editor  (a lead and a ceiling, never value)
InvestED podcast (498 eps) ──► Professor ──► knowledge/invested/episodes + course.json
The Intelligent Investor ───►  (notes)   ──► knowledge/intelligent_investor/chapters
                                   │
                                   ▼ synthesis (curate step, book notes)
                     knowledge/rule1/METHOD.md ◄── LESSONS.md ◄── Professor review ◄── scorecard
                     knowledge/rule1/MARKERS.md      ▲                 ▲                 ▲
                     knowledge/rule1/CHECKLIST.md     │                 │                 │
                     knowledge/index/companies.json   │                 │                 │
                                   │                  │                 │                 │
        ┌──────────────┬───────────┼──────────────┐   │                 │                 │
        ▼              ▼           ▼              ▼   │                 │                 │
     Screener        Radar      RULERS          Editor ──► brief + decisions (home page)  │
   (markers,     (events vs   (dossiers:     (reconciles,                                  │
    flags, Ten    dossier      R-U-L-E-R-S,   watchlist.txt ──► RULERS next week)          │
    Cap, rank)    triggers)    markers,             │                                      │
        │              │       episodes)            └──── verdicts + Radar calls ──────────┘
        │              │           │                      (research/scorecard/calls.csv)
        └──► lists ────┴──► Radar history ──► RULERS, Editor, Holdings
   Engineer: implements METHOD.md "Proposed app changes" + MARKERS.md in the screener; keeps the pipes healthy
File Written by Read by Purpose
knowledge/invested/episodes/NNN.md, course.json Professor (podcast) Professor, RULERS (via the index), Learn page Study notes per episode; course modules
knowledge/rule1/METHOD.md Professor (curate, book, review) all agents The consolidated Rule #1 method, with episode citations
knowledge/rule1/MARKERS.md Professor proposes, Engineer implements Screener, RULERS, Radar, Professor review Ten machine-checkable markers of a wonderful business
knowledge/rule1/LESSONS.md Professor review all agents Process lessons from past calls and from the show's own mistakes
knowledge/rule1/CHECKLIST.md Professor RULERS, site The checklist applied per stock
knowledge/index/companies.json ruleone.knowledge (weekly, and after new episodes) RULERS, Radar, Professor review, stock pages Which episodes discuss which company
lists/latest/* Screener (Saturday) Radar, RULERS, Editor, Engineer, site Prices, Rule #1 prices, Big Five, markers, flags
lists/latest/analysts.csv, research/analysts/* ruleone.analysts (weekdays with Radar; a broad pass on Saturday) Radar (downgrades and target cuts), RULERS (the Street line, growth ceiling), Editor (street vs Rule #1), stock pages, /street/ Sell-side consensus: low/mean/high targets, ratings, revisions, and agree / contrarian / crowded against the Rule #1 prices
research/radar/history.json Radar (weekdays) RULERS (scope and Event), Editor, scorecard, Holdings EVENT / PROBLEM / WATCH / NOISE per ticker
research/rulers/<T>.md, index.json RULERS (Saturday) Radar (triggers), Professor review, Editor, scorecard, Holdings Living dossiers: verdict, ladder, story, sell triggers
research/reviews/latest.json Professor review (Saturday) RULERS (fix majors), Editor (conflicts) Marks each dossier against the method
research/scorecard/* ruleone.scorecard Professor review, Editor Decision journal; calls graded against the S&P 500
research/watchlist.txt Editor (and you) RULERS Names to (re)analyse first next week
research/editor/decisions.json Editor Home page This week's actions and your decisions
ops/health/*, ops/incidents/* Engineer Engineer, Ops page Pipeline health and fixes

Weekly order (Saturday): screen → RULERS → Professor review → Editor → deploy. Radar runs on weekdays, the Professor (podcast) daily, and the Engineer on Sundays and whenever a run fails.

How a lesson travels: a verdict goes into the scorecard. If it goes badly, the Professor review asks whether the method would have warned us. A process lesson then goes into LESSONS.md, and if it needs code, a proposal goes into METHOD.md → Proposed app changes, which the Engineer implements as a screen change or a new marker. From then on Radar and RULERS apply it.