The Method Lab

This is an experiment in finding a repeatable formula for picking stocks: what a business is worth, where to get in and where to get out. AI is part of the process. The goal is results that are repeatable, logic-based and verifiable. The same inputs should give the same call, and every call gets scored on the track record.

The system isn't perfect yet. When I change the valuation or entry formula, I write up what changed and why below, and every past call stays on the record exactly as it was made. That way you can see whether each change actually helped. It may turn into how I invest one day. For now I'm sharing the progress as it improves.

The checks, in order

1. Red flags. I read the filings first, looking for dilution, accounting stretches, insider selling, customer concentration and anything the headlines leave out. The result is Pass, Caution or Fail.
2. Value. Every valuation has three cases, each with a low, base and high. Conservative is a floor I keep for reference and never use for the call. Market-implied puts a growth-tiered multiple on forward owner earnings. Growth-adjusted is a three-stage DCF. The discount rate is the 10-year Treasury yield on the month's first trading day plus 5.75 points, kept between 9% and 12%. The call rests on a 50/50 blend of those two.
3. Timing. I set the entry from the chart: a recent swing low, the point of control on the volume profile (the price where the most shares traded), or the 50% Fibonacci retracement of the last leg up. The exit is the daily close that proves the setup wrong.
4. Size. A full position risks 1% of the account, measured from entry to stop. When earnings fall inside the trade, I size for a gap 1.5 times the stop distance.

The ratings

Every call uses one of five ratings. The rating legend has the full definitions and the scale history.

• Buy (full). The price is at or below my buy-below, which is the blended base value less a 15%, 20% or 30% margin of safety. The less predictable the business, the bigger the margin. Full risk.
• Buy (half). The price is at or below the blended base value, the red-flag review is Pass or Caution, and the chart gives me an entry and a stop. Half risk, which is $500 on a $100,000 account. Most of my buys happen here. Solid companies rarely trade at a deep discount, so I time the entry and keep the stop close.
• Wait. Not a buy at the current price, because the price already assumes too much or the setup isn't there yet. The call says what price or event would change it.
• Steer clear. The filings or the red-flag review fail. I wouldn't own it at the current price.
• Short. The price is well above the high end of all three valuation cases, the red-flag review fails, and I have a fade trigger, a cover stop and the borrow and squeeze data (borrow fee, short interest, days to cover). Shorts get half risk because a short's loss has no ceiling.

A name my short screen turns up is a screen result until it meets all three Short tests. Screen results aren't calls and aren't scored.

The levels on each card

Every level on a card marks where I act or where I'm wrong. Entry is where I buy. Out below is the daily close that ends a Buy. On a Short, the fade trigger is the price action that opens it and the cover stop is the close that ends it. The failed-breakout level is the daily close that says a breakout attempt didn't hold. None of them is a price target, and I don't publish targets. Levels expire after the next earnings report or a move bigger than two times the average true range (ATR).

Worked example: NVIDIA

On Oct 7, 2026, at the $237.47 close, my call was Wait with a buy zone of $185-190. That came from a single value of about $263, which left a 9.6% margin of safety against the 30% I wanted.

On Oct 11 I revised it to Buy (half) at $224-228, out on a daily close below $217, using Oct 9 data (NVDA Deep Dive). The three cases put the blended base value at about $245, with a low of about $127 and a high of about $317. The Oct 9 close of $229.28 sat about 6% under it.

NVIDIA earned a 75.0% gross margin in Q2 FY27, and my ROIC calculation for FY26 is about 134%. Its earnings are cyclical, though. In FY23 revenue was flat and operating margin fell from 37% to 16%. In Q2 FY27 operating cash flow was only 40% of net income. The price is under the blend but well above the buy-below of about $171, and the red-flag review is Caution, so the call is Buy (half). Full size waits for about $171, 30% under the blend.

The entry zone is the 50% retracement of the Sep 14 to Oct 6 leg ($226.15) and the top of the six-month value area on the volume profile ($228.24). The stop sits about one ATR under the 50-day average. These levels expire after the Q3 FY27 report, expected Nov 17.

Every call, including the withdrawn and superseded ones, is on the scorecard with its result against the benchmark.

Method changes

Every report and every buy-list row names the method it used and links to its entry here. I don't edit published reports, so an older report can show a different value than today's list.

Oct 30, 2026: one formula for every name (v1)

Link: /insights/method-lab/#method-2026-10-30

Why it was built. I built v1 because my calls ran too pessimistic. I like shorting overvalued stocks, and that bias pushed my fair values down. Before v1 I graded each name's uncertainty by judgment, and the grade set the margin. v1 runs every name through one formula with fixed settings. The same inputs always give the same value. Exceptions are set by category in advance, and each value names the exception it used.

Inputs.

• OE0: trailing owner earnings per diluted share, after stock comp, with D&A standing in for maintenance capex. MSFT, GOOGL, META, BRO and FLEX were re-based on post-Q3 filings on Oct 30.
• Consensus revenue and EPS growth for this and next fiscal year and the revenue range, refreshed Oct 30. The weight w (months left in the fiscal year / 12) comes from SEC quarter-end dates.
• Caps on EV/(FCF − stock comp): the name's own 5-year median and its peer median.
• Up to 10 fiscal years of SEC financial history for the category tests.
• The red-flag lines below.
• Discount rate r: October's 11.00% (10-year Treasury 5.24% on Oct 1 + 5.75 = 10.99%, rounded to the nearest 0.25 point, held within 9% to 12%). Every report also shows the value at r − 1 and r + 1 point.

Formula. For each case i (low, base, high with s = -1, 0, +1):

1. Market-implied_i = net cash + OEn x (1 + gN + s x spread) x lev / (1 + d) x min(tier multiple_i, cap). Tiers on base gN: under 8% 14/17/20x, 8% to 15% 20/24/28x, over 15% 25/30/35x, one tier per name. spread = the larger of 2 points and half the consensus range. lev = (1 + EPS growth) / (1 + gN), held within 0.97 to 1.05. OEn = OE0 and d = 0 unless an exception changes them.
2. Growth-adjusted_i = non-operating value + a three-stage DCF on OEn. Year 1 grows (1 + gN) x lev − 1, year 2 (1 + next-year growth) x lev − 1, years 3 to 5 at next-year growth capped at 25%. Years 6 to 10 fade in a straight line to 3%, with a 3% terminal. Low = 0.6 x the base path, without lev. High = 1.2 x base, capped at 30% a year and never below base.
3. Blend_i = (Market-implied_i + Growth-adjusted_i) / 2. Fair value = blend base.
4. Margin from grade points, one each for: cyclical or turnaround, negative FCF in 2 or more of the last 5 years, under 5 years of history, net debt above 3x OE, and a blend spread (high − low) / base above 0.75. 0 points = 15%, 1 = 20%, 2 or more = 30%. No consensus = 30%. Add 5 points once for dilution or overhang and 10 points in a Red macro regime, to a maximum of 50%.
5. Buy-below = fair value x (1 − margin).

The Conservative floor carries over unchanged and is never the call.

Exceptions by category. Each is a fixed test on the name's data.

• Cyclical: in the last 10 years (at least 5), a one-year operating income fall of 40% or more from a positive base, or a negative year. OEn = OE0 x F, where F = 10-year average operating margin / latest margin, held within 0.5 to 2.0. Market-implied drops one tier. The low case takes -15% in year 2. Plus 1 grade point.
• Turnaround: 7 or more years of margins and a latest margin at or below zero, or at or below half the cycle average. Same adjustment as a cyclical.
• No own 5-year multiple, or negative FCF in 2 of the last 5 years: the cap comes from peers alone. Negative FCF adds 1 grade point.
• Heavy dilution: diluted shares up 3% or more a year. Growth for the next 12 months and years 1 to 5 is divided by (1 + d), plus the 5-point step.
• Overhang: a government holding, or one holder distributing 5% or more of the shares within 12 months. Plus the 5-point step.
• Short history: under 5 fiscal years. Plus 1 grade point.
• No consensus: fair value = Conservative base, grade High.
• Spin in the current fiscal year: next-year consensus only.
• Compounder: not cyclical, 7 or more years of history, 5-year incremental ROIC at least r + 5 points, reinvestment rate at least 10% and revenue growth at least 5% a year. Growth-adjusted takes the higher of the consensus path and an earning-power value (ROIC capped at 40%, starting growth = reinvestment rate x ROIC capped at 15%, fading to 3% by year 10).
• Unfiled quarter, known only from an 8-K: levels are the lower of this run and the run on the last filed numbers.

Red flags. Each name gets eleven fixed lines, each Pass, Caution or Fail and each tied to a filing, its accession number and its date:

1. Dilution and offering capacity: shares outstanding from 10-K and 10-Q cover pages, S-3 and ATM shelves, equity and convertible issuance, stock comp.
2. Share overhang: forward sales, contingent issuance, holder distributions, lock-up expiries.
3. Insider transactions: Form 4 and Form 144, last 6 months.
4. Related-party transactions.
5. Founder or control: voting structure and any controlling holder (proxy, Schedule 13D/G).
6. Short interest and borrow (FINRA).
7. Debt and liquidity.
8. Cash flow against net income, and free cash flow.
9. Controls, auditor and going concern.
10. Legal, regulatory and government.
11. Off-balance-sheet commitments, impairments and restructuring.

The overall verdict is also Pass, Caution or Fail. 12 names still carry a single overall verdict from Oct 20.

Buy rule. A name at or under its buy-below is a buy unless its red-flag check is a Fail. A Caution is a warning only. A Caution on the dilution or overhang line also adds the 5-point margin step. Buy (half) at or below fair value also needs a set entry and stop.

Levels.

• Swing low: the lowest daily low of the 5 sessions on each side, searched over 1 year. A low without 5 later sessions is labeled unconfirmed with the date it would confirm.
• Volume profile: 5-minute regular-session bars over the last 60 sessions, 100 price bins, each bar's volume spread across its range. Spike cleaning was dropped. A post-earnings profile and an earnings-day profile (40 bins each) are drawn as well.
• 50% retracement: the midpoint of a leg between two named pivots, both stored with their dates.
• Leg ends, zone width, stop placement and the breakout level are set by hand per name and cited on the card.
• Out below: a daily close under the stop.
• Confirmation: an up close on volume at least 1.2x the 50-day simple average, through the last close.
• Breakout: a daily close above the breakout level on confirmation volume. Failed breakout: a daily close back below it.
• Expiry band: from the stop minus 2 ATR to the failed-breakout level plus 2 ATR (ATR is the 14-day simple average true range). Levels expire at the next earnings report or on any daily close outside the band.
• Redraw after an earnings report or a move of more than 2 ATR. The old levels stay on record, marked superseded. Levels with no expiry date or band count as stale.

Worked example: MSFT valuation, Oct 29 close $530.14.

1. OE0 = TTM operating income $163,191M x 0.82 / 7,446M shares = $17.97. Net cash $3.90 a share.
2. gN = 0.75 x 19.21% + 0.25 x 19.98% = 19.40%. lev 0.980. spread 2.85 points.
3. Exceptions: compounder only (5-year incremental ROIC 37.1%, reinvestment rate 40.7%, revenue growth 14.6% a year).
4. Tier 25/30/35x. Cap 51.63x, not binding.
5. Next-12-month OE $20.53 / $21.03 / $21.54. Market-implied $517.16 / $634.88 / $757.61.
6. Growth-adjusted at 11%: $378.46 / $551.24 / $671.30. The earning-power value ($261.00 / $315.05 / $346.26) is lower in every case, so the consensus path stands.
7. Blend $447.81 / $593.06 / $714.45. Fair value $593.06.
8. Grade points 0 (spread 0.45), margin 15%. Buy-below = 593.06 x 0.85 = $504.10, 4.9% under the price. Status: Buy (half) zone.
9. Red flags: the run read a single overall Caution from the Oct 20 check. It has no dilution or overhang line, so no margin step.

Worked example: INTU levels, Oct 29 close 331.19. Entry 310-319 on the 50% retracement (310.18) of the Sep 28 swing low 262.32 to the Oct 22 high 358.03. Nearby: the Sep 10 swing low 309.80, the 20-day average 317.76, the 50-day average 319.79 and the value-area low 306.58 (59 sessions, Aug 7 to Oct 29). The rule is 60 sessions, but only 59 were available. Out below a daily close under 298, under the 61.8% retracement 298.88 and the Oct 14 low 298.83. Confirmation: an up close on at least 4.43M shares (1.2 x the 3.69M average). Breakout: a close above 358.03 on at least 4.43M. Failed breakout: a close back below 358.03. ATR 13.55, so the band runs 270.90 to 385.13. Expiry at the Nov 19 earnings report (estimated) or a close outside the band.

Effect on the list. At Oct 29 closes, with consensus refreshed Oct 30, across 27 names, the median buy-below moved from 45.0% under the price to 37.0%. Names whose buy-below needs a drop of more than 35% fell from 17 to 15.

Testing. v1 has no back-test on history. A same-day re-run reproduces all 27 published fair values and buy-belows to the cent. Terminal value is 44.5% to 58.7% of the DCF leg across the 26 names that have one. Three names move more than 25% when growth changes by one point: AZO, BRO and HONA.

Limits. Growth for years 1 and 2 comes from one free consensus source and is not refreshed after every report. Peer sets are fixed per name with no written selection rule. The thresholds (40%, 3% a year, 3x, 0.75, 5 points) are set, not calibrated.

Used in. The META and INTC Deep Dives and the buy list from Oct 30.

Oct 11, 2026: three cases for every name (pre-v1)

Link: /insights/method-lab/#method-2026-10-11

Why it was built. Before Oct 11 each name had one value: half a 10-year DCF at 10% and half 14x, 17x or 20x trailing owner earnings, plus net cash. Buy-belows set 20% to 30% under that value needed a recession to reach.

Inputs.

• Owner earnings (OE): trailing, per fully diluted share, after stock comp. I set the definition per name: free cash flow less stock comp, or GAAP operating income after tax.
• Consensus revenue and EPS growth for this and next fiscal year, and the consensus revenue range.
• EV/FCF: the name's own 5-year median and the median of a fixed peer set.
• Net cash or net debt, excluding leases, per share.
• Discount rate r: 10% in the Oct 11 runs. The monthly rule (10-year Treasury yield on the month's first trading day plus 5.75 points, rounded to the nearest 0.25 point, held within 9% to 12%) was approved that evening and applied from Oct 18 at 11.00%.

Formulas.

• Next-12-month growth gN = w x this-year growth + (1 − w) x next-year growth, where w = months from the last reported quarter end to fiscal year end / 12.
• Operating leverage lev = (1 + EPS growth) / (1 + gN), held within 0.97 to 1.05.
• Market-implied = net cash + OE x (1 + gN) x lev x multiple. The multiple comes from a tier on gN: under 8% 14/17/20x, 8% to 15% 20/24/28x, over 15% 25/30/35x (low/base/high). Each is capped at the lower of the own 5-year median and the peer median EV/FCF. Low and high growth are gN minus and plus half the consensus range, at least 2 points.
• Growth-adjusted = non-operating value + a three-stage DCF on OE. Years 1 and 2 use consensus growth times lev, years 3 to 5 next-year growth capped at 25%, and years 6 to 10 fade in a straight line to 3%, with a 3% terminal. Low is 60% of the base path without lev, and cyclicals take -15% in year 2. High is 120% of base, capped at 30% a year.
• Conservative, a floor I keep for reference only: half a 10-year DCF and half 14x to 20x trailing OE.
• Fair value = (Market-implied base + Growth-adjusted base) / 2.
• Buy-below = fair value x (1 − margin). The margin is 15%, 20% or 30% for a Low, Medium or High uncertainty grade set per name, and 25%, 30% or 40% in a Red macro regime.
• Cyclicals are never valued on Market-implied alone. A name with no consensus takes the Conservative base as fair value.

Red flags. Each name got one overall verdict, Pass, Caution or Fail, read from SEC EDGAR filings (8-K, Form 4, 10-Q, S-3 and 424B, Schedule 13D/G, NT filings, 8-K items 4.01 and 4.02) and FINRA short interest. The check covered dilution, cash flow against net income, debt, off-balance-sheet commitments, legal matters, insider sales and short interest. Each check states its sources and the date the data was pulled.

Ratings. Buy (full), Buy (half), Wait and Steer clear. Buy (full) is at or below the buy-below with a red-flag Pass. Buy (half) is at or below fair value with a Pass or Caution and a set entry and stop.

Levels.

• Swing low: a daily low that is the lowest of the 5 sessions on each side, searched over the last 126 sessions.
• Volume profile: 126 daily bars in 80 price bins, each bar's volume spread evenly across its range. A day above 6x its 21-day centered median volume is cut to that median. The point of control is the busiest bin. The value area grows from it until it holds 70% of volume.
• 50% retracement: the midpoint of the leg from a recent swing low to the highest high since.
• I placed entry zones and stops by hand around these levels.
• Out below: a daily close under the stop. An intraday trade under it does not count.
• Confirmation: a bounce day on volume at least 1.0x the 50-day average.
• No failed-breakout level was set.
• Expiry: the next earnings report, or a move of more than 2 ATR from the close on the day the levels were set (ATR is the 14-day simple average true range). A name with no valid swing low, point of control or 50% level under value has its levels withdrawn with a dated note.

Worked example: MSFT valuation, Oct 9 close $535.07.

1. OE $17.14 = FY26 operating income $155.2B x 0.82, per diluted share.
2. gN 17.91%, next-year growth 19.60%, lev 1.1440 / 1.1791 = 0.970, half-range 3.45 points.
3. Tier 25/30/35x. Cap = lower of own 42.15x and peer 46.47x = 42.15x, not binding.
4. Next-12-month OE = 17.14 x 1.1791 x 0.970 = $19.61. Market-implied base = $4.99 net cash + 19.61 x 30 = $593.24.
5. Growth-adjusted base at 10% = $583.61.
6. Fair value = (593.24 + 583.61) / 2 = $588.42.
7. Grade Low, margin 15%. Buy-below = 588.42 x 0.85 = $500.16, 6.5% under the price.
8. At 11% (Oct 18 re-run): fair value $547.61, buy-below $465.47.

Worked example: INTU levels, Oct 9 close 302.75. Entry 284-290, from the 50% retracement of the Sep 28 swing low 262.32 to the leg high 309.76 (286.04) and the Oct 6 close 289.81. Out below a daily close under 276, about 0.75 ATR (10.91) under the 284 shelf. Confirmation on volume of at least 3.80M shares (1.0x the 50-day average). Expiry at the Nov 19 earnings report or a 2 ATR move. A +2.05 ATR move redrew the levels on Oct 19.

Why it was replaced. I decided too much name by name. The OE definition, the peer set and the uncertainty grade had no written rule. The cap compared EV/FCF before stock comp with OE after it. For very fast growers the high case could fall below base. The levels also stayed out of reach. At Oct 29 closes the median buy-below sat 45.0% under the price, and 17 of 27 names needed a drop of more than 35%. v1 replaced it on Oct 30.

Used in. The SPOT Deep Dive and the buy list until Oct 30.