Reason 2 is the twin of reason 1, and the distinction is temporal and about you. In a broken thesis (reason 1), your analysis was reasonable at the time and the world subsequently changed. In a mistake (reason 2), the thesis was wrong when you formed it. The world did not change; your understanding of it was flawed from the start, and you have only now recognized it. You misjudged the moat (it never really existed); you mis-modeled the unit economics; you missed a red flag that was sitting in the filings (the ALM mismatch in the NBFC, M5.02; the aggressive revenue recognition, M2.06); you trusted a management that the proxy and related-party notes should have warned you about; you overpaid on a valuation built on an assumption that was never plausible. The recognition often arrives as new information that reveals what was always true: a restatement, a disclosure, a competitor filing that shows the market was never what you thought.
Advertisement
What is the test for cutting a loss?
The decision test: given what I now know, was my original analysis wrong, not merely unlucky? If yes, take the loss now. This is the Assassin move, and it is, per Freeman-Shor, the single highest-value habit in the whole discipline: cut mistakes fast. The instinct to do the opposite is overwhelming, because selling crystallizes the loss (loss aversion) and admits the error (ego, Munger's commitment-and-consistency tendency), and because the mind offers a seductive alternative: average down to "lower your average cost" and "make it back faster." This is the cardinal sin of reason 2. Averaging down into a broken or mistaken thesis is not the Hunter buying a gift; it is throwing good money after bad, doubling your bet on a premise you now know is false, and converting a small, survivable mistake into a large, portfolio-threatening one. The whole point of the mistake taxonomy is that a mistake is a signal to reduce exposure to your own error, not increase it. Add only to positions whose thesis is intact and strengthening; never to a position you have just recognized as a mistake.
How do you tell a mistake from bad luck?
Mistake versus bad luck: the process-over-outcome filter. The hardest and most important judgment inside reason 2 is separating a genuine mistake (flawed reasoning, which you can learn from and should cut) from genuine bad luck (sound reasoning and a bad outcome from irreducible uncertainty). This is M6.03's process-over-outcome discipline applied at the exit. A mistake means that, with the information reasonably available at the time, a good process would have reached a different conclusion. You missed something knowable. Bad luck means that a good process, honestly run, would still have bought, and an genuinely unforeseeable event (a true black swan, a fraud with no detectable footprints, an exogenous shock no diligence could have priced) caused the loss. The distinction matters twice over: it governs whether you cut (you cut a recognized mistake regardless of price; a bad-luck loss with an intact thesis you may hold or even add to, because the reason to own survives), and it governs what you learn (see the mistake taxonomy: a mistake generates a checklist line; genuine bad luck does not, and pretending it does is superstition). The danger runs both ways: over-attributing losses to bad luck protects the ego but kills learning ("I was just unlucky" when you actually ignored a red flag); over-attributing to skill error breeds superstition and overfitting, adding useless checklist lines that fight the last war. Most losses that feel like bad luck are, on honest inspection, an analytical or behavioral miss in disguise; reserve the "bad luck" verdict for the genuinely undetectable, and demand evidence for it.
Worked example: a lender that missed a repayment
Worked example (India/NBFC archetype, illustrative, the M5.02 failure mode). You bought a fast-growing non-bank lender two years ago on a thesis of "25% AUM growth at a 3% ROA with a widening spread — a compounder in retail credit." The stock has fallen 60% and now the company has missed a commercial-paper repayment. Reason 1 or reason 2? Go to the journal, and be honest about what was knowable at purchase. If the asset-liability mismatch (long-tenor loans funded with short-tenor commercial paper, a negative short-bucket ALM gap) was disclosed in the filings you could have read (it was; it is in the ALM statement, M5.02), then this was not bad luck and not a world-changed thesis break: it was a mistake, an analytical miss, a red flag you failed to check in a business whose entire risk is its funding structure. The recognition is that you were wrong at purchase about the quality of the balance sheet. The decision: sell, take the loss, and do not average down on the seductive logic that "it's fallen so far it must be cheap." A lender facing a funding run can go to zero (IL&FS/DHFL, M5.02), and adding is the Rabbit funding a mistake. The checklist line this generates (the mistake taxonomy, via M6.05): "For any lender, read the ALM gap and funding mix before the growth story — never own a financial whose funding I have not stress-tested."