Outcomes are made of judgement and luck in unknown proportions. Judging the decision by the result means learning from the luck, which teaches the wrong thing in both directions.
13 min read · Level 4
The separation of process from outcome is usually presented as a consolation - a way to feel acceptable about losing - and that framing is both wrong and slightly corrosive, because it makes the idea sound like something you reach for when things have gone badly. Its real function is the opposite. It is far more useful applied to wins, and it is almost never applied there, because a win generates no discomfort and therefore no demand for explanation. The uncomfortable implication is that an investor's most dangerous holdings are frequently the ones that worked, since those are the ones whose reasoning was never examined and whose method will be repeated at larger size. There is a limit worth being honest about. Process cannot be assessed in a vacuum, and an investor who explains every loss as variance and every win as judgement has built a position that no evidence can touch - which is not rigour, it is the same bias with better vocabulary. What keeps it honest is aggregation and writing things down in advance: enough decisions, with stated likelihoods, to see whether the reasoning ever resembles reality. That is a slow instrument. It is also the only one an individual investor has, and refusing to use it because it is slow leaves nothing but the outcome, which is the thing that was never about you.
Every result is produced by two things: the quality of the reasoning, and what the world happened to do. Neither is observable on its own. All you ever see is the sum.
Outcome bias is the habit of reading the sum as though it were the first term. It produces four categories, and only two of them are ever discussed.
Right and won - the reasoning was sound and the world cooperated. Wrong and lost - the reasoning was poor and the world did not rescue it. Right and lost - the reasoning was sound and the unlikely branch happened anyway. Wrong and won - the reasoning was poor and it worked out regardless.
The first two are the ones people describe. The second two are where nearly all the available learning is, and both are systematically mislabelled: the third gets filed as a mistake and something correct is unlearned, while the fourth gets filed as skill and something incorrect is reinforced.
The fourth is the more dangerous by a wide margin, because nothing about it feels like a problem. A loss at least prompts an examination. A win that should not have happened prompts a repetition.
Because an investor who judges decisions by results is not learning from experience. They are learning from variance, and variance has nothing to teach.
This is not an abstraction. The measurable consequence is a portfolio that drifts toward whatever recently worked and away from whatever recently did not, regardless of whether either was correct - which is a description of buying high and selling low arrived at through entirely reasonable-sounding steps.
And because you already have the tool. A written entry records the reasoning separately from the result, which is the only way the two can ever be judged apart. Without one, the distinction in this lesson is philosophically true and practically unusable, because there is no record of the reasoning to examine.
Educational, not personalised financial advice. Nothing in the Investor Academy is a recommendation to buy, sell or hold anything — what is right for you depends on your own circumstances and the rules where you live.
Two decisions, each committing 10,000. The probabilities are stated because this is a constructed illustration; in real investing they are estimates, and that is a difficulty rather than an objection.
Decision A: 70% chance of gaining 50%, 30% chance of total loss. expected value = 0.70 x 5,000 + 0.30 x (-10,000) = 3,500 - 3,000 = +500
Decision B: 5% chance of gaining 400%, 95% chance of total loss. expected value = 0.05 x 40,000 + 0.95 x (-10,000) = 2,000 - 9,500 = -7,500
A is a sound decision. B is a poor one, by 8,000 of expected value.
And yet: three times out of ten, A loses everything. One time in twenty, B returns 40,000 and the person who made it is congratulated at length.
This is not a rare edge case. Take ten independent decisions all shaped like A - every one of them sound:
chance that none of the ten loses everything: 2.82% chance that at least three of the ten lose everything: 61.72%
So an investor making consistently good decisions of this shape is more likely than not to suffer at least three total losses out of ten, and is almost certain to suffer at least one. If each of those is filed as a mistake and the reasoning revised accordingly, the revisions will be made in response to nothing at all - and the most likely revision is to stop taking A-shaped decisions, which were the good ones.
Now the part that stops this from being a licence. If process were unexaminable and only outcomes existed, 'judge the process' would be untestable comfort. It is testable, but only in aggregate and only slowly: across enough decisions, a genuinely sound process produces results that a poor one does not, and the entries themselves are what let you check whether your stated probabilities resembled reality. One loss says nothing. Thirty decisions whose stated likelihoods never resembled the outcomes say a great deal, and they say it about the process rather than about luck.
The specific wrong move is changing an approach because a single decision lost.
It is the most reasonable-sounding error available. Something went wrong, you examine it, you find the thing you would do differently, you resolve to do it differently. Every step is what a thoughtful person does - and if the decision was sound and the unlikely branch simply happened, the change is a revision made in response to noise, applied permanently, to a process that was working.
It is tempting because losses demand explanation and 'the unlikely thing happened' is an unsatisfying one. It sounds like an excuse even when it is the literal truth, and it is socially unusable: nobody explains a loss that way out loud, so nobody hears it said, so it stops being an available category.
The mirror error is worse and much quieter. A decision that was poorly reasoned and worked anyway is filed as evidence of skill, the method is repeated with more money behind it, and nothing signals a problem until the arithmetic finally asserts itself. Survivorship bias makes this near-universal at the level of public examples: the people describing their methods are disproportionately the ones whose gambles landed, and their accounts are sincere.
At review, ask whether the reasoning was sound given what was knowable at the time - not whether the position is up. These are different questions with frequently different answers, and only the first one is about you.
Right and won, wrong and lost, right and lost, wrong and won. Forcing the four-way classification is what makes the two invisible categories visible, and most people find they have never knowingly placed anything in the fourth.
If the outcome had gone the other way, would the reasoning have looked sound? A win that only looks like judgement because it won is the counterfactual test failing, and the test costs one sentence.
One outcome is not evidence about process. Before revising an approach, find the same failure in at least three entries, in the reasoning rather than in the result. Three is not a statistical threshold; it is a barrier against revising in response to noise.
This is the one genuine test of a process available to an individual, and it needs the numbers to have been written down in advance. Across enough decisions, stated probabilities that never resemble the outcomes are a finding about the reasoning that no single result could give you.
At every review of a closed decision, and particularly after an unusually good result - which is the moment this is least likely to be applied and most likely to matter.
Do not use it to avoid examining a genuine error. 'The process was fine, I was unlucky' is sometimes true and is also the most comfortable sentence in this subject, which is exactly why it needs the four-way classification and the entry to support it rather than an assertion.
Public accounts of investing success are dominated by people whose concentrated positions happened to land, and their accounts are sincere - they genuinely did reason their way to the position, and it genuinely did work. What the account cannot contain is the set of people who reasoned identically, sized identically, and were wiped out, because those people do not publish and are not asked. The method therefore propagates with an apparent success rate assembled entirely from survivors, and the sincerity of each individual telling is part of what makes it persuasive.
When a method is supported mainly by examples of it working, ask what the same method looks like when the central judgement is wrong. If nobody can describe that case, you are looking at survivors rather than at evidence.
Investors who keep entries and classify outcomes generally populate three of the four categories readily and leave 'wrong and won' almost empty. It is not usually dishonesty. A profitable outcome supplies a reason retroactively, and the reason is often genuinely plausible - the position did rise, and there is always an account of why. Without a contemporaneous entry there is nothing to contradict the account, and with one the contradiction is uncomfortable in a way that a loss somehow is not.
If no decision has ever been placed in 'wrong and won', treat that as a finding about the classification rather than about the decisions. Start with the biggest win and ask what the entry would have looked like had the outcome gone the other way.
Create a free account to read the rest and apply it to your own investing.