The entry is only half the practice. The other half is going back to it later and reading what you actually wrote, rather than what you now wish you had.
11 min read · Level 4
There is a quiet assumption underneath this practice that deserves to be stated rather than smuggled: that an investor can study their own judgement the way they would study a company. Mostly they cannot, and the reason is sample size. A person makes perhaps a handful of consequential investment decisions a year, the outcomes are dominated by variance, and the feedback arrives years later attached to a market that has changed in the meantime. Twenty entries over three years is a realistic pace and it is not enough to establish much of anything about skill. That sounds like an argument against the practice and it is not, because the useful findings are not statistical. They are specific and repeated: four entries that all assumed things would happen faster than they did, three that all skipped the price paid, two that both relied on somebody else having checked something. Those are visible at a handful of observations because they are patterns in reasoning rather than in outcomes, and reasoning is the thing the journal actually records. So the honest claim is narrower than the usual one. A decision journal will not tell you whether you are good at this. It will tell you what you reliably forget to consider, which is more useful and considerably easier to act on.
A review is a structured reading of an old entry, done in a fixed order, with one rule: the entry is never edited.
The order matters more than anything else here, and it is counterintuitive.
Read the entry first, before looking at what happened. Reconstruct the state of mind it describes. Only then look at the outcome. Then, separately, write a new note below the old one saying what was accurate, what was not, and what you now know.
The old text stays exactly as written. Corrections are appended, never applied. An entry that can be revised is a document that will agree with you forever, which is the one thing a record must not do.
Over time, a set of entries supports something a single entry cannot: calibration. If you wrote down how likely you thought each outcome was, you can eventually compare the likelihoods you assigned with how often those things actually happened. That comparison is available to almost nobody, because almost nobody wrote the numbers down.
Because the entry alone does not protect you. It only makes the protection possible.
An unread journal is a filing cabinet. The value arrives at the moment you set your memory of what you thought beside the document recording what you thought, and notice the gap. That gap is the only direct evidence you will ever get about how your own judgement actually behaves.
And because the order of reading is what prevents the review from being worthless. Look at the outcome first and the narrative fallacy takes over immediately - the entry is then read as a story whose ending you already know, and every ambiguous sentence resolves toward the ending. Read the entry cold and you get the genuine experience of not knowing, which is the state you are trying to study.
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.
A review of twenty entries written over three years. Each one recorded a likelihood band alongside the claim, which is the part that makes this possible at all.
Marked around 90% likely: 6 entries, 5 happened = 83.33% Marked around 70% likely: 9 entries, 5 happened = 55.56% Marked around 50% likely: 5 entries, 3 happened = 60.00%
Total: 20 entries, 13 happened = 65.00%
What this does and does not say.
The 90% band came out at 83.33%, which is close, and rests on six entries - far too few to conclude anything at all. One different outcome would move it to 100.00% or 66.67%. Six is not a sample; it is an anecdote with a percentage sign.
The 70% band is the interesting row: nine entries, 55.56%. That is a meaningful gap in the direction almost everybody's is - things marked as likely happening less often than stated. Nine is still a small number, and this is a signal worth watching rather than a finding.
The 50% band came out at 60.00% on five entries, which says nothing whatsoever. At that band you are declaring a coin, and five coins land three-two more than half the time.
The honest summary of twenty entries over three years is therefore: one row worth watching, two rows too thin to read, and a total of 65.00% that mixes three different claims together and means nothing on its own. That is a genuinely useful result, and it is useful precisely because it refuses to conclude more than twenty observations can support.
Which points at the real finding: three years of investing produced twenty checkable statements. Most of the value of this review was in discovering how few decisions had been recorded in a form that could be examined at all.
The specific wrong move is opening the outcome first and then reading the entry.
It feels efficient - you want to know whether you were right - and it destroys the review. Knowing the ending makes every ambiguous sentence in the entry resolve toward it. A line that was genuinely balanced at the time now reads as prescient or as obviously mistaken, depending on what happened, and you will not be able to tell that this has occurred.
It is tempting because the outcome is the emotionally interesting part and the entry is not. The discipline of reading four dry lines before allowing yourself the answer is small and real, and it is the whole method.
The second mistake is editing. Softening a claim, adding a caveat you did not make, or quietly deleting the sentence that turned out to be embarrassing. It happens semi-consciously, in the spirit of making the record more accurate, and it converts the one document that could disagree with you into another one that does not.
The third is reading a single entry and drawing a conclusion about yourself from it. One decision contains almost no information about judgement, because the variance in outcomes is enormous relative to any difference in skill. Patterns live across entries, not inside them.
A date chosen in advance, matched to the timeframe of the claim - a three-year thesis gets read at one year and three years. Reviews that happen when something goes wrong are a biased sample of your own decisions, which produces a biased conclusion about your own judgement.
Close the price, close the statement. Read what you wrote and try to genuinely occupy the position of not knowing. Write one sentence describing the state of mind it records, and do it before you look.
Claim by claim, not entry by entry. An entry with four claims is usually not simply right or wrong; it is commonly two accurate claims, one wrong, and one that turned out to be irrelevant - and which of the four is which is the whole content of the review.
Date the note, say what you now know, and leave every original sentence standing, including the embarrassing one. The embarrassing sentence is the highest-value object in the entire journal, and its value is destroyed by exactly one keystroke.
One loss teaches almost nothing. Four entries that all overestimated how quickly something would happen teach you a specific, nameable bias in your own thinking, and that is the kind of finding worth changing behaviour over. Check how many entries a pattern rests on before believing it.
On a schedule you set in advance, at the review dates written into the entries themselves, and once a year across the whole set to look for repeated patterns.
Do not review in the middle of a sharp move in either direction. Anchoring on a price that is currently unusual will distort the reading, and the entry will still be there next month.
Journals that are kept without scheduled review dates tend to be opened when something has gone badly. The entries that get examined are therefore the ones attached to painful outcomes, and the ones attached to decisions that quietly worked, or that were correctly declined, are never revisited. The picture of one's own thinking that emerges is assembled entirely from its failures, which produces two opposite errors at once: an exaggerated sense of how often the reasoning was poor, and no information whatsoever about what the reasoning gets right.
Write the review date into the entry itself, and keep the reviews that were scheduled even when nothing has happened. An uneventful review is not a wasted one - it is the part of the sample that a loss-triggered review always omits.
Entries written as prose without quantities read well and turn out to be unfalsifiable in practice. 'I expect this to do well over the next few years' is compatible with almost any outcome, and at review it produces a discussion rather than a finding. Entries that contain a number, a threshold and a date produce an answer instead, and the answer is frequently surprising - including, often enough, that a decision now regarded as a mistake contained a forecast that was met.
Before filing an entry, check that at least one sentence in it could be shown false by a fact. If none can, the entry will not be reviewable, and the effort of writing it has bought nothing.
Create a free account to read the rest and apply it to your own investing.