How to journal trades so an AI can find your psychology patterns
An AI can only read what you wrote. Most trade notes record the outcome, which is the one thing your broker already knows.
Write the decision, not the outcome. At entry, record why you took the trade, what you expected, what would prove you wrong, and how you felt about the size. Write it before you know the result. Across dozens of notes, an AI can read those words back as a pattern.
Why can’t an AI learn anything from “+$300, good trade”?
Because it records the result, which your broker already knows, and nothing about you. There is no reason, no expectation and no state of mind in it. An AI reading a year of notes like that can only restate your P&L in sentences. Nothing new comes back.
Every broker export already contains the outcome. Symbol, size, entry, exit, fees, result. If your note adds “good trade” to a row that already says +$300, you have written a label, not a record.
The problem is not that the note is short. It is that the note contains no variable. A pattern needs something that changes between trades and can be compared: a reason, an expectation, a feeling about size. Outcome-only notes have none, so a hundred of them carry the same information as one.
What should a trade note actually contain?
Four things at entry: why this setup, what you expected to happen, what would tell you the idea was wrong, and how you felt about the size you chose. Anything else is optional. Four honest sentences beat a form with twenty fields you will abandon.
Four fields, and none of them is a number:
- Why. The setup, in your own words. “Third test of the same level, volume drying up.”
- Expectation. What you think happens next, and roughly how far.
- Invalidation. The thing that would tell you the idea was wrong — not just the stop price, the reason behind it.
- State. How you felt about the size you chose. Comfortable, stretched, slightly too big because of yesterday.
That last one is the one people skip, and it is the one that carries the psychology. If size is the part you keep getting wrong, work it out before entry with the position size calculator and then write down whether you actually used the number.
What do I write at entry versus at exit?
At entry: the thesis, the invalidation, the size and your state. At exit: what actually happened, whether you followed the plan you wrote, and what you did differently in the moment. The gap between those two notes is the material an AI can work with.
Two notes per position, not one. The entry note is the plan. The exit note is what you did with it.
What makes the pair valuable is the difference between them. A trader who writes “stop below the level” at entry and “moved it down twice” at exit has recorded something no equity curve will ever show. Written as a single note after the close, that difference disappears — the plan gets remembered as whatever ended up happening.
Why does it matter that I write before I know the outcome?
Because a note written after the close is written by someone who knows the answer. Hindsight quietly rewrites the reason. A note written at entry cannot flatter you, so the record of your thinking stays honest, and honest input is the only kind worth reading back.
Nobody writes “I took this because I was bored and wanted to be in something” after a winner. The reason gets upgraded on the way out. This is not dishonesty, it is how memory works, and it is why journals written on Sunday about the week gone by are pleasant to read and nearly useless to analyse.
Writing at entry removes the incentive. You do not yet know whether the note will end up flattering you, so there is nothing to protect. That is the only reason the record stays honest.
What does a good trade note look like?
Short, specific and written in your own voice. It names the setup, the level you were watching, the thing that would end the trade, and one honest line about how you felt sizing it. Two or three sentences is enough if all four are there.
AAPL long. +$300. Good trade, stuck to the plan.
AAPL long, third test of 224 with volume falling into it. Expect a push to 231 over two or three days. Wrong if it closes back under 222 — that would mean the level is supply, not support. Size is a full position and it feels fine, which is worth noting because I was under-sizing all last week.
Out at 228 for +$300. Took it early because the tape got choppy at lunch, not because anything I wrote was invalidated. Third time this month I have exited on discomfort rather than on my own level.
The first version supports one sentence back: you made $300. The second supports a question about why discomfort keeps ending trades before the thesis does. Unfamiliar terms in your own review are worth looking up in the glossary rather than guessing.
How many notes before patterns show up?
Dozens, not a handful. A pattern is a thing that repeats, so it needs repetitions to sit on. Ten notes give you ten anecdotes. Once the same phrase keeps turning up across different setups and moods, a reader, human or machine, can name what recurs.
One note is a diary entry. The value arrives later, when the same words start appearing on their own — the same three-word phrase about size, the same time of day, the same setup you keep taking when you are already down for the week.
This is the honest cost of the method: nothing useful comes back in week one. Nothing useful can. What you are doing in the first month is building the corpus, and the reflections get sharper in proportion to how much of yourself is actually in it.
What can an AI actually tell me, and what can it not?
It can read your own words back and describe what keeps appearing in them. It cannot diagnose you, predict a trade, or improve your returns. Iris in sizeforzero reads the notes you wrote and answers in plain language. It is a mirror, not an oracle.
Iris reads what you wrote and reflects it back in plain prose. It does not score you, it does not forecast, and it has no opinion on what you should trade tomorrow. Anything claiming an AI improves returns from a trade journal is selling something we would not buy.
The useful version is unglamorous: a reader with perfect recall of every note you have written, telling you what it noticed. You already knew it somewhere. Seeing it in a sentence is the difference.
What does sizeforzero include?
The journal is free forever — unlimited trades, every account, full analytics, calendar P&L, CSV in and out — and two AI reflections a month. Pro turns the AI always-on: a daily reflection, the Sunday letter, a read on every position, and a monthly digest.
The journal itself is not the paid part. Trades, accounts, analytics, the equity curve, the calendar and CSV import and export are free forever, with no card. Free also includes two AI reflections a month, so you can read one before deciding anything.
Pro is the always-on tier: a daily reflection, the Sunday letter written from your week, a read on every position, and a monthly digest — never metered or rationed. Letters are kept indefinitely, because a read of your own mind from eight months ago is the part that gets more valuable with time.
If you trade in a registered account, the tax side is worth reading too: when the CRA calls TFSA day trading business income. Canadian traders can also start from the Canadian journal overview.
A note on what this is
sizeforzero is a journal, not a broker and not an advisor. It does not execute trades, hold funds or recommend positions.
For informational purposes only. Not financial advice. Past performance does not guarantee future results.