How to build a trading journal that actually changes decisions

A trading journal is not a record of wins and losses. It is a decision database. The objective is to preserve the information available at the moment of the trade, so a later review can distinguish a flawed process from an ordinary uncertain outcome — a distinction that is invisible in a profit-and-loss column and obvious in a well-structured record.

Record the setup before it changes

Everything that describes the decision has to be captured before the outcome is known, because memory rewrites it afterwards with complete confidence and no awareness of doing so.

Screenshots help but are not searchable. Written context is what lets you ask, six months later, how a specific setup behaved when funding was elevated and a macro release was inside the horizon — a question no folder of images can answer.

Create fields that answer future questions

Design the journal backwards from the questions you will want to ask. Setup type, market regime, session, catalyst, entry reason, invalidation, planned size, actual size and plan adherence are the fields that generate answers. Add a short note on state of mind only when it produces something actionable — "entered late after a loss", "moved stop after checking price four times" — rather than as a running commentary. The purpose is evidence, not self-criticism.

Use tags consistently or not at all

A journal with ten names for the same setup cannot answer whether that setup works. Keep a small, closed vocabulary — five to eight setup names, four or five regime labels, a fixed list of execution errors — and resist adding a new one for a trade that is almost the same as an existing category. The value of the record is entirely in its groupability.

• Setup: one of a fixed list, no free text.
• Regime: trending, ranging, event-driven, thin liquidity.
• Catalyst: scheduled, unscheduled confirmed, developing, none.
• Error: none, late entry, oversized, stop moved, plan not written, event ignored.

One entry, filled in

An abstract template is easy to agree with and hard to use. Here is what a single row looks like when the fields are doing their job — written before the position was opened, in about ninety seconds.

Everything in that table is a fact at the moment of the decision. Nothing in it requires knowing the outcome, which is precisely why it is still worth reading in six months. The fields added afterwards are short: actual size, actual exit, whether the plan was followed, and one line on what happened.

What not to track

Journals die from overhead. Every field that takes effort and never answers a question is a reason not to write the next entry, and an abandoned journal is worth less than a small one. Two specific traps.

• Indicator readings you do not trade off. Recording RSI on every entry produces a column you will never filter by.
• Long emotional narratives. One tagged error is queryable; three paragraphs about how the trade felt is not.
• Screenshots as the primary record. Useful as an appendix, useless as a dataset.
• Metrics that are properties of the market rather than the decision, unless they are your regime tags.

The test for any field: name the question it will answer at the monthly review. If you cannot, delete it. Fifteen fields you fill in every time beat forty you fill in for the first two weeks.

Review process metrics, not only outcomes

Track whether the entry followed the plan, whether the position was sized correctly, whether the stop or target was changed, and whether a scheduled event was ignored. These are controllable. The outcome of a single trade is a noisy score for a decision, and — as the sampling arithmetic in losing streaks and drawdown maths shows — so is the outcome of twenty.

Questions for a monthly review

1. Which setup had the clearest written thesis and the best plan adherence?
2. Where did actual size exceed planned risk, and what preceded it?
3. Which losses were normal outcomes and which were execution errors?
4. What one rule would remove the most repeated avoidable mistake?
5. Did the regime change, or did only the results change?

Keep reviews narrow. A weekly pass should look only for repeated execution errors; a monthly pass can compare setup types and regimes. Drawing conclusions from a handful of trades is how a sound process gets abandoned and a lucky one gets adopted.

The records worth keeping most

A journal becomes valuable when it contains losses that were handled correctly. Those entries are the reference class that protects against the two symmetrical mistakes: abandoning a sound process during a normal losing run, and adopting a poor one after a favourable one. Both feel like learning at the time.

The standards that make a personal record credible are the same ones worth demanding of anyone else's — timestamped before the outcome, every entry included, a fixed scoring rule, the sample size stated. How to check whether a trading signal record can be falsified works through the full list.