- A backtest log records what the model does. A journal records what you do. They answer different questions, and most traders keep the first while believing they are keeping the second.
- Log the criteria, not just the result. Which conditions were present and which were missing turns a record of wins and losses into a record of which rule you keep bending.
- The skipped column matters as much as the taken one. Setups you correctly passed on, and setups you wrongly passed on, are invisible in a normal journal — and they are half of your decision-making.
- Grade execution before outcome. A losing trade taken exactly to plan is a good day. A winning trade that broke three rules is a loan, and the journal is where you find out when it gets called.
Search for an ICT trading journal and you find journal software. TradeZella, Treydly, spreadsheets and Notion templates, all built to log entry, exit, size and profit. None of them is wrong. None of them is built for this methodology either.
A generic journal answers one question: did this trade make money? For an ICT trader that is close to the least useful question available, because the methodology is a set of conditions, and money arrives or does not for reasons that often have little to do with whether you applied them. A trade can break every rule and win. A trade can follow every rule and lose. If the journal only knows the result, it will teach you to repeat the first and abandon the second.
This page is about a journal that records the thing you can actually control.
Journal and backtest log are different tools
The site already covers the backtest log — the record you keep while replaying historical sessions. It is worth being precise about how the two differ, because people often keep one and assume it does the job of both.
| Backtest log | Live journal | |
|---|---|---|
| Question it answers | Does the model have an edge? | Do I execute the model? |
| Who makes the errors | Mostly the model | Mostly you |
| Emotional state recorded | None — there is nothing at stake | Essential — it is often the cause |
| Skipped setups | Rare; you see every one in replay | Common, and revealing |
| What a bad entry means | The rules need changing | You did not follow them |
The gap between the two records is where most of the useful information is. If the backtest says a model works and the journal says you are losing with it, the difference is you — and the journal is the only instrument that can show you which part.
What to log for every setup
Every setup you identify, not every trade you take. That distinction is the whole design.
| Field | What goes in it |
|---|---|
| Date, session, window | Which kill zone or macro the setup formed in |
| Bias going in | Written before the session, copied here unchanged |
| Draw | The named level you expected price to reach |
| Sweep present? | Yes / no, and which level was taken |
| Confirmation | Which event: MSS body close, CISD, displacement — or none |
| Entry type | FVG, order block, OTE, breaker |
| Stop reference | Which structural price the stop sat beyond |
| Taken or skipped | And in one line, why |
| Execution grade | A, B or C — how closely you followed the plan |
| Outcome | In R, entered last |
| State | One word: calm, rushed, tired, chasing, bored |
Two things about that list. Outcome goes near the end, and you fill it in last, so that the grade is not contaminated by knowing how it turned out. And there is a row for setups you did not take, which almost no template has.
The skipped column
Every morning you look at several candidate setups and take one or none. The ones you pass on vanish from a normal journal completely. That is a large blind spot, because passing on a setup is a decision as consequential as taking one.
Skipped setups fall into two groups, and the journal should let you tell them apart:
- Correct skips — a criterion was genuinely missing. No sweep, no confirmation, against bias, outside the window. These are the process working, and on a no-trade day they are your only record that you did anything right.
- Hesitations — every criterion was present and you did not take it. Usually because the previous trade lost, or the move looked too fast, or you were not watching closely enough.
Hesitations are expensive and invisible. A trader who takes every qualifying setup and a trader who takes two-thirds of them are running different systems, even with identical rules — and only the second one does not know it. The skipped column is how you find out which you are.
Grade execution, then outcome
The single change that makes a journal useful is grading each entry on whether you followed the plan, separately from whether it made money. Trading psychology covers the reasoning; here is the scale.
| Grade | Meaning |
|---|---|
| A | Every criterion present, entry and stop exactly to plan, managed as written |
| B | Valid setup, one deviation — early by a candle, stop nudged, target changed live |
| C | A rule was broken to take it: no confirmation, against bias, outside the window |
Once each entry carries both a grade and a result, four combinations exist, and they mean very different things.
- A and a loss is a good trade. The market did something the model allows for. Nothing to fix.
- A and a win is the process working as intended.
- C and a loss is the obvious lesson, and the one people learn anyway.
- C and a win is the dangerous one. It teaches you that breaking the rule was fine. Enough of them, and the rule stops being a rule.
A generic journal records the last category as a success. That is how a trader with a sound method slowly stops using it.
Walkthrough — what two weeks of entries showed
The entries below are constructed to show how the pattern surfaces. It is a teaching illustration rather than a record of real trades.
| Day | Window | Sweep | Confirmation | Taken? | Grade | R |
|---|---|---|---|---|---|---|
| Mon | NY AM | Yes | MSS close | Taken | A | −1.0 |
| Tue | NY AM | Yes | None | Taken | C | +2.1 |
| Wed | London | No | — | Skipped — no sweep | A | 0 |
| Thu | NY AM | Yes | None | Taken | C | −1.0 |
| Fri | NY AM | Yes | CISD | Skipped — “too fast” | B | 0 |
| Mon | NY AM | Yes | MSS close | Taken | A | +1.8 |
| Tue | NY PM | Yes | None | Taken | C | −1.0 |
| Wed | NY AM | Yes | Displacement | Skipped — after Tuesday’s loss | B | 0 |
| Thu | NY AM | Yes | None | Taken | C | −1.0 |
| Fri | NY AM | Yes | MSS close | Taken | A | +2.4 |
What a generic journal would say. Seven trades, three winners, net +2.3R. A mediocre fortnight. Perhaps the model is not working. Perhaps try a different one.
What this journal says. Sort by grade and the picture changes completely.
- The three A-grade trades returned +3.2R between them.
- The four C-grade trades returned −0.9R. Every one was an entry without confirmation. One of them won, and that single win on Tuesday is almost certainly why the habit continued through the second week.
- Two qualifying setups were skipped — Friday’s because it “looked too fast,” the second Wednesday’s because of the previous day’s loss. Both had every criterion present.
So the trader does not have a model problem. The trader has exactly one execution problem — entering before confirmation — and a related hesitation problem that shows up after losses. Neither is visible in a profit-and-loss column. Both are obvious in ten rows once the grade and the skipped column exist.
The fix is equally specific. Not “trade better” or a new model, but one rule for the next fortnight: no entry without the confirmation written in the plan, and log every hesitation with its reason. Entry confirmation covers what that event should look like.
The weekly review
A journal nobody reads is a diary. Once a week, with nothing open on the chart, sort the entries three ways and answer three questions:
- By grade. What did the A trades return, and what did the C trades return? If the C trades are not clearly worse, look harder — small samples flatter bad habits.
- By the missing criterion. Among the C trades, which rule was broken most often? It is nearly always the same one. That is next week’s only focus.
- By skip reason. Were the skips correct, or were they hesitations? Do hesitations cluster after losses, on certain days, in certain windows?
Then change one thing. A review that ends in five resolutions ends in none.
Spreadsheet, app or paper?
Whichever you will actually fill in every session. That is the only criterion that matters, and it rules out most elaborate setups by the second week.
What does matter is that the tool can do three things: hold the custom fields above, record skipped setups as rows, and sort by grade. A plain spreadsheet does all three. Most trading journal apps do the first and third but are built around executed trades, so skipped setups have to be forced in as zero-size entries — workable, but check before paying for one. Paper is fine if you transfer it weekly; the sorting is the part you cannot do by hand for long.
Screenshots help. One at the moment of decision, marked with your levels, is worth more than any number of fields — because in a month you will not remember what the chart looked like, only what happened after.
Five ways journals fail
| What people do | Why it fails |
|---|---|
| Log only trades taken | Half your decisions are invisible, including every hesitation |
| Record the result first | The outcome contaminates the grade; winners get remembered as well executed |
| Log profit instead of R | Different sizes and instruments become impossible to compare |
| Write the bias after the session | Hindsight rewrites it to match what happened |
| Never review it | A record that is never sorted cannot tell you anything |
Log every setup you identify, taken or not. Record which criteria were present. Grade execution before you look at the result. Sort by grade once a week and change one thing. The journal’s job is not to tell you how much you made; it is to tell you which rule you keep breaking, which is the only thing in the whole process you can actually fix.