The Weekly Review: Turning a Trading Log into Lessons

A trading journal collects data. A weekly review turns it into a decision. Without the review, the journal is a record of what happened; with it, the journal becomes a feedback loop.

This guide gives you a 30–45 minute weekly routine that works with a small sample, a structure for using AI to find patterns without inventing them, and a rule for turning the output into one change for the next week.

It is the companion to The Trading Journal Fields That Matter and the Trading Journal Template.

A weekly review should produce one change, not a new identity.

Why Weekly, Not Daily

A daily review is about the session: did I follow the plan, what worked, what failed. A weekly review is about the pattern: which behavior repeats, which setup is actually working, and which rule is being ignored.

Daily reviews are too close to the emotion. Monthly reviews are too far from the behavior. Weekly is the interval where a pattern can be seen and still changed.

If you trade part-time, the review can be every four or five sessions rather than every calendar week. The point is the interval, not the calendar.

The Four Questions

The review answers four questions.

Did I follow the rules? Average compliance, number of breaches, most common breach.
Where did the breaches happen? Time of day, setup, instrument, emotional state, after a win or a loss.
What did they cost? In R, in P&L, in opportunity.
What one rule will I change or reinforce next week?

If the review does not answer all four, it is not finished.

The Routine

Step 1 — Aggregate (5 Minutes)

Fill in the week’s numbers.

Metric	This week	Last week	Notes
Trades
Win rate
Average R
Profit factor
Largest win
Largest loss
Average compliance
Trades below 8/10 compliance

The numbers are descriptive, not a verdict. A losing week with high compliance may be a better week than a winning week with low compliance.

Step 2 — Cluster The Deviations (10 Minutes)

Group the low-compliance trades.

Time: do they cluster at the open, the close, or around a news release?
Setup: is one setup consistently improvised?
Instrument: is one market producing most of the breaches?
State: what emotional state precedes the breach?
Sequence: do breaches happen after a win, after a loss, or after a long quiet period?

The output is a short list, not a narrative. For example:

“Five low-compliance trades; four in the first 15 minutes; three after a loss; all were ‘quick scalps’ that were not in the plan.”

Step 3 — Compare Plan To Behavior (10 Minutes)

For each trade, ask:

Which scenario was it from?
Was the trigger present?
Was the invalidation respected?
Was the size within the plan?
Was the exit consistent with the plan?

This is where the journal and the plan meet. If the plan says “only short the retest” and the journal shows a market short at the low, the gap is the lesson.

Step 4 — Choose One Change (10 Minutes)

One change. Not five. It must be specific, observable and testable.

Bad change	Good change
“Be more disciplined.”	“No entries before 09:45.”
“Stop revenge trading.”	“After a full loss, leave the screen for five minutes and take only a plan setup.”
“Trade better.”	“Only the primary scenario; no ‘quick scalps’ for five sessions.”
“Risk less.”	“One contract on the first trade of the day for the next week.”

The change goes into the next daily plan as a rule. It also gets a measure: how will you know whether you followed it?

Step 5 — Write It Down (5 Minutes)

Record:

The week’s numbers.
The three most frequent deviations.
The one change.
The measure.
The date of the next review.

The record is the memory. Without it, next week’s review starts from zero.

Using AI In The Weekly Review

AI is useful for steps 1–3 because it can count, cluster and quote. It is not useful for step 4, because the change is a judgment about what matters.

A safe prompt:

“Here are 20 anonymized journal entries from this week. Produce:

1. Aggregate metrics: trades, win rate, average R, profit factor, largest win and loss. 2. Average compliance and the number of trades below 8/10. 3. The three most frequent deviations, with quotes. 4. The times of day and emotional states associated with the deviations.

Do not give advice. Show the evidence and the sample size.”

A follow-up:

“For each pattern, tell me whether the sample is large enough to be meaningful and what data would be needed to test it.”

Do not paste account numbers, balances, broker credentials or personal data into a consumer AI tool. Anonymize the entries and use an approved business-tier tool.

The model is a clerk and a mirror. The trader is the reviewer.

A Worked Example

A trader reviews 14 trades.

Aggregate: 6 wins, 8 losses; average R 1.6 on winners and −1.0 on losers; profit factor 1.2; average compliance 6.4; five trades below 8/10.

Deviations:

Four of the five low-compliance trades were in the first 15 minutes.
Three were “quick scalps” that were not in the plan.
Two followed a full loss.
The emotional state before the low-compliance trades was “rushed” or “frustrated.”

Cost: the low-compliance trades produced −3.1R; the plan trades produced +2.4R.

Change: “No entries before 09:45 and no quick scalps for five sessions. Measure: zero low-compliance trades in that window.”

The change is specific and testable. Next week’s review will show whether it worked.

What If The Week Was Profitable But Compliance Was Low?

Treat it as a warning. A profitable rule-break teaches the wrong lesson and is often followed by a larger loss when the same behavior meets a less forgiving market. The review should reward the process, not the outcome.

What If There Were Only A Few Trades?

Review the process instead of the outcome: did you write the plan, mark the levels, check the risk and complete the journal? Process metrics work with small samples; outcome metrics need more data.

What If I Missed The Review?

Do it before the next session. A review that is two days late is still useful; a review that is skipped is a lost week.

Common Mistakes

Reviewing only the losses.
Changing five things at once.
Choosing a vague change (“be more disciplined”).
Using the P&L as the only measure.
Writing a narrative instead of counting.
Letting AI give advice instead of showing evidence.
Not putting the change into the next plan.

Frequently Asked Questions

How Long Should A Weekly Review Take?

Thirty to forty-five minutes. If it takes longer, the journal is too detailed or the review is trying to do too much.

What If I Trade Part-Time?

Review every four or five sessions. The interval matters more than the calendar week.

Should I Review Every Trade?

Review the low-compliance trades in detail and the rest in aggregate. The pattern is in the deviations, not in every entry.

Can AI Do The Review For Me?

It can aggregate, cluster and quote. It cannot decide what matters or choose the change. Use it for the evidence, not the judgment.

What Is A Good Compliance Score?

Eight or higher is a useful target. The score is not a moral judgment; it is a measure of how closely the trade followed the plan.

How Do I Know If The Change Worked?

Define the measure before the week starts. “Zero entries before 09:45” is measurable; “be more patient” is not.

Sources

Barber, B. M., and Odean, T. (2000), “Trading Is Hazardous to Your Wealth,” Journal of Finance, 55(2), 773–806. Overconfidence and excessive trading are associated with worse net returns.
Barber, B. M., Lee, Y.-T., Liu, Y.-J., and Odean, T. (2014), “The Cross-Section of Speculator Skill: Evidence from Day Trading,” Journal of Financial Markets, 18, 1–24. Less than 1% of day traders predictably earned positive abnormal returns net of fees.
Chague, F., De-Losso, R., and Giovannetti, B. (2020), “Day Trading for a Living?” SSRN 3423101. Most persistent day traders lost money.
NIST, Artificial Intelligence Risk Management Framework (AI RMF 1.0), for the human-review principle.

No statistic in this article is invented. Verify figures against the primary source before republication.

Trading involves substantial risk of loss. This article is educational information, not financial, investment, legal or tax advice, and it is not a personal recommendation or solicitation. AI output can be inaccurate or incomplete; keep personal data out of consumer AI tools and verify every figure. Past performance and backtests do not guarantee future results.