The weekly review
Learning objectives
- Explain why one week's P&L cannot grade your process, and read the numbers you control instead: adherence and deviation tags.
- Run the 30-minute weekly review in a fixed order and end it with at most one dated plan change.
- Judge a plan change by the behaviour it targets, not by next week's net R.
- Grade the review's claims: house rules, psychology measured in other domains, and what is not yet measured.
Friday, 16:00 ET (04:00 ICT Saturday, winter). Last week your journal closed at +7.5R and you felt sharp. This week the same setups, the same size and the same rules closed at −5.0R, and you feel like a different trader. You are not. On Sunday, with the market shut, you get thirty minutes with the journal and one question: which of these numbers says anything about you? This lesson is about the state of mind you review in. The journal fields come from M7.03; the full review tables (per-setup expectancy with its interval, the tag Pareto) belong to the Desk course’s review lesson, DESK1.06.
Key term
The weekly review (đánh giá hàng tuần) is a fixed, timed session, once a week with the market closed, that reads your journal and ends with at most one written change to the plan House rule. It separates two kinds of number Definition. Process numbers (số liệu quy trình) count what you did: adherence % (the share of trades taken exactly as the plan said), deviation-tag counts, planned vs realised R. Outcome numbers (số liệu kết quả) count what the market paid: net R, win rate.
Four weeks, one set of rules
DefinitionHow it works
One week of P&L is mostly noise
In the figure the rules never changed; only the order of wins and losses did. That is the normal case, not bad luck Definition. With these rules a 15-trade week has an expected net of about +1.9R and a spread (standard deviation) of about 4.8R, so roughly one week in four ends negative and one in five ends flat, on a method that has a real edge. M7.07 and M7.10 have the arithmetic.
The psychological trap is judging a decision by its result. People rate the same decision as better when it happened to work out (outcome bias), which is measured in experiments on general decisions MeasuredOther market; that it distorts traders’ reviews the same way is plausible, not measured on traders PlausibleOther market. M7.05 covers the bias itself. In a review it does two kinds of damage: the lucky week’s rule breaks get rewarded, and the unlucky week’s correct trades get “fixed”.
Review the numbers you control
A week can show behaviour. If the plan says no new positions after 12:00 ET and the journal shows four, that is a count of your own actions, not an estimate from a small sample Definition. So the review reads in this order, every week House rule:
- Numbers (10 min): trades, adherence %, deviation-tag counts, planned vs realised R, and the week against your house loss lines (M7.08). Net R last, read as noise.
- Biggest leak (10 min): the top deviation tag is the week’s lesson. One lesson, not three.
- One change at most (10 min): written, dated, with the plan’s version number raised (M7.04). Then stop.
The one-change rule
Change the entry, the size and the session hours in the same week, and next week’s numbers cannot tell you which change did what House rule. This is attribution, not superstition: one variable at a time is what keeps any test readable. Be clear about what a change can show and when. Next week shows whether the behaviour changed (the tag count). Whether the P&L improved takes months of trades, not a week.
Monthly and quarterly: same mindset, longer clock
Once a month the per-setup numbers come out, each read with its sample size (M7.03 shows how loose 20–30 trades are). Moving a setup back to sim when it hits a written trigger, such as the −15R line or a per-setup expectancy below zero after a minimum count you set in advance, is a house rule House rule, not proof the setup is dead: 60 trades at −0.05R carry an interval of roughly ±0.3R, wide enough to include a small edge. So bench it; don’t delete it. Watch the adherence trend too: falling adherence while P&L holds up means the results are arriving despite the plan, not because of it. Quarterly is the only time to add a setup or an instrument, never mid-month.
What the data says
So the honest claim is narrow: the review is a way to see your own behaviour clearly and change it one piece at a time. It cannot tell you in a week whether your method works.
Worked example: week 6
A synthetic trader’s Sunday review. Numbers: 16 trades, adherence 81% (90% the week before), net +0.6R. Top tag: NO_SETUP ×4, all after 12:00 ET (00:00 ICT, winter) on Thursday and Friday. Leak: boredom trades after the plan’s session window. One change, written into the plan’s session box and dated: “no new positions after 12:00 ET”.
Week 7: 11 trades, adherence 93%, NO_SETUP ×0, net +2.2R.
What the review may conclude: the rule did its job. The behaviour it targeted disappeared, and that is readable in one week because it is a count of your own actions. What it may not conclude: that the +2.2R came from the change. On 11 trades, net R swings several R either way under the same rules, as the figure shows.
When the review fails
Run week 7 again, but it ends at −3R, with adherence 93% and no NO_SETUP tags. A P&L reviewer reverses the rule: “cutting the afternoon cost me”. A process reviewer keeps it. The change did exactly what it was for, and a −3R week sits well inside what unchanged rules produce (week 2 in the figure was −5R). Judge a rule by the behaviour it was meant to change; judge the method on months of trades.
Common mistake: The review that is really a P&L post-mortem: staring at the red number and feeling bad, or at the green one and feeling smart. Without adherence, tags and planned vs realised R it is mood journaling, not a review.
Common mistake: Reading one good week after a change as proof the change works. Eleven trades cannot separate a better rule from a lucky week; that is treating noise as an edge. Only the behaviour count can be read that fast.
Common mistake: Fixing everything at once. Three changes in one week leave nothing you can attribute next Sunday.
Drill: rank four weeks without the P&L
Take your last four weeks of journal. If you have none yet, use a sim journal.
- Cover the net R column.
- For each week, write adherence %, the top tag with its count, and planned vs realised R.
- Rank the weeks from best to worst process.
- Uncover net R and compare the two rankings.
Solution
Expect the rankings to disagree, at least in part: with 10–20 trades a week, net R is mostly order-of-trades noise. If your best-process week lost money, the review’s job is to keep that process, not to change it. If the rankings match perfectly, do not read that as proof either: four weeks are four data points. Finish by writing the one change your top tag points to, with the date.
Recap
- One week of 10–20 trades cannot grade a method: under fixed rules with a real edge, about one week in four still loses.
- Review the numbers you control first (adherence, tag counts, planned vs realised R) and read net R last.
- Thirty minutes, fixed order: numbers, biggest leak, at most one dated plan change.
- Judge a change by the behaviour it targets; benching a setup is a house rule, not proof the setup is dead.
- That a weekly review improves results is a house belief, not yet measured; the biases it guards against are measured in general psychology.
Further reading
- Baron, J. and Hershey, J. C. (1988). Outcome bias in decision evaluation. Journal of Personality and Social Psychology 54(4), 569–579: the same decision rated better or worse depending on how it turned out.
- Fischhoff, B. (1975). Hindsight ≠ foresight: the effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance 1(3), 288–299: why a review written after the fact remembers more certainty than there was.
- Our lab: a lesson drawn from a trading journal is treated as a hypothesis and only becomes knowledge after it has been measured; the weekly review itself has not been tested against alternatives.
Connects to: M7.10 (the validation the monthly review keeps up) → M7.12, the psychology of size, which the review’s loss lines and size changes run into. Later, the Desk course’s DESK1.06 turns this review into tables and charts.
Educational content only — not financial advice. Futures trading involves substantial risk of loss.
Saved in this browser (account sync comes at launch).