Twenty minutes, once a week, market closed
Most traders review constantly and change nothing. A glance at the P&L after every exit, a mental replay on the drive home, an 11 PM scroll through the day's fills — that isn't review, it's rumination. A real review is short, scheduled, and produces exactly one output: a rule you can check next week.
Here is the whole routine: Sunday, 20 minutes, market shut, your own journal open. Nothing in it predicts Monday's NIFTY. Every minute of it measures what you did last week — because that's the only thing the review can change.
Minutes 0–5: read the calendar, not the candles
Start with the month's calendar heatmap — the grid of green and red days. Don't open a single trade yet. You're answering three questions:
- Which days made money, which lost? Not how much in total — which days. Losses that cluster (every Thursday expiry, every Monday open) are a pattern; losses scattered randomly are variance.
- Did any day have an outsized trade count? If your profitable days average 4 trades and Tuesday shows 14 trades for −₹9,800, Tuesday wasn't a market problem. It was a behaviour problem, and it has a timestamp.
- Is this week's shape different from the month's? One red week in a green month is normal. A red week that looks exactly like the last three is a process telling you what it produces.
Don't diagnose anything yet. Circle the ugliest day and move on.
Minutes 5–10: open your three worst trades
Sort the week by P&L and open the bottom three. Read the notes you wrote at trade time — not the story you remember now. Then tag each one honestly: revenge, oversized, moved stop, held loser.
Honestly is the hard part. The week's worst trade — say −₹6,200 on a NIFTY weekly put, entered at 9:31 AM IST, four minutes after a stop-out on the same strike — is a revenge trade whether it felt like conviction or not. The timestamps decide, not your mood. And if one of the three was properly sized, had a stop that you respected, and simply lost? Leave it untagged. That's a clean loss, and clean losses are tuition, not mistakes.
Minutes 10–15: read the invoice
Now zoom out from three trades to the totals: what did each tagged habit cost this week, and this month? This is where the review stops being reflective and starts being financial. A month of tagging might read:
- Oversized: −₹15,300 across 2 trades
- Revenge trades: −₹11,400 across 6 trades
- Moved stop: −₹7,900 across 3 trades
Two oversized trades cost more than six revenge trades — you would never guess that from memory, because the revenge trades felt worse. Pricing each mistake in rupees is what makes the next step obvious: the biggest line on the invoice is next week's target. Not all of them. The biggest one.
Minutes 15–20: one rule, written down, checkable
First, check last week's rule. Did it hold? Count the violations from the data, not from memory — if the rule was “no entry within 15 minutes of a stop-out”, the timestamps give you a number, and the number is either zero or it isn't. A broken rule carries over. It doesn't get replaced by a shinier one.
Then write next week's rule, aimed at the biggest line on the invoice. The test of a good rule: next Sunday, the data alone can say kept or broken with no judgment call. “No entry within 15 minutes of a stop-out.” “Maximum 2 lots on expiry day.” “No new trades after 2:30 PM IST.” Compare with “be more disciplined” — unfalsifiable, therefore useless. If your worst habit was oversizing, a lot-size cap. If it was overtrading, a daily trade cap. One habit, one rule, one week.
Three ground rules for the review itself
- Never review mid-session. Reading your journal at 12:40 PM with an open SENSEX position isn't review — it's hunting for permission to move a stop. Sunday exists because nothing is open on Sunday.
- One rule at a time, not five. Five rules is a manifesto; you'll keep none of them and feel bad about all of them. One checkable rule, kept for a week, is a changed behaviour.
- Judge decisions, not outcomes. An oversized punt that made ₹4,000 is still a mistake. A well-sized trade that hit its stop is still a good decision. Options P&L is noisy enough that outcomes lie weekly; process only lies if you let it.
The honest objection to all of this is the prep work. Pulling the heatmap, sorting the losers, totalling the tag costs — that's the boring half of the 20 minutes, and it's why most Sunday reviews die by week three. It's also the half a machine should do. PnL Book keeps the heatmap, the tags and the rupee totals current as you journal, and its weekly AI-written review reads your week's numbers and points at the pattern — three of your five losers share a tag, your trade count doubled after Wednesday's stop-out. It reads your own past trades; it doesn't predict markets or hand out calls. The pattern gets found for you. The rule is still yours to write — and yours to keep until next Sunday.