What counts as a pattern in your trading journal: sample size, the win-rate gap and the five-trade slice
A pattern needs 20 trades in the journal, 5 closed trades in the slice and a 15-point win-rate gap. Why a five-trade flag is still a question.
Published by TradeFlow Quantum. Filed under statistics, patterns, journal.
Any journal that slices your history by hour, weekday, setup and tag will always have a worst slice. Whether that slice is a pattern or the noise of a small sample is a question of counts and thresholds, so this post gives the ones TradeFlow Quantum's pattern flags use, read from the detector's code, and what each one does and does not protect you from.
What the journal slices
A slice is a group of your trades that share one thing. The detector cuts your history these ways: hour of day in US Eastern time (pre-market, the opening hour, mid-day, afternoon and the close), weekday, setup, side, two bands of logged psych score (4 or below, and 5 to 6), the trade straight after a realised loss, the trade straight after a realised win, the top quarter by notional size (entry price times quantity), symbol, and each mistake tag and emotion tag you put on a trade. Each slice is compared with your whole history.
Gate one: 20 trades in the journal
Below 20 trades the flags stay empty. The count is rows, open or closed: an open fill counts toward it, though it has no outcome yet. Win rates are worked out on closed trades only, so an open fill is never counted as a loss, in a slice or in your overall rate.
Gate two: 5 closed trades in a slice, 8 for a symbol
A slice is compared only once it holds 5 closed trades. A single symbol carries a higher bar: 8 trades in the slice, open or closed, on top of the 5 closed, because one ticker traded a handful of times says little. A slice under the bar is not reported as fine; it is not judged at all.
The test: a 15-point gap, or a losing average
The gap is your overall win rate minus the slice's. A slice is flagged when that gap is 15 percentage points or more, or when its expectancy, the average P&L per closed trade, is below zero. A slice is never flagged for winning more often than you usually do; it can still be flagged for losing money on average.
The second test matters on small slices. Four winners of $50 and one loser of $250 is a win rate of 80%, and it is flagged anyway: the slice loses $10 a trade on average. One large loss can carry a five-trade slice over the line.
Why a five-trade flag is a question
In a slice of 5 closed trades, each trade is 20 points of its win rate. That is more than the 15-point gap, so one trade going the other way can raise a flag or clear one. In a slice of 20 closed trades each trade is 5 points, and it takes several trades to move the slice across the line. The detector does not run a significance test; its gates are fixed counts and a fixed gap, not a confidence level. It also compares each slice with an overall rate that includes the slice's own trades.
Flags are ranked by a severity score: the gap times the natural log of one plus the slice's closed trades, plus one point for every $100 the slice lost. At the same gap, a slice of 20 closed trades scores about 1.7 times one of 5, so the larger slice sits higher. At most 10 flags show, strongest first, and each card on the mistakes page shows the slice's sample size. Read the count before the gap, and look at the trades behind a flag before deciding what it means; the flags describe your history and recommend nothing.
One finding, right after an import
When a CSV import finishes, or a broker sync brings in new trades, the result panel shows one line from the same detector: its strongest flag, with the slice's closed trades, the slice's win rate against your rate across all your trades, and its net P&L, plus a link to every flag and the trades behind it. Under 20 trades the line says how many the journal holds so far; when no slice is flagged it says that instead. Sample trades are left out, so the line describes only your own trades.
The export guide has each broker's CSV menu path, and the supported brokers page lists the connections that sync directly.
Where the flags fall short
- No significance test. A flag is a count and a gap. At 5 closed trades, chance alone can produce one.
- Underperformance only. The flags look for slices below your baseline; they do not list where you do best.
- Notional size, not risk. The size slice ranks trades by entry price times quantity, so it has no view of a futures contract's multiplier or a stop distance.
- Tags are only as consistent as you are. Tags are trimmed and lowercased, so FOMO and fomo are one slice, but fomo and chased are two.
The flags sit beside the P&L calendar, the weekly digest and the rest of the journal features. The five-trade rule covers sample size for a strategy as a whole, and the beginners' checklist covers what a new journal can tell you early.
$15/month or $150 a year, one tier, 7 days free with a card at signup and nothing charged during it. The ten direct broker connections and CSV import work during the trial.
Related posts
Not financial advice. This post reflects the author’s opinion based on publicly-available information at the time of writing. Mention of third-party products is not an endorsement; product features and prices change over time. Past performance does not guarantee future results.