Every number in a trading journal is an estimate from the trades you have taken so far. Sample size is how many trades that estimate rests on. With ten trades a win rate of 60% could easily be a 40% strategy on a good run; with three hundred trades it almost certainly is not.

How uncertain is a win rate?

A rough 95% margin of error for a win rate p measured over n trades:

margin ≈ 2 × √(p × (1 − p) / n)

For a measured win rate of 50%:

Trades Margin of error Plausible true win rate
20 ± 22 points 28% to 72%
30 ± 18 points 32% to 68%
100 ± 10 points 40% to 60%
300 ± 6 points 44% to 56%

Rules of thumb

  • Below 30 trades: treat every statistic as a rumour. Profit factor and expectancy can swing wildly with one trade.
  • 30 to 100 trades: direction is visible, size is not. Good enough to stop doing something clearly broken.
  • 100+ trades of the same setup: enough to compare setups, sessions or days of the week with some confidence.

Slicing makes samples smaller. 200 trades split across five setups and four sessions leaves cells of ten trades each. Be careful with conclusions from small cells.

In SetupBook

Any metric card computed from fewer than five trades carries an n= badge, and the psychology cards do the same for the days or trades behind them. The trader score waits for at least ten closed trades in the period. The breakdown tables in Analytics always show the trade count next to each row, so a 100% win rate from two trades is visible as exactly that.