A win rate from a backtest is one draw. This puts a 95% range on it, so you can see whether the trade count is big enough to mean anything at your payoff.
The interval straddles break-even (50.0%): this many trades can’t tell the strategy from one that breaks even.
28 winners out of 50 trades is a 56% win rate. The 95% Wilson interval on that runs from 42.3% to 68.8%. With wins and losses the same size, break-even is 50%, so the bottom of the range loses money: fifty trades can’t tell this strategy from a coin flip.
If the whole interval sits above your break-even win rate, the sample is large enough on win rate alone. If it straddles break-even, you need more trades before the result says anything.
The usual formula, win rate ± 1.96 × √(p(1 − p)/n), falls over at small samples. Nine winners out of ten gives 71.4% to 108.6%, which is impossible. The Wilson interval on the same data gives 59.6% to 98.2%: still wide, and at least possible. That is why this calculator shows both.
It assumes the trades are independent. Ten trades opened together on assets that move together count as fewer than ten. It also says nothing about costs, fills or how many variants were tried before this one; those are other checks on the checklist. To see how many trades you’d need, use the trades-needed calculator.
The one-page checklist, free. Confirm the email and the PDF opens straight away; one worked check each Tuesday after that.
Interval arithmetic only; nothing here is a result of any strategy or investment advice.