Trade Sample-Size Calculator
An edge you can see over a handful of trades might be real — or it might be a coin that landed heads a few times running. This shows the minimum number of trades below which an edge your size simply can't be told apart from luck, and at your pace, how long that is. It's a floor to clear, not a finish line: reaching it means the sample is finally big enough to take the result seriously — never that the edge is proven, will last, or is safe to trade.
You'd need at least ~97 trades before an edge this size could be separated from zero-edge luck at 95% confidence. Treat it as a minimum, not a target: real wins and losses vary in size, trades cluster in streaks, and if you picked this strategy after trying several, the honest number is higher — often 1.5–2×.
“Statistically significant” here would only mean luck alone would rarely fake a result this size — not a 95% chance the edge is real, not a promise it lasts, and not a green light to trade.
The only trades that count toward telling luck from edge are ones your rules couldn't have been fit to. Forward-test on paper: accumulate them out-of-sample, at live prices, with no hindsight.
The longer it runs, the more it's worth to you — as evidence you can trust your own rules, not as something to sell. We never touch your funds and never place an order; you decide what, if anything, to do on your own broker.
- • Your win rate and average win/loss become an expectancy per trade (the average R you earn) and a per-trade swing (how much a single trade bounces around).
- • We ask how many trades it takes before that expectancy stands clear of zero at your confidence level: N = (z ÷ (expectancy ÷ swing))², with a two-sided z (95% → 1.96).
- • We floor N at a few dozen trades — significance on a handful is noise — and label it “at least”: the simple model assumes every win and loss is the same size, and real ones vary, always pushing the true number higher.
- • Open this from a real backtest and it uses your actual trade-by-trade results instead of the average — closer to the truth, and almost always a bigger number.
Worked example: a 50% win rate with a +1.5R average win and a −1R average loss needs at least ~100 trades before the edge could be separated from luck.
Clearing the bar removes one doubt — “too small a sample” — and none of the others. It says nothing about whether the edge survives new data, costs, or a change in the market, or whether you found it by trying many strategies. For the luck this can't catch — picking a good-looking strategy out of many (the Deflated Sharpe, in Batch scan, is the only check that knows how many you tried), and whether random entries would have scored as well (the Noise Floor, on any backtest result) — use those checks too.
New to this? Real signal or just luck? walks through all four independent ways a good-looking backtest fools you — sample size is only the first.
Educational tool, not advice. We never touch funds or place orders.
A calculator shows what a rule should do on paper. Whether a strategy actually beats simply buying and holding — costs on, losses shown, no hindsight — is a different question, and the only one that pays. Test one on real data, free, no sign-up.