Sample size: how many trades you need

What it is

A strategy's results over a handful of trades are mostly luck. As a rough guide, 30 trades is a bare minimum to see anything, and 100 or more is much better. The maths shows why: with a true win rate of 50%, the win rate you'd measure over 50 trades could easily land anywhere from about 36% to 64% (a 95% range of roughly ±14 points). Over 200 trades that narrows to about ±7.

So a strategy that 'won 70% over 20 trades' may well be a 50% strategy on a lucky run.

How to spot it

• Check how many trades sit behind any statistic before you trust it, including filtered stats such as one session or one setup.

How to use it

• Keep logging and testing until each setup you rely on has at least 30, ideally 100, trades.
• Be extra careful with filtered results: splitting 100 trades into five groups leaves 20 in each.

Common mistakes

• Changing a strategy after a few losses.
• Trusting a great result from a small sample.

Related