Sample size in gold trading

Sample size is how many observations, here trades, a result is based on. The smaller the sample, the more a win rate, profit factor or expectancy can differ from the strategy’s true long-run value purely by chance.

How it applies to gold trading

A gold strategy that wins 12 of 20 trades shows a 60% win rate, but its true rate could plausibly be anywhere from about 40% to 80%. With 200 trades at the same 60%, the plausible range narrows to roughly 53% to 67%. The same applies to expectancy and profit factor.

Market regimes matter too. Gold behaves differently in trending years, ranging years and around major Fed cycles. A large sample from one regime can still mislead, so good tests include different years and volatility conditions. Most traders aim for at least 100 trades, and preferably several hundred, before trusting a statistic.

Worked example

Example (hypothetical numbers): two gold setups. A: 15 trades, 10 wins, average +1.2R. B: 180 trades, 82 wins, average +0.25R. A looks better but rests on 15 trades and could easily be luck. B’s smaller edge is far more likely to be real.

Journal statistics follow the same logic. Twenty live trades say little about a strategy, but they can reveal execution problems, such as late entries or moved stops, that do not need a large sample to spot.

A common mistake

Changing rules after every few losses. That resets the sample each time, so you never learn whether any version works.

See also: Backtest, Forward test, Win rate, Curve fitting.

Frequently asked questions