A rough hint arrives around 30 trades, a believable read around 100, and real confidence when the sample also spans different market conditions. The number alone is not the point - a hundred trades from one calm month prove far less than a hundred spread across trends, ranges, and volatility.
Why small samples lie
Trading outcomes are noisy. Flip to a run of three lucky winners and a 15-trade test shows a dazzling win rate that has nothing to do with your edge. The fewer trades you have, the more a single outlier dominates the averages. This is why a beautiful result over 12 trades should excite you far less than a merely decent result over 300.
Conditions matter more than the count
A strategy can look brilliant in a trend and fall apart in a range. If your whole sample comes from one regime, you have not tested the strategy - you have tested the market's mood that month. Aim to include trending, ranging, and high-volatility periods, plus at least one genuinely bad stretch. That mix is what separates a robust edge from a lucky fit, and it is the backbone of a proper backtest.
Important: the sample must contain pain. If your backtest never shows an uncomfortable drawdown or a real losing streak, you have not stress-tested the strategy or yourself. A test with no bad periods tells you nothing about survival.
Match sample size to your trading style
- Scalpers take many trades, so 100+ arrives quickly and you should aim higher, 300 to 500.
- Day traders can gather 100 to 200 trades over a few weeks of replay.
- Swing traders take fewer trades, so reaching 100 means testing a longer span of history - do not shortcut it just because trades are scarce.
Sample size and the metrics
Every metric gets more trustworthy with more trades. Expectancy, profit factor, and win rate all stabilize as the count grows and swing wildly when it is small. Treat any figure from under 30 trades as a rumor, not a fact.
Gather the sample without waiting months
Live trading produces trades slowly. Backtesting compresses time: by replaying historical charts in a simulator, you can take a hundred trades in an afternoon and jump across different years and conditions to build a genuinely mixed sample. That is how you reach a trustworthy count quickly, without risking a cent to find out whether the edge is real.
Sample size FAQ
How many trades should I backtest before trusting a strategy?
About 30 for a hint, 100+ for a serious read, and more when spread across trending, ranging, and volatile conditions with at least one bad stretch.
Is 20 trades enough to judge a strategy?
No. Twenty trades can expose execution problems but is too small to trust for expectancy or win rate, because one outlier can dominate.
Why does sample size matter so much?
Trading results are noisy. Small samples can show a great edge by luck; larger samples average out randomness so the numbers reflect reality.