A backtest with 40 trades and a beautiful equity curve is one of the most convincing things in trading, and also one of the least reliable. The question of how many trades are enough doesn't have one exact answer, but there are a few ways to get a feel for whether the sample is telling you anything.
Look at how much one or two trades contribute. If removing the single best trade turns a profitable test into a flat or losing one, the result is resting on an outlier, not on an edge.
Split the sample in half by time. If the first half and the second half look completely different, the combined result is an average of two unrelated periods rather than a stable pattern.
Do a rough confidence check on the win rate. With 40 trades and a 55% win rate, the uncertainty is wide enough that the true rate could plausibly be anywhere from the mid-40s to the mid-60s. That range includes both a solid strategy and a losing one, depending on the reward-to-risk ratio.
Count the market conditions covered. A thousand trades from one quiet three-month period may be less informative than 200 trades spread across trending, ranging and high-volatility months.
As a loose rule of thumb, many people want at least a few hundred trades across different conditions before trusting the average, and more if the strategy has a low win rate with occasional large winners, because those results are naturally lumpier.
What minimum sample do others use before taking a backtest seriously?
How to tell whether a backtest has too few trades to mean anything
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm