LEARN · EN · easyquanttrading.com
How many trades do you need to trust a backtest?
Most traders look at total return and ignore the trade count — that is the mistake that makes a backtest meaningless. With too few trades, any result is noise. This page explains why the number of trades is the first thing to check, what a defensible minimum looks like, and how statistical confidence scales with sample size.
- A result built on 20 or 30 trades is mostly luck — the confidence interval is enormous
- Rough rule of thumb: you want at least 100 trades, and 300+ to start trusting the edge
- Statistical significance is a function of both trade count and effect size, not return alone
- A high Sharpe on 25 trades is worth far less than a modest Sharpe on 300
- Check expectancy and its standard error, not just total profit
- Out-of-sample and walk-forward tests multiply your effective sample, not just your confidence
Why trade count decides whether a result is real
A backtest is a sample of what a strategy might do. The smaller the sample, the wider the range of outcomes that are still consistent with pure chance. With 20 trades, a profitable result tells you almost nothing: a coin-flip strategy with a small edge could easily look great or terrible over 20 flips.
The trade count — not the percentage return — is what determines how much you can conclude. A 300% return over 15 trades is less trustworthy than a 40% return over 300 trades.
The rule of thumb: 100 to 300 trades
A common floor is around 100 trades before a result deserves any real attention, and 300 or more before you should trust the edge. Below 100, the standard error of your average trade is still large enough that the sign of your expectancy can flip.
This is a guideline, not a law. A strategy with a very large, consistent edge needs fewer trades to be convincing; a strategy with a thin edge needs many more. The point is that you cannot look at return alone — you must ask how many trades produced it.
What matters more than the raw count
Two metrics are more informative than trade count by itself. Expectancy — the average profit per trade — and its standard error. If the average trade wins 0.3R with a standard error of 0.05R, the edge is clear. If it wins 0.3R with a standard error of 0.3R, you know nothing.
You should also ask whether the trades are independent. A strategy that places 300 nearly identical trades in one trending month is less convincing than 300 trades spread across regimes — bull, bear, and sideways.
How to get a trustworthy sample
First, count the trades honestly: exclude overlapping or duplicated signals. Second, split the data — keep a held-out period the optimizer never touched. Third, use walk-forward testing so the out-of-sample trades come from many independent windows, not one lucky stretch.
Platforms that report trade count, expectancy, and out-of-sample windows per strategy make this check fast. You should be able to see, at a glance, whether a candidate was built on 30 trades or 300 before you spend another minute on it.
FAQ
- How many trades is enough for a backtest?
- At least 100 as a floor, and 300 or more to trust a thin edge. Below 100, the result is mostly noise.
- Is 30 trades enough to evaluate a strategy?
- No. Over 30 trades the confidence interval is so wide that a profitable result can easily be luck.
- Does a high win rate make up for few trades?
- No. Win rate and trade count answer different questions. A 90% win rate over 10 trades is still statistically weak.
- What is a good number of trades for statistical significance?
- There is no single magic number, but 300+ independent trades across different market conditions is a defensible target for most strategies.
- How do I get more trades for testing?
- Use more history, smaller timeframes, or walk-forward windows that generate independent out-of-sample trades.
More guides
Not investment advice. Historical results do not guarantee future performance. EasyQuant is a research factory — you execute on accounts you control.