How many trades before a backtest means anything?

There is no single number, and the useful question is a different one: how wide are the error bars around what you just measured?

Why a fixed answer does not exist

The honest answer depends on how variable your results are. A strategy whose trades cluster tightly around a small average needs far fewer of them to establish that the average is real than one whose result is carried by a handful of outsized winners.

That second case is common and badly served by rules of thumb. If removing your best few trades turns the strategy negative, the sample is describing those trades rather than the strategy, and no amount of additional testing on the same history fixes it.

The test that costs nothing

Sort the trades by profit. Remove the top few. Look at what is left. Then do the same with the worst few and see whether the strategy becomes wonderful. A result that swings from good to bad on a handful of trades either way is a result about those trades.

This is cruder than a confidence interval and it takes a minute, which is why it gets done. Most strategies that fail live fail this first.

Small samples are not useless, they are just not conclusions

A short sample can rule things out. If a strategy is already unprofitable across forty trades after costs, more data is unlikely to rescue it and you have saved yourself a week. What a small sample cannot do is confirm an edge, because the range of outcomes consistent with random noise is wide enough to contain a good one.

Treat a thin result as a reason to keep testing rather than a reason to trade or to abandon.

More trades is not the same as more data

Dropping to a lower timeframe multiplies the trade count without adding independent information: the same week of market behaviour is being sampled more finely, and costs rise with every extra trade. A thousand trades from one quarter tells you about one quarter.

What actually widens the evidence is more distinct market conditions: a trending stretch, a ranging stretch, and at least one shock.

Questions

Is thirty trades enough to evaluate a strategy?

Rarely. The range of outcomes consistent with pure chance across thirty trades is wide enough to contain both a real edge and nothing at all, so a positive result at that size is not evidence either way.

Does a higher win rate mean I need fewer trades?

Not on its own. What matters is how variable the outcomes are relative to the average. A high win rate with occasional very large losses can need a larger sample than a lower win rate with consistent, similar-sized results.

Can I get more trades by testing more instruments?

It helps more than dropping to a lower timeframe, because different instruments are closer to independent. Be aware that correlated instruments are not independent: the major dollar pairs largely tell you the same thing at the same time.

Sonar Sciences is a research and validation platform, not an investment adviser. Nothing on this page is investment advice or a recommendation to trade any instrument or strategy. Platform names are trademarks of their respective owners and we are not affiliated with them. Trading leveraged instruments involves substantial risk of loss.