How many trades does a scalping backtest need

[Amanhasnoname]

A strategy that made 40 trades and won 28 looks great. It also tells almost nothing, because the sample is far too small to separate skill from luck.

With a 55 percent true win rate, the observed rate in 40 trades can easily range from 40 to 70 percent. Even a coin flip can produce 65 percent over that many trades.

Scalpers have an advantage here, since they take many trades. A few hundred is a reasonable minimum, and a thousand gives far more confidence.

But number of trades is not the only issue. Trades must also cover different conditions: trending days, ranging days, news days, quiet days. A thousand trades from one calm month is less informative than three hundred across a year.

Beyond the count, look at the distribution. A strategy that earns most of its profit from five outliers is fragile, even with many trades.

How many trades does it take before a backtest result is believed, and what else is checked?