How many trades before an edge shows itself

[Amanhasnoname]

The standard error of average R per trade is roughly the standard deviation divided by the square root of the number of trades. With a standard deviation near 1R and a true expectancy of +0.1R, the picture looks like this:

- 100 trades: standard error 0.1R, so a measured average of 0 and one of +0.2 are both ordinary results
- 400 trades: standard error 0.05R, and the true +0.1R sits two standard errors from zero
- 1,600 trades: standard error 0.025R, four standard errors

At 40 trades a day, 400 trades is about 10 trading days and 1,600 is about 40. For someone taking 3 trades a day, 400 trades takes over four months. That is the real advantage of high frequency: the edge can be measured sooner, not that it is bigger.

Several caveats apply. The maths assumes trades are independent and the edge is stable, and neither is quite true. Mixing regimes in a long sample blends different edges together. A clean result also says nothing about whether the edge persists next month, and costs that vary by session muddy the average.

Over how many trades did a real edge stop looking like noise in your own stats, and did it survive the next sample?