Bootstrapped confidence intervals around expectancy

[Amanhasnoname]

An average profit per trade of 0.3 pips sounds positive. Whether it is distinguishable from zero depends on the variance and the number of trades. Bootstrapping provides an honest answer.

Resample the trade results with replacement, compute the mean for each sample, and repeat thousands of times. The spread of these means forms a confidence interval. If the interval includes zero or negative values, the edge is not established.

For scalping strategies with wide variability in outcomes, intervals are often wider than expected. A thousand trades might yield an interval of plus or minus 0.2 pips around the mean, enough to swamp a small edge.

The analysis also shows how many trades are needed to reach a desired level of confidence. Doubling the sample reduces the interval by about 30 percent.

It supports decisions such as whether to increase size, whether to continue a test, or when to stop a failing system.

Care is needed with clustering and serial dependence, which make the effective sample smaller than the trade count.

What does the confidence interval around the average trade look like for the system being run?