Hi Maximus! The technical details are here:
https://docs.growthbook.io/statistics/power#growthbook-implementation
The short answer, is that we take the
mean and uncertainty in your metric (e.g. the usual conversion rate for a 0/1 metric, or the mean for a continuous metric like revenue), and your estimated sample size, and then use the formulae above to estimate how often you can observe your stipulated
effect size with that big of a sample.