If you repeatedly perform a hypothesis test when the null hypothesis is true, p-values will randomly drift between 0 and 1, and therefore, if you test often enough, you will always find a significant effect. https://lakens.github.io/statistical_inferences/errorcontrol.html#optionalstopping This is why preregistration is important. You can test multiple times as data comes in, but then you need to adjust the alpha level: https://lakens.github.io/statistical_inferences/sequential.html For a single simulation of the p-values you could get, see: