To remove the effects of this inequality, we randomly sample in every dataset the exact same number of bets from heavy gamblers.

Check out my blog: here