Why not checking the contributors instead of the forks? Just pick the most regular ones, or the ones that contributed the most (in number of PRs).