I suppose PG really should have said: >the groups of applicants you're looking at have exactly equal distribution of ability. Rather than "roughly equal". Bu

I suppose PG really should have said: >the groups of applicants you're looking at have exactly equal distribution of ability. Rather than "roughly equal". But obviously that makes the whole thing infeasible.