My apologies. Yes that's correct. My mistake was reducing the set of choices to n + (|H| mod n). Your worst-case estimate of a total bias of n/|H| seems to be c

My apologies. Yes that's correct. My mistake was reducing the set of choices to n + (|H| mod n). Your worst-case estimate of a total bias of n/|H| seems to be correct. I guess the question now becomes whether that bias is acceptable for the purpose of the lottery (I would say probably yes). Thanks for sticking with this thread :)