create a running average of all chi squares and if its below 1 s.d (or something similar) then run a more exhaustive test