What level of significance for a chi square test would you require for a distribution to fit "good enough"? 90 %? 95 %? 99 % ?
Of course, the larger the better, but one should probably demand a higher significance level the more data one has, right?