alanback: I've just tested Wetware's data against a Binomial distribution with n=137, p=1/18.

I think getting 38 double matches is more than 11 standard deviations away from the expected number!

#sd's = [38-np]/sqrt[np(1-p)] = 11.33