[bramj]

If you are interested in testing whether the zero comes up more often than a different numer you shouldn't test on the variance. Let P7 be the probability of a 7 in a random round of roulette. Than you have zero hypothesis H_0: P7 = 1/37 versus H1: P7 > 1/37. This can be tested with a standard test on sample average. This would make more sense than testing the variance.

The latter would be possible however. In this case you need the variance of a uniform distribution on 0,...,n. I'm suprised you didn't google it up, but it is (n^2+2n)/12. Now you could do a ratio test o something, out of my head, I suppose you should use an $F$ statistic.

However, don't bother. Your roulette game is fair.