I would suggest that you generate an empirical distribution for some statistic of the seperation of random balls by generating a large Monte-Carlo sample (say the mean square seperation of random balls from the 23 in the sphere), and then compare with the value for the actial distribution.

CB