Hmm. This seems like one of those things that may be statistically true, but the statistic may not be relevant to a given situation. I think there's a term for stuff like this but I forgot it, regression to the mean, maybe?