Don't have the book any more, but I believe Kai Lai Chung says in "Elementary Probability with Stochastic Processes" that J.M. Keynes made a similar point - if the roulette wheel is red ten times, then you should suspect that there is a bias, and bet on red for the next.