Interesting point. I view YC's strategy as playing the central limit theorem -- taking a high-volatility random variable (startup returns) and increasing the sample size until the sample mean tends arbitrarily close to the population mean.

Smart, guys.