I think he's saying that, because many first time entrepreneurs would happily take a $XX million offer to sell, most will sell early rather than holding out for a chance to be the next Dropbox. So even if the potential/expected value to be a billion dollar company is there, factors that are harder to calculate might ultimately determine exit value.
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.
