Of course, most people who think they are two standard deviations above the mean in expected returns, are not in fact that much out of the ordinary (monetarily,

Of course, most people who think they are two standard deviations above the mean in expected returns, are not in fact that much out of the ordinary (monetarily, anyway). So the a priori expected return also needs to adjust for the risk that your self-assessment is incorrect. You are making a common mistake inherent in backwards-looking risk assessment, considering the Bill Gates example after Bill Gates is already massively wealthy. Prior to Bill Gates founding Microsoft, or even in its early years, his main advantages were in coming from a wealthy family, and few people bet on his company getting exceptionally huge (as evidenced by the early Microsoft market cap, which was not exactly tens of billions). I would argue the ultimate outcome there, whether Bill Gates "succeeds" into $50b or "fails" into a mere $10m, is closer to rolling dice, and not reliably predictable. In any case, Bill Gates was a trust-fund kid with a multi-million inheritance backstopping him, so is excluded from the field of consideration I'm positing here, which is what decision is best for kids who come from less-wealthy families.