I don't have evidence to back up my assertion - only the theory that in an efficient market riskier assets should be priced appropriately. So I don't know for a fact that the expected return is higher. That said, the fact that the median fund underperforms the S&P doesn't necessarily prove that the expected return on an individual startup investment will do so. For one thing, funds will have overhead - probably significant overhead in startup investing. (Of course, they provide valuable diversification, but as mentioned above, that shouldn't be the priority in this particular case.) Also, even if the median fund underforms before costs, the average might still outperform, if the winners are relatively concentrated.

The application of the EMH to a market as overhyped, information-poor and illiquid a market as startup investment is questionable. And another consequence of competitive markets is that funds with board seats, insider information and ability to negotiate preferential terms based on expected follow-on investment take on less risk than you when investing, but they shape the market prices for the asset.

$20k is friends & family or crowdfunding money, and I suspect that whilst there will be some winners, the median return on those asset classes will be -100%

All good points. And I agree that the median return will likely be -100%. Honestly I think you're probably right that even the average return will probably be underwhelming, even if theoretically it shouldn't be. I am curious though. I certainly wouldn't invest my money in startups on the theory that on average I would end up with more to donate, even if I intended to donate 100% of the proceeds regardless...