I don't think its unreasonable to be able to pick 80% winner startups, as long as you're willing to wait long enough. As an employee, you only need one company

I don't think its unreasonable to be able to pick 80% winner startups, as long as you're willing to wait long enough. As an employee, you only need one company every four years in your "portfolio." If you narrow to startups that have hit, say 50MM active users before employee #50, you might get to 50/50 success rate. If you analyze potential market size, red flags, comparable companies, and talk to three employees to get a sense of the internal workings of the company, it seems pretty reasonable that you can weed out most of the companies who are capped on their potential market, and the engineering or other disasters waiting to happen. Incidentally, I think friendster didn't hit 50MM actives. The semi-fails of the rocketship class are Zynga and Groupon, both of which still have market caps logarithmically closer to 10B than $500MM. Foursquare might be both a rocketship and a fail, but I'm not quite sure. Who else grew like mad in the last decade? I guess my core point is that you can tell really, really early whether something has very strong network effects, and if it does, the arbitrage is currently very easy.