Earlier quoted context omitted.
Or his model is wrong. So many folks assume, since they succeeded, they are responsible for that success. Market conditions, accidental sales, and Brownian motion may in fact be more important. Even considering YC companies, who's to say how many would have succeeded anyway? There's no control, so no experiment.
Even considering YC companies, who's to say how many would have succeeded anyway? There's no control, so no experiment. I guess you could look at rejected applicants as a sort of control group (not perfect though)? How many of those have gone on to be a success (I have no idea)?
Assuming that PG is not a complete idiot and can identify obvious skill and/or obvious product home runs, thus removing them from the "rejected" list, your control group isn't very accurate.