I think that's generally true but I've noticed that there are a couple of people in every cohort who have the right stuff. They're good and everyone knows it, including themselves.

Glad this post was revived after being dead for some reason.

That's true if the cohort is MIT grad students. It's probably not true if the cohort is Random Western State Institute of Technology. On the other hand the relatively high percentage of postdocs who survive according to the cited article probably reflects that in the analogy many of the hiring institutions are closer to second division clubs than to first division ones. The analogy breaks at this point because salary structures are quite different. The marginal difference between working at Research I and Research II is not so great, while the marginal salary difference between playing in Division I and Division II is huge.