That was not my reading of it. To me, given a large enough pool (~20+), you can statistically assume you'll get a top 10% candidate with this method, at least that is what i read.

Still, this method at least has some structure to it. As opposed to the anecdotes that the rest of the thread is.