One of the astute comments on this by Terence Lam: Terence Lam
Stage one: invest $25K in 100 startups ($2.5M)
Stage two: pick 20 of them and invest $100K in them ($2M)
Stage three: pick 4 of the 20 and invest $500K ($2M)
Stage four: pick 1 of the 4 and invest $5M in that ($5M)
the "expected" exit value of a startup is different in different stage, and most often this is non-linear: the expected exit value increases quadratically when a startup move from stage one to stage four.
$11.5M invested ;) (ignoring the management fee :p)
Another hidden benefit: startups in stage three or four may "acqu-hire" startups in stage one or two, and the VCs who know and engage with them are the good middlemen here to facilitate the deals.