The math of "what to do if startup valuations vary multiple orders of magnitudes less than startup outcomes" is very easy.
If "what to do" was "very easy", everybody would do "it". But that doesn't happen. Investors are not all created equal, and investor returns show massive variance (ie, they are not observed to be uniform.) So something is very wrong with your approach/level of abstraction.[1]
You might want to also read this comment from paulsutter that PG responded to, if you think this is a personal attack on you.
http://news.ycombinator.com/item?id=4959964
______
This point is so obvious that I find it hard to believe you wrote the above comment in good faith.