[webmaven]

Even if I accept your model of reality, there are some pretty big problems that a grant could address:

1) the categories of 'brilliant' and 'delusional' aren't mutually exclusive, especially since both are spectra rather than binary. They aren't entirely orthogonal, however.

2) 'brilliant' and 'delusional' are each qualities that are very hard to evaluate except in hindsight.

3) Finally, it is possible for someone who is brilliant and non-delusional to still fail (or be 'insufficiently driven' and give up rather than dying in poverty), or to succeed with no-one noticing (because they lack resources or skills for self-promotion).

There is no way to reliably sort the wheat from the chaff, except to give them space and time to succeed or fail.

[nickff]

I agree with you on all three points; I was trying to point out that there's an adverse selection problem that the post doesn't seem to take into account. Even most VCs have a very difficult time making money by finding brilliant visionaries, and they have a number of factors working in their favor.

[webmaven]

Right. Most attempts to select winners end up filtering out the high-risk candidates that also represent the highest potential impact should they succeed (which also makes applicants fudge their proposals to seem lower risk high reward). So, you need to consciously allow the selection of some candidates randomly from a pool that only filters out the obviously deranged or fraudulent (in other words, quality of proposal counts, perceived odds of success does not).

There are a number of research funding agencies that are starting to use just this sort of grant funding lottery, the OP seems to be groping toward something similar.