Ever considered setting aside say 5% of your funding allocation for wildcard acceptances? Every company that gets rejected from YC could go into a lottery and depending on the absolute size of that 5%, a number of companies could make it in solely because their names were drawn from a hat.
Because the odds are so low that a rejected contender will win a wildcard spot, it removes any unconscious scoring bias you might implement. In other words, I was thinking of the scenario in which you might only make some of the rejects eligible for a wildcard spot based on a scoresheet. You might unconsciously bias your scoring to make some rejects eligible and others not. The beauty of the wildcard is that it's totally random.
EDIT: Another thought. Wildcard entries would set a nice benchmark for merit-based entrants to YC to compete against. They would have more to prove because they've been written off. I'd be curious how this would affect the larger group dynamics of the programme.
There are two phases of acceptance, the application and interviews. The first phase is where most mistakes occur, because written applications are so much lower bandwidth than face to face conversations. So if we were going to insert randomness, we should randomly invite people to interviews, not randomly fund them. But the trouble with that is that, unlike funding, interview slots are constrained. There is a physical limit to the number of interviews we can do. So to invite someone randomly we'd have to not invite someone who was (as far as we could tell) more deserving, and we are very reluctant to do that.
Even if this were true, which I wouldn't argue against, it doesn't mean that it's where the mistakes of the highest aggregate value occur.
For example, what percentage of first phase rejections try again, as compared with second phase rejections?
But the trouble with that is that, unlike funding, interview slots are constrained.
You could apply the usual interview scaling technique of adding another phase (e.g. shorter pre-screen with a single interviewer) or delegating. This has the obvious disadvantages, which is, arguably, one of the major reasons big company hiring processes are so unreliable.
We don't do much formal analysis, but I do regularly look up the applications of people we rejected who later succeed in order to figure out how we missed them.
Because the odds are so low that a rejected contender will win a wildcard spot, it removes any unconscious scoring bias you might implement. In other words, I was thinking of the scenario in which you might only make some of the rejects eligible for a wildcard spot based on a scoresheet. You might unconsciously bias your scoring to make some rejects eligible and others not. The beauty of the wildcard is that it's totally random.
EDIT: Another thought. Wildcard entries would set a nice benchmark for merit-based entrants to YC to compete against. They would have more to prove because they've been written off. I'd be curious how this would affect the larger group dynamics of the programme.