I think this is a good point. Right now YC avoids using social proof as an evaluation tool. They look at the team's ability to work together, that they can ship, and their idea. As YC grows they need to make sure they don't look at who recommends them, how much influence they have, if they're famous.
They can get a %50 success rate, because they have developed a model of evaluating. As the value of getting in to YC increases, the ability to game the model does as well. Every blog post which tells you how to apply, shifts the companies accepted towards those who are able to win at that game.
Every applicant who goes in to that interview knowing it's not a presentation, but a demo followed by 10 minutes of rapid fire questions / brainstorming, has a big advantage. Every applicant who can get current or former YC founders to review their application has an advantage. Every applicant who spends time reading each and every 'applying to YC' blog post, has an advantage.
The big question is, are those advantages things which correlate with being a good startup founder. In some cases yes, in some no.
This is the real signal to noise problem that YC faces. Not that there might be 2000 startups applying each cycle, but that applicants learn how the theater works. They know the game, and are actors which present the perfect YC application. Then YC has to figure out, are these guys and gals really good at what they do, or are they telling me the story I want to hear.
That said, i bet even with the noise, YC has a better way of evaluating the potential future of startups than the 'who else invested, who's the advisors' model many angels use.