It's also a way to test new investment strategies as well. Basically, with what amounts to play money for them, they can test out new ways of evaluating teams and startups at an earlier phase and see what works and doesn't. They have a good idea how to evaluate companies that already have users and customers, but no idea when it comes to raw, unproven ideas. The value of that information when applied to their main batches could be pretty big... if it makes them catch one good startup they would have missed otherwise, it will have already proven a nice return.
I doubt many if any VCs see as many applicants as YC, and they're finding every way to leverage that data advantage they can against the competition