I don't doubt that most competent self-taught hackers could learn to do CS research if they were interested. A passionate interest in the learning new things, without waiting for someone else to teach you, is fundamental to teaching yourself to hack, and that same attitude is the main component of doing research. imho the biggest difference between hacking and doing 'research' is the mental shift from building something to creating new knowledge, whether it is a way of solving a problem, or something more abstract.

The most valuable things I've gotten out of the formal process has been the built-in mentorship provided by a good supervisor, the community of fellow students, and the opportunity to work on something with no immediate economic value. If you have the time/money, and don't mind harassing strangers via email you can get some of that without going to school for it. Still, if that mentorship, community, and money wasn't valuable, no one would be applying to ycombinator.