I've experienced a bit of both worlds—I picked up books on Data Structures (including Red Black trees), attempted to implement Rjindael in C++ (without understanding any of the underlying math) in high school, learnt PHP/JS/CSS/C++ and started building from there.
What I've observed is—as a self taught programmer the biggest disadvantage is that you don't know what all is available/standard—as someone else mentioned in the thread, I did not know about parsers until I managed to take a course in Programming Languages.
Since I realized that there was so much I did not know, I've tried to make it a point to keep asking my batch-mates in CS what they are reading, picking up textbooks recommended by professors in class; reading other books as recommended on HN, and elsewhere (SICP, Learn you a Haskell..., started reading TAOCP, etc.) and have attempted to at least match my batch-mates who are CS undergrads. At the very least—try to get a fairly broad idea about known solved and unsolved problems.
The biggest advantages an avg CS undergrad has over me is that s/he has the credentials (which, I've tried to match by having my own open source project, doing interns in well-known companies, etc.), and that s/he is exposed to much more theory than I would ever be without actually having to do much beyond attending classes (for which I have to go the extra mile).
The advantage _I_ have over the other, standard CS guys is that I have the freedom to find out more about what I really want to do, and explore/follow certain areas which I find more interesting without being bound by academia (tests, reports, stupid & pointless assignments), and in general, perhaps—have a lot more fun while learning.