The thing with "built from scratch" games is that the people making them don't have time to refine the high-level stuff like AI. That's because for most of the project, you can't even play the game (getting playfields up, getting renderer working, implementing animations, etc.). Until those things are working properly, it's really not possible to have any sort of AI. I'd be very impressed if either MVP or Inside Pitch could pull off the "complete package" in their first year.

Online baseball is different from other online sports because of the serious consequences of latency. The two normal methods of implementing online suport don't really work for bsaeball. Synchronization is used in games like NFL and NBA, where only 2 players are in the game at once. This method essentially just passes control inputs back and forth, forcing both games to remain identical on both screens all the time. This method was used in WSB2K2; the problem there was that the time it took to sync every control input caused a split-second delay that made it extremely difficult to time pitches and nearly impossible to field. The master/slave method is the one used by games like Quake, where synchronizing controls for up to 64 players would require an impossibly huge amount of data transfer. This method allows the games to be slightly different on different players' screens. Basically, the server assumes the last input is the current input until it hears otherwise from the client. Then when it hears something different from what it assumed, it "backs up" and puts things how they should be. If this system were used for baseball, you'd see some strange things. You might get a strikeout on your screen, when in actuality the other guy hit the ball (and your game hadn't gotten that data yet). You'd see guys tagged out on your screen, when in fact the guy had turned his baserunner around and gone back safely. As you can see, online baseball isn't easy. It'll be interesting to see how well ASB can do with it.