http://bulbapedia.bulbagarden.net/wiki/Pok%C3%A9mon_data_structure_in_the_GBC
Assuming Stadium doesn't gut the storage format from GB, try searching for the following consecutive byte string, which would represent R2 Rival's Mewtwo, and see if other 48-byte strings nearby might represent other Pokemon. I'm not sure, but the first set of ??s might be 07 D0 in case you can't do a consecutive-byte search with wild cards in the middle.
96 52 5E 42 55 69 ?? ?? 13 12 D0 FF FF FF FF FF FF FF 0A 19 0F 14 ?? ?? ?? ?? 64 ?? ?? 01 9F 01 9F 01 5E 01 16 01 66 01 96 01 16
As time goes on, though, I'm becoming less convinced that a Stadium TAS could ever be interesting enough. In low level cups, it's quite possible to get a
perfect cup (all OHKOs, opponent never takes any action) unassisted and with minimal luck manipulation, while still using a diverse set of Pokemon. All a TAS can bring to the table for that is streamlining the attack selection process down to the absolute fastest, and possibly choosing the attacks with the shortest possible animations that still get in for lethal.
However, once you get into Poke and Prime (and I guess GLC by extension), too many Pokemon are resilient enough to withstand any standard damage-dealing attack, and you can't get any significant edge attempting to out-level them. Therefore the only viable strategy for a speed-oriented run would be to blast moves like Horn Drill through most opponents, and by the time you get to using something with a Quick Claw and Fissure and Horn Drill, it's like you're not even playing Pokemon any more, you're just playing an entirely predictable solo target practice game that inserts 30-second time delays after each stab for no real reason and that just looks a lot slower than it ought to be in a game like that, not to mention drives most of the audience to sleep or to watching something else by the time they get around to the 96th turn of that crazy little game.
