Pick a game, and check in which revision it works the fastest, provided proper tinkering with settings. Commonly, revisions around rev1000 seem to be claimed to perform faster than the latest ones, but obviously the difference in compatibility would be huge. Quite the tradeoff.
I used to keep track of the speeds for SSBM for many revisions, but I lost track of it all around the 300s or so.
The best way to figure it out without playing every game on every version would probably be to look at the svn logs and see what changes were made in each revision. If the change is a speedup, it is probably comparatively fast, but if it's a compatability fix, it's probably slower.