this isn't a perfect metric, it over-focuses on attacks, probably has some noise because a button press doesnt necessarily cause any given state change, and the average probably gets raised by stuff like mashing FD, i'm surprised at how clean it seems to be for characters with a lot of match samples
