noone’s hating. he came here for advice.
the CPU can be extremely predictable, so even while you think you’re consciously anti-airing on reaction, you’re actually sub-consciously figuring out the patterns and limiting your own defensive options when you reckon the probability of a jump is high. the AI does not jump randomly. all it’s moves are based on built-in scripts or something like that. eg. if you retreat beyond a certain distance against a projectile user, they start throwing fireballs. if you do a certain move within a certain range, it increases the probability that they will jump (or react with a certain move). hell, even just moving into a certain area increases that probability that certain AI characters will jump. furthermore, you don’t always want to react with an anti-air to a jump. what if a human player does a magic jump (one that you can only hit if they press a button in the air), or they do a move that changes their trajectory?
trying to throw the AI whenever you get close to it is an exercise in futility. i tested the AI in vanilla, and basically you only get 1 or 2 free throws per round. it will tech everything after that. and besides, why would you always go for a throw when you get close to the opponent? you’re just screaming to get punished.
if you want to train your whiff punishment, set the dummy to walk back and forth and to poke randomly, while maintaining forward movement (meaning that if you don’t poke it back or whiff punish it will push you to the corner). once you start punishing it’s whiffs successfully, record another setup with different timing (maybe even with a different char). repeat until you’re confident you’re reacting to the move and not the pattern.