Notes on progress updoots: AI Rewrite

I starting working on AI again. Starting specifically with rewriting Gejhi's AI. His previous AI was rushed because I wanted to get the character out for folks to finally play with, but after some patches I made, it also seemed to weirdly effect his behavior. 

SO, I'm rewriting his AI very carefully this time, and working out the kinks and weird glitches I overlooked and already fixed a LOT of the basic issues with the AI. I also have new ideas on how I'm going to implement the AI's combos with few tricks to make the AI do well put-together combos rather than leaning so heavily on the RNG to handle that.

I actually fixed the parry behavior this time. So even at max difficulty, he wont be an annoying parry god anymore (but I'm going to put in a flag to still let him parry god other AI, lol, it was something I intended to do a long while ago, but never got around to). So far I just have a much less frustration set of math for it, and I also restricted how often it can try for parries, and limited it to certain states. but I also plan to focus this into certain "modes" or special behavior sets where he's specifically waiting fishing for a parry and the likely hood for this behavior will be influenced by his well-being in the match. Also will be tossing in a flag for when attacks/actions finish, he'll try to parry a punish attempt if he recovers fast enough, of course, this will scale with the difficulty. 

With that said. I'm also making him respond better to throw attempts. He can tech throw attempts farely often now, even when throwing them in with a reset, which I noticed lots of AI seem to struggle with. I have Gejhi calling out my throw attempts in mix-ups and I nerfed the math a bit to not have it be aggravating. I will make this more complex as well to feel more natural.

SIMILARLY. I'll be making sure the AI also knows how to deal with command grabs. Mainly by avoiding the situation as the baseline by dashing away or jumping, or using throw-invincible moves in common situations where frame-1 grabs are a threat and also recognize when there's a grabbing attack he has a few frames to respond to. Command grabs are the destroyer of a lot of AI, and I've seen a lot of other AI with command grabs do that... a LOT.

Somewhere along my progress with this. I'm actually gonna put that "brainplace" helper I made a while ago (but never used) to use. It's mostly just gonna be there to alter the AI's behavior with variable flags to make him do certain things more or less often with a soft randomizer or alter the frequencies to avoid the AI being too strongly patterned or conditioned. That concept of having each palette behave differently will be more obviously presented in this form, where all the variables have different base values.

I'm rather happy with the progress and issues I've ironed out so far for it. Fixing all the weird movement blocking. I'll post up some progress on that soon once at least have most of my starter/neutral actions and combo trees set back up.