[TAS] Final Dreadnought 4 - Carol in 1:22.20

Here it is finally! The first sketch of Final Dreadnought 4 as Carol.

The first mention on this commentary is to IceColdSight, who figured the new dreadblocks 3.0 skip. Dreadblocks was already a known zip, double subpixel based and pretty much IL only, but he figured that using the brute force and pouncing all the way it skips entirely the laser room instead of being softlocked as we thought in the past, no one else tried that hard up there until he did, nice job man. Pounce TAS speed abuse was the main reason that motivated me to start this work, 30 presses per second.

The second mention is to Avikaielef, current WR holder in the stage. After starting the work I got lazy about the first phase in the boss fight, I gave him the input file and he made the model on this run. Without his help I'd probably never finish the work I started.

About the level run:

The first part is getting the dreadblocks zip at 4 seconds, the earliest possible cycle. The frequency of the cycle is exactly 4 seconds, the first cycle begins with the level. At this moment only once I had that cycle in a level run, normally the cycle we use in our runs is at 8 seconds with any character, but Carol is surprisingly the character with better chances of getting the earlier cycle, also confirmed for Milla on TAS and not yet for Lilac.

Once in the wall, taking the earth shield is a strategy in order to not needing to wait for the DNA cannon at the end of the screen, then the boss fight model was built around the fact Carol had still one hit in the shield. Once in the top, in the very first frame possible is where TAS shines more, with a total of 625 pounces in 1250 frames to escape from the wall and then the way to the boss fight is simple.

About Brevon 1 or Absolution 2: There were a lot of possible strats and comparisons, Carol on foot not having an upwards movement is not able to get a one cycle kill. Maybe with some luck she would be able to get a 1.5 cycle, which means killing absolution in the rush between cycles 1 and 2, testing without a new idea that came up recently she was 5 frames away to make it , but at which cost?
There is a particular part in the first cycle that is very inconvenient, when Absolution rises and starts flying very slowly to the left before finishing the first cycle. It's really worth trying to deal one or two hits more if he will be so much not being hit in the height? The answer we arrived after testing some models is that if the first cycle is finishing anyways, making it shorter is the best choice.
The first fight model is not 100% confirmed as the fastest possible option, it's just the best we figured in the time we worked. I tried an alternative model but actually Avi's model was being faster by 10 frames so I decided to just leave it exactly like he made it.

Brevon 2 was a real headache, conceptually super easy but in the moment of scripting the fight an unexpected enemy appeared: RNG.
There is a long story about the RNG on this boss fight, but to make it short the RNG was removed and the pattern is always the same since the version 1.21.5 of the game. But even if the pattern RNG was removed there is another kind of RNG I call "micro-RNG", and it consists on the time between actions, sometimes Mecha-Brevon does his movements with up to 5 frames difference, which in runs is probably not noticed at all but TASing was being a really huge problem.
Another worth mention is to the creator of the trainer and the TAS tool, Fladervy, because he made a way to calculate and lock RNG seeds. That was really useful to make this fight work after a lot of issues.
The first hit in the fight is more a swag hit than a faster hit, indeed it makes the fight a few frames faster but not a huge difference.
Every phase in Brevon 2 has 10 health points except the first one, a wild claw is good enough for every phase since the second, but is truly important the hitbox. Looking closely is noticeable how sometimes seems like Brevon is not being hit in the first possible frame and that is because the way wild claws work, earlier wild claws are not finishing the phase, waiting some frames for a single good claw is faster than an earlier claw and a second claw chained.

Brevon 3 was actually pretty easy, the only problem was synching with the file but it was a really minor issue comparing with Brevon 2.
I tested a few different models, being this one the fastest (6.18 seconds total). It's based on wild claws abuse, not all of them with entire hitbox and based also in Brevon's movement, trying to be more efficient even in the moments not hitting him. Parrying attacks is key on this fight, especially when Brevon does the grab. Being grabbed means some frames not being able to attack, the parry there works in a specific position and allows Carol to keep fighting during those frames and secure an earlier kill.