Submission #8607: NinjaCocktail's GBA Rayman 3 "100%" in 1:15:19.63

Game Boy Advance
100%
(Submitted: 100%)
(Submitted: 0985 - Rayman 3 - Hoodlum Havoc (U)(RDG).gba USA)
BizHawk 2.8
269946
59.7275005696058
70019
PowerOn
Submitted by NinjaCocktail on 9/17/2023 8:12 PM
Submission Comments
This is an update of my previous 100% TAS of Rayman 3 GBA.
A new glitch in Vertigo Wastes Screen 1 found by the current WR holder Payman resulted in rerouting the backtrack section. The new route saved 26 seconds! A slightly slower RNG seed on Final Boss made me lose 50 frames, oh well good enough. The RTA run would start at frame 867 and end at frame 269609, or a time of 1:14:59.49.
Both sub 1:15 and sub 270000 frames milestones are achieved!
Details about the new improvements:
In Vertigo Wastes Screen 1, it is possible to hit two different purple lums with two consecutive punches. Rayman starts swinging after hitting the first purple lum. When his other fist hit the next purple lums, Rayman starts swinging in the other direction. The result is that Rayman's sprite disappears and the camera starts glitching. Despite being invisible and offscreen, Rayman is still there. He can jump, walk and grab ledges. What he can't do is killing pirates, that's why the first pirate is killed (slower) before performing the glitch.
When the camera is far away, the wooden floor and trampolines are not loaded, as long as they are not refocused by the camera. There is no more collision with Rayman. Thus, he can walk and go through the unloaded wooden floor to collect the 4 lums underneath. Once the camera refocuses on Rayman, the glitch ends, everything turns normal again. If Rayman didn't go out of the pit after the refocus, then the wooden floor will block the way out.
During the glitch, the camera has to be manipulated efficiently to save as many frames as possible. Unfortunately, the behavior of the camera isn't well understood yet. I hit both purple lums and jumped so that the camera goes upwards and to the right. By holding Right, Rayman walked until he reached the 4 lums. Then, the inputs are an optimal mix of Down, Jump, Left and Right to refocus on Rayman while going out of the pit.
The segment can probably be done faster from the purple lums. I tried to hit them from lower heights, but the camera went to the left or refocuses immediately on Rayman. Performing the glitch from the selected height feels pretty optimal, so I'm satisfied with it.
Here, the glitch cost 595 frames (about 10s) until the pole of Vertigo Wastes Screen 1. My best optimized glitch cost 591 frames, but I deoptimized the segment by 4 frames in order to find good RNG seeds for both Jano and the Final Boss. After Vertigo Wastes Screen 2, I saved 3 frames thanks to the framerule.
The transitions of Void of Bones are unchanged. Jano is as fast as the previous TAS. Nothing changed until I reached the Backtrack section. I directly entered Prickly Passage. Skipping the backtrack of Vertigo Wastes saved 2145 frames (about 36s). I lost 2 frames because of the framerule after Prickly Passage.
The rest of the TAS hasn't changed until the Final Boss, I entered the curtain 1551 frames sooner. About the seed I selected:
  • I lost 49 frames during Phase 1 because I got one more set of beams.
  • Phase 2 is the exact same as my previous TAS. However, I had to wait 1 more frame to be able to skip the final cutscene.
So this seed is 50 frames slower compared to my previous 100% TAS. If I compare this seed to all the seeds I tested (Any% and 100% combined), this one is the 10th fastest out of 43 seeds. It is 129 frames slower than my best seed ever, 280 frames faster than my worst seed ever.
Those improvements shortened the previous 100% TAS by 1501 frames or 25.13s! Enough to achieve both sub 1:15 and sub 270000 frames milestones! The final minute barrier has been broken considering the current knowledge of the 100% category.
I hope you'll enjoy this updated TAS. Regards, NinjaCocktail

nymx: Claiming for judging.
nymx: Congratulations on the improvements. Everything looks good here.
Accepting.

despoa: Processing...
Last Edited by despoa on 9/28/2023 1:17 AM
Page History Latest diff List referrers