New trick on non-flow levels
3 years ago

It seems well-known among the community that on Level 1, hopping at the start of the level consistently saves 1 frame (4.17) compared to simply holding right (4.20). After staring at the source code for 3 days last month, I figured out why this works.

The bug: this comes from the code that handles speed limiting. When not burning flow, if you're not at max speed at the start of a jump, then you can accelerate up to 10% faster than the intended max speed while in the air. (This does not work when burning flow, so I don't expect this to be useful outside of Levels 1, 2, and 17.)

The execution: jumping at the start of a level, or slowing down right before a jump. This GIF shows (afaik) the first recorded 4.13 on Level 1, using this trick: Here the slowing down was achieved by releasing right for 1 frame, before holding up and right the next frame. However, I later realised that tapping down before jumping is a much easier method; you don't have to release right.

The time savings: after learning this trick, it is fairly easy to get 4.10 or 4.13 consistently on Level 1. It is possible, but rare, to get a 4.07. Using this trick, my individual level PBs on Levels 1, 2, and 17 are 4.03, 7.33, and 6.97 respectively. (For reference, WR splits are 4.20, 7.53, and 7.27.) I estimate that I can get each of these with 1% chance, so realistically this trick might give only a half-second time save throughout the entire run.

Edited by the author 3 years ago
SubOptimal likes this
Colorado, USA

I'm a few weeks late in replying, but this is really interesting. I always wondered why jumping at the start saved a frame, but I had no idea that more could be saved. This is the type of thing that makes me really wish there was a TAS, it seems that there would be so many small timesaves using what you found.

Maybe one day someone will actually use this (also all the other new strats) to break the recorded world record.

Game stats
Latest threads
Posted 4 years ago
1 reply
Posted 4 years ago
0 replies