[TAS] TMNF A07-Race | 26.10 | Thanks Bice!
Here we go again, A07.
I have never been able to improve trabadia's start by myself, so even my 26.12 was built upon trabadia's 26.17 start.
But recently, Bice decided to try A07 from scratch. He managed to improve from the very start and basically did everything even or better, and he got the .11 entirely by himself. Incredible job! Here is his video in case you missed it: • [TAS] (Outdated (.06)) Almost .10! | A07-R...
But Bice sadly couldn't get .10, so I gave it a try, using his superior start.
The dropdown was the 1st place where I thought I had a chance to gain a bit, so I went and cleared the jittery inputs and did some manual tweaking. I ended up ever so slightly ahead in the slalom, but it was unclear if the faster way to the uphill was degrading the uphill entry, so I kept manually adjusting until I was pretty confident I was ahead and in condition to get hopefully a similar uphill.
Then I started working on the uphill. And in the process, I forgot to bruteforce the dropdown. Woops. The upside is that the drop down inputs are not jittery at all (there are only 2 steer commands between 15.07 and 18.19, perfect for a fresh bruteforce LOL)
Back to the uphill, I tried bruteforcing with custom scripts, changing every parameter that could make sense: min inputs time, max inputs time, modify prob, evaluation time, evaluation goal (distance, distance/velocity balance)... At some point, I saw the bruteforce finding a big gain over the base run, I checked it and YEP bf magic, new god uphill.
This uphill is very wide, but the speed kept is more than enough to compensate the extra distance traveled, and reach the finish sooner. It is so much better that after the bruteforce found this uphill, I only had to manually change 1 steer input timing post-uphill, and it was instantly 26.10. It was a 26.1079, to be precise.
Last straight tweaking and precise finish bruteforcing weren't enough for the 26.09, obviously. We'll see if I'll find the patience to try to get a better uphill or finally bruteforcing the dropdown and re-doing entirely the uphill.
But yes, thanks to Bice's improved start, 26.09 looks very possible. I never thought I would say it, but A07's limit is not 26.1x. TAS can reach 26.0x. It's time to screenshot the Youtube comments that were joking about it in the previous A07 TAS videos.