(Thanks for the console verification, TiKevin83!!!!!)
Anonymous here. For once, I don't have too long of a submission note since this builds on two previous submissions. For more info on the ACE glitch, please read the
previous submission.
This run is 10 frames faster than the previous run, and syncs on console!
I noticed an improvement in the deck edit; any time we exit a category (which we do 7 times), the "B" input can be done on the same frame as the last left or right input. So, this saves 7 input frames. I am a bit embarrassed to have not noticed this error the first time, but at least it gives a reason to resubmit on the new console-accurate BizHawk, as well as apply my lag reduction brute force technique to the run.
Like the last run, we have 3 frames to kill for advancing RNG. The time we do these, as well as the tile we pause on, will change the number of cycles executed, which can make the credits warp succeed or (99.9+% of the time) fail.
I had to wait one extra input frame to have the warp succeed (possibly due to the shorter edit, or due to the console accuracy). I tested absolutely all possible no-extra-input-frames options and NONE of them activate the credits. Literally. All. Possible. I think.
Building off the brute forcer I made for the
glitch'less' run, I took every possible combination of 3 RNG advancement frames (around 8500), every possible tile you can pause the game on plus every path you can take going to Aaron, and a list of possible spots to insert one extra (so, any spot where it won’t increase RNG: pre-tutorial-exit, the Settings menu, the deck edit, and in the duel).
So, I had three vectors to modify the run in search of the successful credits warp with the least lag: 8436 permutations of the 3 RNG waits; 56 movement variants, in 2 groups (26 pause steps, 30 extra variations of paths); and 86 spots I could wait 1 additional frame, in 3 groups (8 mid-duel frames, 37 earlier options post-tutorial-exit, 41 early options pre-tutorial-exit).
Well, I can’t test all 40,627,776 options, but I ran through many variations of in-depth-ness of these options over the past month. I made the later groups later on, so I was more thorough with the first groups. I tried ALL variants of “no extra waits”, with no results. Group 1 of steps + group 1 of waits (mid-duel waits) got an earliest end of 10585. Once I moved to settings/deck edit waits, it quickly found a 10582 input end, which was my goal (my base movie had a 10582 end, without a successful credits warp). Not expecting any better results later on (early = more lag potential!), I moved on to the even earlier options.
But testing the pre-tutorial-exit options takes a LOT longer, because I was starting the brute forcing process from frame 7570 for everything post-tutorial, but now I was starting from frame 300, so every attempt would take 3 times longer, HUGELY reducing my RPM (runs per minute). I ran that script for about a week, but it only found a few 10584 ends.
I should be completely done with Pokemon TCG TASes now. Well… maybe a co-authorship on an All Cards run…
ThunderAxe31: File replaced for a corrected CycleCount value.
ThunderAxe31: The improvement is valid. Regarding the entertaining side, despite the fact that this movie appears aesthetically indistinguishable from the current publication, I have to acknowledge that the entertaining ratings for that movie are overall less enthusiastic than the reactions that were gathered on the Workbench forum at the time.
