Рет қаралды 347
edit: Check out the new record here!: • Happy Wheels DOTD Worl...
Happy Wheels Dawn of the Dead.lv1 Speedrun World Record in 11.56
I didn't really have a reaction other than thinking "Oh? What?" because I was mostly really confused on how I got it on the 5th attempt of the day. Learning this was much harder than I expected, but now that I got it once getting it a second time should take wayy less time. I think with a *lot* of practice this can get consistent enough to use in full game runs, but for now I think it's best to use a mix between pause buffering and not. The first 3 inputs for me are really consistent, those definitely do not need to be pause buffered. The zombie inputs are (in my opinion) the second hardest since it requires 2 frame perfects in one frame, and shortly after another frame perfect for grabbing. For me the last 3 inputs (especially the two immediately after the zombie) are the hardest since you're moving really fast and it's hard to time and when you're first starting out you don't expect to hit the zombie inputs correctly so you end up not even getting close to pressing the buttons on the correct frames lol.
There have also been some massive improvements to the brute forcer. It used to only have one mode, "Human-like brute forcing", where it would only try inputs that are simple and easily replicable in rta. That mode was just quickly made in a few hours and it was really limited and not as good as it sounds. It was suuper lucky that it was capable of assisting in finding the current helicopter skip inputs, if the TAS file Dray had sent originally wasn't set up exactly as it was then it probably would have never worked in its limited state. Recently I adding a new mode, "Full brute forcing", that can test every single possible input permutation of any given size with any given list of button presses it can use. I also fully automated the brute forcer by making a "bot" (essentially just an advanced macro with image recognition and lots of conditional statements). The bot is still in development, but is like 99% finished and shouldn't take more then an hour or two to finish it (whenever I get around to it). With the new bot I can currently check about 5184000 / (30 * X + 67) - 75427200 / (30 * X + 67)^2 seeds per 24 hours, where X is the amount of igt you want it to wait while watching each TAS. So for brute forcing a helicopter skip, it should be able to check around 9500 seeds every day. For something short like IKH, it should be able to check around 40,000 seeds every day. For reference it took me iirc 4 days of doing almost nothing but manually checking seeds (with help from the brute forcer in its old form) to find the 3 inputs needed for this version of the helicopter skip, which ended up taking around 3500 - 4000 seeds not including my previous attempts or all the seeds Dray checked. Based on that it seems like it will take less than a day to find a new seed, but I highly doubt that since it can't just be used anywhere. We need to do some manual TASing first to get to a good spot for brute forcing. I got a couple ideas on where to use it, but it's still a low chance of them working. The first spot I'll probably be using it on is IKH to try and see if a 0.83 with segway guy is possible. This is also all just external programs. As much as I would have liked to, I never modified the game to make an internal brute forcer. An internal brute forcer would be really good, like we could probably easily check hundreds of thousands if not millions of seeds every day using an internal brute forcer. The main reason I didn't do that is because I don't think I have the rights to do so. I'm not a lawyer or anything so I don't actually know the laws behind it but from what I can tell, the Totaljerkface terms of use tell me not to modify or decompile the game. So unfortunately I won't even be trying to do that (even if that is something I'm allowed to do, I still want to respect the terms).