This is an improvement to my previous SMB1 warp run. It is 22 frames faster, due to much better luck manipulation combined with a new acceleration trick in 8-2, which helped bust the 21-frame rule, and 1 frame saved in the very first section in 8-4, using the same trick used in this movie.
Thanks to HappyLee, andrewg and adelikat, as their movies prove the 8-2 improvement feasible, though it's extremely hard without the fireball upgrade.
mmbossman: Accepting as an improvement to... you know the rest.
This is one epic improvement. Not sure if we'll see this improved again. Though, I've now said this 3 times, at this point it really does seem perfect.
I can't believe that we're 1 frame short of improving level 1-1 by 21 frames. And it's like impossible to improve it. I really doubt anyone will ever save that frame and improve this TAS either.
Congrats klmz on a "perfect" TAS.
Super Mario Bros. console speedrunner
- Andrew Gardikis
Entirely incorrect all games have a defined minimal input to complete the game. Although it is generally arrogant/ignorant to claim that certain runs are perfect, even with the Saturnesqe notion of assuming it's perfect till proven otherwise. it is just as wrong to claim that no TAS is ever perfect.
Apples: if you don't mind waiting for the heat death of the universe, we can do that. Say there's 8500 frames in the fastest possible movie, and the game has six inputs that matter. So to check every possible combination of inputs, you'd need to do 6^8500 checks. The actual number is smaller (we can assume, for example, that any inputs that result in Mario dying are unproductive, and that input during points where you have no control over Mario don't count), but even 6^20 -- every possible combination of the first twenty frames of input -- is 3656158440062976. If you could simulate 100000 frames per second, you'd need 1160 years to check all of those.
Pyrel - an open-source rewrite of the Angband roguelike game in Python.