Thanks to Denial140 for finding 11 more frames, mainly with the wrong warp in 8-4c. Also to negative seven and chatterbox for their work in the previous movie.
A +- 2 frames are lost/won on emulator differences. We argue that any obsoleting submissions need to be done with Bizhawk to avoid bad CPU/PPU timing and lag detection emulation to distort the movie length.
New Tricks
Fake Vine
For some reason, it is possible to fool the game into believing there's a vine in the middle of the level by pressing L+R at a precise moment, with the effect of having Mario shoot up to the right, saving a bunch of frames.
New Wrong Warp
In 8-4d it is faster to take the wrong warp (exactly as the in the original SMB any%)
General Optimizations
Better execution all over the place, especially 1-2
Pretty unexpected improvement, yes vote!
Some information on the spawned vine tiles would be nice, though; I noticed that you were able to catch a couple of those and warp to the right side (on World 8-1 and World 8-2).
Knowing the first Superfast Mario Bros. as being the most unstable of the series, it seems fascinating now to find this glitch exists.
Also, something minor: traditionally, SMB1 TASes end when Mario touches the last axe. Any reason why this TAS doesn't? Not necessarily against it or anything.
I'm Asumeh, semi-expert SMB1 TASer. :)
Check out some of the TASes I don't submit/upload to TASVideos on my YouTube channel, if you'd like.
In progress:
- Extra Mario Bros. (redo) (with HappyLee and w7n) - I'm currently assisting with finding improvements before the boss. On hold; we're currently struggling to confirm that we have the fastest route. Some debugging may also help with finding glitch exits, but neither of us are experts in that field.
- Record my older TASes (excluding any obsolete TASes) and upload to YouTube.
Also, something minor: traditionally, SMB1 TASes end when Mario touches the last axe. Any reason why this TAS doesn't? Not necessarily against it or anything.
TASes end when input ends, For this romhack, you don't have to hold right or A for very long for Mario to touch the final axe, so input can end earlier than Mario touching the axe.
Here is an 11f improvement.
I saved 2 frames in 4-2, but these gave different rng in 8-2. It was very painful to find an 8-2 that lost at most 2 frames to the submission's because the game seemed to decide that this rng is laggy, even when killing some of the enemies. I saved a lag frame in 8-3 and a few minor improvements in 8-4, plus fixing what was a somewhat embarrassing omission from my first version. If I recall correctly, we *did* try this wrong warp back in 2014, but either we couldn't didn't make it work or we ended up running into a firebar in the water room after it. Anyway, it's fixed now.
Current TAS:
[SNES] Jelly Boy
[NES] Street Fighter 2010
Some information on the spawned vine tiles would be nice
Thanks! I'll see to add some decent sub notes. It was getting late last night and just wanted to get this sub in.
Denial140 wrote:
Here is an 11f improvement.
That's amazing! I thought there was nothing left to squeeze. May I add you as co-author and request a judge to replace the movie file? If so, I'll create a comparison video using your movie, instead of mine.
May I add you as co-author and request a judge to replace the movie file?
Yes, that's fine!
I'm wondering whether it may be worth doing a deeper search on (say) 8-2 and 8-3 with the different rng seeds, if that would make sense to do with how your run was created - of course, I am not familiar with the specifics :)
Current TAS:
[SNES] Jelly Boy
[NES] Street Fighter 2010
Also, something minor: traditionally, SMB1 TASes end when Mario touches the last axe. Any reason why this TAS doesn't? Not necessarily against it or anything.
TASes end when input ends, For this romhack, you don't have to hold right or A for very long for Mario to touch the final axe, so input can end earlier than Mario touching the axe.
The final encode (if the movie gets published) will contain the full ending sequence. For the temp encode I wanted to reflect how early inputs end.
eien86 has conquered many epic games with the Jaffar bot. Prince of Persia, Ninja Gaiden, Castlevania. And now: Superfast Mario Bros.
Are we really sure this isn't an April Fools joke?
BTW, here is an encode of the latest TAS, with the ending included. Also in 1080p:
Link to video
This new wrong warp is very nice! I tried jumping from the pipe itself but there are a couple of clever ideas I didn't have to make it faster than what I found previously. Excellent work!
Out of interest, did you rerun the bot on 8-2/8-3 with the new RNG values from the improved 4-2? Is this something that would potentially be worth doing if not? I don't really know how the bot works in practice but I wanted to point it out in case.
Current TAS:
[SNES] Jelly Boy
[NES] Street Fighter 2010
Out of interest, did you rerun the bot on 8-2/8-3 with the new RNG values from the improved 4-2? Is this something that would potentially be worth doing if not? I don't really know how the bot works in practice but I wanted to point it out in case.
I did re-run the bot on the 8 stage all over again but no gains. Unfortunately, there's practically nothing that can be done regarding lag as it has to do with where and how the enemies behave and that's determined exclusively by the frame at which you arrive, as far as I know. Nevertheless, I don't discount ever finding more frames :)
