I finished it!

And guess what? It's a MINUTE faster than the cancelled submission!

Currently making the submission text, as well as a userfile that ends input on the save dialog, like the un-optimized submission.

I got a benchmark run from a user on the TASVideos discord of 4:46.

I'm trying to improve my previous userfile.

7 sec IGT (~420 frames) improvement from this userfile!

There is a GitHub repo that I made for these:

https://github.com/darkdevilTAS/SMW2PQ-TAS

EDIT: GitHub repo for all the tests and such.

Comments

just some ram watch stuff

Some info: http://tasvideos.org/forum/viewtopic.php?p=463578#463578

Reupload because the other one has been lost in the masses and I got no feedback.

This is a BizHawk movie (bk2 format) that syncs.

It's 1 frame longer due to how BizHawk and lsnes apply the input, but there's no difference in emulation.

Originally, this is intended to be an improvement run over Vitor I.S (https://www.youtube.com/watch?v=r8eDZko7Iu8). So far, there are not too many major improvements over this. Now on Front Door.

Improvements:

1. YI2: N/A
2. YI3: N/A
3. YI4: No improvements here, just used the same technique of shell jumping to reduce lag
4. Iggy: Using 6/5 trick to keep speed in the first section. Included more entertainment and not collected flower in second room.

Length 1:14:58.858 (270376 frames), around 21 seconds slower than Snes9x's wrong time.

Original TAS: http://tasvideos.org/2307M.html

If it desyncs for you, please PM me.

Length: 1:18:33.954 (283303 frames), around 10 seconds slower than Snes9x's wrong time.

Original TAS: http://tasvideos.org/1868M.html

If it desyncs for you, please PM me.

Original: http://tasvideos.org/1944M.html

Edit: there's a warning about the movie starting from a dirty SRAM. However, this SRAM is a null file. Therefore, there'll be no desync. This internal SRAM can be removed as well.

EDIT 2: This the movie file and the anchor savestate file, Lsnes can work like that. Go to File > Load > Movie and select this file. After that, go to File > Load > State and select this file again, so the anchor state will be loaded.

Playing this movie from the beginning will result in a desync, because I used some cheats in Yoshi Island 2. Those cheats, however, have no influence in Yoshi Island 3.

Encode: http://nach.tasvideos.org/smwsmb-tasv2-masterjun,nach.mkv

I'm not HappyLee, nor did I even put much effort into this to perform all kinds of crazy glitches and get perfect play everywhere. However, the play it still pretty solid, and is within 20 seconds of what HappyLee's latest version would get on this version of SMB.

I did put some effort into entertainment with interesting enemy kills and near misses. So you might find this more entertaining than several other SMB TASs, even though this isn't a polished TAS. Of course a proper TAS would put even more effort into entertainment, although many established TASers in the past have failed to do so for this game.

Note: This is the first time I tried rerecording on lsnes, so rerecord count is way higher than it should be as I was playing with hot-keys and such, and the movie finishes way later than actual game-play too. The rerecord count also includes Masterjun's SMW rerecord count. Read: Not much effort was put into the SMB side of things.

Encode: http://nach.tasvideos.org/smwsmb-tas-masterjun,nach.mkv

In-game level ending times, HappyLee (~5:02) vs this (~5:18) vs first SMB TAS (~5:20):

• 1-1: 370 - 368 - 368
• 1-2: 347 - 345 - 346
• 4-1: 340 - 339 - 340
• 4-2: 356 - 344 - 354 (time units very misleading versus real-time due to beanstalk)
• 8-1: 200 - 197 - 199
• 8-2: 338 - 337 - 340
• 8-3: 244 - 242 - 242 (intentionally slows down to avoid fireworks)
• 8-4: 317 - 309 - 306

All in all, not bad for about an hour of TASing, not aiming for perfection. Although as a judge, I'd reject this for sloppiness. There's even a point where I didn't realize I wasn't running in 8-1 where I should have, not to mention other issues that could be fixed with an extra 5-10 minutes of work.

What's wrong with this TAS:

• 1-1: Crashes into pipe in underground.
• 1-2: Goes over the ceiling instead of through the wall.
• 4-1: Misses many coins, slows down near the end.
• 4-2: Goes between the bricks instead of through the wall, which further requires beanstalk manipulation instead of being able to enter pipe directly.
• 8-1: Slows down in the middle for no good reason, misses enemies.
• 8-2: Slow start. Then big jump followed by Piranha collision glitch could be done faster. In fact I did do it faster and had a better looking glitch, however, right afterwords I stupidly pressed load instead of save, and didn't redo the effort into getting that part to be optimal in entertainment and time.
• 8-3: Didn't kill enough Hammer Brothers.
• 8-4: Didn't wall jump into pipe, swimming suboptimal.

Lastly, issues aside, this run is still faster in real-time than the first few SMB TASs we published when re-synced for this version.

Rom Used: Super Mario World (USA).sfc

Re-record Count: 740

Length: 07:44

Emulator: Snes9x 1.53

if you see any optimization you can tell me