Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
negative_seven wrote:
Interestingly, this submission is old enough to feature this now-incorrect comment:
Submission text wrote:
Furthermore, so far, nobody in the world has been able to write a computer program that plays Go better than an average club player.
There is this completion of level 14 (as opposed to the level 10 chosen by Bisqwit):
Link to video
It is slower than Bisqwit's run, but not by much - possibly faster if you were to discount menuing and thinking time. Although it is played on some English-language version, which could conceivably have a different AI.
Aww, shoot, I did t find this video. So level 10 isn’t hardest difficulty? Is 14 the max?
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
The movie has literally no TAS related entertainment value. However, I’m not sure what vault rules it violated on its 2nd rejection, but regardless, I think this meets all current rule requirements, thoughts?
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
I unrejected this submission, because we allow board games now and this is a good TAS for this game. It's been a long time, so I looked, and current human speedruns are 11 minutes, and I could not find a TAS that has improved this submission. As far as I can find, it still beats all known records.
Looking for some feedback, thoughts?
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
Bigbass wrote:
Thanks Randomno for compiling these lists. Worth noting however that these changes can only be performed by judges or other staff. Editors cannot modify submissions (except for catalog changes).
Some of these issues, like wrong judges, even staff can't do. They are data issues from migrating the old (and bad) data. I'll fix them direclty in the database
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
Dang, didn't notice this one during the festivities, thanks for the submission. Unfortunately, I think we have officially gone too far with this gag. Fortunately, I live for going too far with gags.
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
User movie #638148478993374149
Another update, got through Chapter 4, and am up to the first encounter of Chapter 5 (with the 2 Liclicks).
Improvements in Chapter 4:
- Getting the gunpowder jar then Orin, this way the death warp that needs to kill off 2 characters is with the more powerful enemies, which saves an action
- Better luck manipulation in the level ups, I get no vitality/hp in every level
- Transition lag manipulation, particularly in the ending scene of sailing to Endor, a surprising amount of frames were found just in a scripted cut scene
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
User movie #638144270152713770
Chapter 3 done! This was the major blocker that stopped my progress years ago, and the last of the super incredibly unlikely manips. I get the 3 max armor offers without declining a single one, which saves some seconds, but the odds of this are about 1 in a million or more. I'm 2475 frames ahead of the published movie with about 600 because I end 1 level ahead but will still end up on the same final level. The strat here is to get a 2nd metal slime so that I have more strength for the drop encounters and not need to do critical hits. The experience works out that it will not cost another level (but 3rd metal slime would have), so this will be realized as saving by the end of the TAS
I'm officially optimistic about this TAS and super motivated to keep going! The worst is behind me at this point
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
So I managed to finish a redo of Chapter 1: User movie #638117571365360120
It is 227 frames faster, about 180 from staying at the inn. I also got an even more absurd level up. About 10 million bot attempts went into the level ups at the end of this chapter.
Other frames were scavanged from screen transitions. More and more I am finding that to truly optimize this game, I have to do more and more precise bots that bot larger chunks of the game. At this point, almost nothing is done by hand anymore.
The good news is the most unlikely events of the whole TAS are done. Bad news is that there is stil an awful lot to do, and the Chapter 2 boss fight is also absurdly unlikely, will probably be a few million bot attempts as well.
Finishing this TAS will be a long journey, yeesh
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
Wow, nice improvements! I really enjoyed watching this, I thought the use of the hiding spots for a position boost was clever. I can't believe how quickly you were able to glitch through some of those barriers too.
Yes vote from me.
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
FractalFusion wrote:
In case you need to do any testing on the hardest difficulty (the CPU take a really long time sometimes), you can use my form-based Lua script to help with this. I used it to make a (not very useful) Stockfish vs Battle Chess TAS (in my userfiles somewhere). Note that Stockfish is non-deterministic so other attempts at using Stockfish to beat Battle Chess may differ.
Also QuickNES is reasonably accurate for this game. The CPU thinking time is virtually the same and the only difference is NesHawk is slightly faster on some screen transitions (at the beginning of the game, and also when using capture animations). So you can optimize it on QuickNES and then make the final movie on NesHawk. Although probably everyone uses QuickNES anyway.
Yeah, QuickNES is what I used to refine strategies and NESHawk once I knew what I wanted to do.
I'm not sure how interesting a non-glitched run is, as it doesn't do much to set it apart from any other chess game. But I would be curious what you come up with.
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
User movie #638051848574916873
I improved this run. It is 684 frames faster at the fastest crossing of the finish line of the final race.
I think early input end is fun and interesting here so I am going for that, manipulating an invulnerability can greatly shorten input, I'll work on that too. Not sure which would be preferred for submission.
I'll submit once I have that worked out and an encode, might be a few days
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
I updated the movie file with one with the Bald Bull realtime improvement. Note that I did not update the encode, because I didn't not want to take the time to make a new encode and upload it, especially since no in-game times were improved.
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
DJ Incendration wrote:
Nice work, yes vote! There's a real time improvement you missed. On Bald Bull, it's actually faster in real time to knock him down with an uppercut when he's dizzy.
Bleh, I don't even care about realtime improvements and that's so early in the movie. However, I'm working on the update now, it will be a 28 frame realtime improvement, for what that's worth.
Awosomeandy wrote:
Zallard1 did not find this new adjustment, it was Lixunis.
https://youtu.be/Ao-ZP7-mxkI
here was the original TAS video
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
Yes, I tried a number of things, and I should document them here for future reference.
1) Lose a turn, probably not helpful but worth trying. You can play your first move, then do a take back during the CPU turn. It will now be the opening board position but black is going first.
2) Play as black.
3) Variations using e3
4) Absurd moves early on in hopes of more moves with more take backs but no long waits.
This submission manages to only have one "long wait" with 2. Nf6, being over 11k frames, vs 5k on 3.Nxh5 as the next longest. Some moves are even around 1k or 3k. In theory, there is potential to do a few short moves instead of one long one, but nothing I tried worked out. It would help if this game had at least some opening book theory, but alas.
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
Replaced the movie file with an updated version. New version uses the 2+2 takeback strategy, also moved to NesHawk for better accuracy and quicknes has a lot of extra lag. This movie is 14 frames slower than Fractal's but I confirmed it is all emulation differences. All frame differences were in screen transitions where no input is happening.
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
FractalFusion wrote:
adelikat wrote:
FractalFusion wrote:
IIRC, this TAS is easily beaten by ~1m30s, somewhere on this forum.
Ah, this: https://tasvideos.org/Forum/Posts/108220
Very clever, looks like 2 take backs + 2 take backs avoids the computer thinking for one more move, I'll make an updated version
For a moment, I thought no one actually wanted to take the credit that I was practically giving away for free.
Cool, it's yours.
It's funny, and embarassing, that I missed this movie of yours by simply not reading the entire thread. I remembered Dacicus' TAS and found it and began working to improve it. Your move order is exactly what I stumbled on too. I redid the movie, it's now 115 frames slow than your, that I am confident is emulation differences based on where the frames are lost. I used quicknes so I can fast forward through painfully long thinking times, I'll have to redo with NesHawk to see what happens there.
I also want to look for more possibilities as well, so I am not ready to update this movie.
Your words worry me that you are upset, but I am not sure. I didn't mention authorship, but now that I have redone the movie, I see it is the same moves and take backs. I would not feel comfortable with this movie, as is, being a sole authorship of mine.
Emulator Coder, Published Author, Site Developer, Site Owner, Expert player
(3650)
Joined: 11/3/2004
Posts: 4758
Location: Tennessee
FractalFusion wrote:
IIRC, this TAS is easily beaten by ~1m30s, somewhere on this forum.
Ah, this: https://tasvideos.org/Forum/Posts/108220
Very clever, looks like 2 take backs + 2 take backs avoids the computer thinking for one more move, I'll make an updated version
