Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
feos wrote:
slamo wrote:
To make things less onerous for judges, the proof that the score really is maximized could just be that it beats all other known scores and there are no obvious ways to increase it further.
BTW I don't even think we need this proven. We just need it to look optimized, which means not easily improvable by a lot.
I'll agree with that. After all, we have the same standard for judging optimization in every other run. However, I still think the score should match or beat all known scores. There's not much of a point in calling the branch "maximum score" if somebody can just point to a higher score during judgment.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
I talked with EZGames a bit about this but I'll also put it here: The order of the levels is just bizarre in this game. In the longplay, the level order goes 1 - 2 - 1 - 3 - 2 - 1 - 3 - 2 - 1 - 2 - 1, which has no pattern at all. I'm not sure what would be considered a "loop".
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
Saved 101 frames on the first several levels:
https://tasvideos.org/UserFiles/Info/637927253341869477
It's not perfectly optimized, just a proof of concept. The most confusing improvement is that the enemy cycle on the top floor of level 2 was easily makeable. Some other optimizations from delaying some jumps up to ladders to get a bit higher up.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
1. Yup, sounds good to me. To make things less onerous for judges, the proof that the score really is maximized could just be that it beats all other known scores and there are no obvious ways to increase it further. I also think that if a score is beaten, the higher score obsoletes the lower score, even if it doesn't beat the time (and other obsoletion rules apply, i.e. don't be needlessly sloppy in unimproved areas).
I don't really have a problem with allowing timed sports games for this. Trying to score as much as possible in sport seems like a formidable, albeit repetitive challenge.
2. Trying to look at this from a casual viewer's perspective, I think using the display score is the more appealing option. It just seems more approachable to me. Viewers should not require knowledge about RAM watch to comprehend the goal. A maximum internal score run just seems like reaching the max display score, continuing past that with no perceivable progress, and then randomly ending with the implication "ok, it's maxed out now, just take our word for it".
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
The movie syncs on this good dump:
Identified from BootGod's database
------
BEGIN NES rom analysis:
Found iNES header:
pr=128,ch=128,wr=8,vr=0,ba=0,pa=0|1,brd=MAPPER001,sys=
Found iNES V2 header:
pr=128,ch=128,wr=0,vr=8,ba=0,pa=0|1,brd=MAPPER0001,sys=NES-NTSC
Since this is iNES we can (somewhat) confidently parse PRG/CHR banks to hash.
headerless rom hash: SHA1:5C5A9D8B12DB810567E3F4EADFD0005956A2890F
headerless rom hash: MD5:C88A679DB40A606F93EDB697C9422CC2
Could not locate game in bizhawk gamedb
Chose board from nescartdb:
pr=128,ch=128,wr=0,vr=0,ba=0,pa=0|0,brd=NES-SLROM,sys=NES-NTSC
Final game detection results:
pr=128,ch=128,wr=0,vr=0,ba=0,pa=0|0,brd=NES-SLROM,sys=NES-NTSC
"Three Stooges, The"
Implemented by: class SxROM
END NES rom analysis
------
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
I'm not familiar with this game, but if this ending requires different enough routing from the "good ending" (i.e. routing levels without having to pick up certain shards) then I think that's worth having as a separate run.
Personally I think we should be allowing different endings as Standard branches in general. I don't think that would be too controversial.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
Double posting because this is important!
feos narrowed down the desync to the "flags" byte in the SRAM. If bit 2 is 0, then it syncs. According to this, that bit is responsible for the mono/stereo setting. You can adjust this in the default SRAM file, and you do not need a hex editor. To do this the right way...
1. Get a NTSC Gamecube BIOS file, name it IPL.bin, and put it in the \Dolphin Emulator\GC\USA\ folder (if that's already there then skip this).
2. Go to Options -> Configure and go to the Gamecube tab and make sure Skip BIOS is unchecked.
3. Start the game (do not run the movie yet) and hold the A button on the Gamecube bootup screen to access the Gamecube BIOS.
4. Open the menu on the left and set the audio to Mono instead of Stereo. Back out to the first BIOS menu and stop emulation.
5. Go to Options -> Configure and go to the Gamecube tab and check Skip BIOS.
Run the movie and it should sync now.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
This appears to be an SRAM issue (the SRAM file is in \Dolphin Emulator\GC\SRAM.raw). If I delete this file and start the movie, it desyncs on the default SRAM file Dolphin generates. However, if I take the default SRAM.raw file and completely 0x00 it out in a hex editor, then the movie syncs fine for me. Would this be a reasonable state to start a Gamecube in? It's easily reproducible, at least.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
I've been thinking about a "Codes" branch for Standard for the past month or so, and I do support it. Figuring out the best codes to use is an actual optimization challenge, and I don't mind it as long as it's very clearly labeled and doesn't obsolete any Fastest Completion runs.
We talked a bit about it in staff Discord, and I like the following guidelines for the Codes branch:
The codes should be entered as part of some in-game system, external devices like Gameshark are not allowed.
The benefit of the code must manifest itself as part of some intended game mechanic. For example, codes that give you items, characters, unlockable abilities, or stat boosts are allowed. Things that are not allowed are mechanics the developers specifically hacked in for the code, such as noclip, infinite god mode, infinite air jumping, etc.
Codes that skip content are not allowed, i.e. level skip codes, unlocking every level, or warp cheats. If a code unlocks something that enables the ability to do a new skip, such as getting an intended ability early, that's still allowed.
And yeah, allowing AR codes is a bit too powerful for Standard. I would be fine with AR codes in place of verification movies, but the burden of proof must be on the author to show that the game state they are creating is actually possible.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
So we figured out that the floppy creation process on different PCs can result in different file structures on the floppy which causes desyncs. I've added a warning to the PCem page not to make your own floppies for TASes, using the ISO method is more reliable. However, if there's a good dump of the original game floppy available, it's recommended to use that.
I agree that the late 80s setup is more appropriate. The early 90s one was super glitchy for me if I tried to play it casually. This is to be expected, some older games have problems with faster processors, I had the same problem with Ford Simulator.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
The dump this movie uses is strange, but the movie does sync on this verified good dump:
Identified from BootGod's database
------
BEGIN NES rom analysis:
Found iNES header:
pr=16,ch=8,wr=8,vr=0,ba=0,pa=1|0,brd=MAPPER000,sys=
Since this is iNES we can (somewhat) confidently parse PRG/CHR banks to hash.
headerless rom hash: SHA1:A5030E82D7C38F2D51016D947CB29E6FF788367C
headerless rom hash: MD5:1CA706896A8D4F2A2B5480D941130A4A
Since this rom has a 16 KB PRG, we'll hash it as 8KB too for bootgod's DB:
PRG (8KB) + CHR hash: SHA1:059454A557AE902E6A43130D27B8482CBDFFF7C7
PRG (8KB) + CHR hash: MD5:0DF19EA102F9013EE7494F809F83C66D
Could not locate game in bizhawk gamedb
Chose board from nescartdb:
pr=16,ch=8,wr=0,vr=0,ba=0,pa=1|0,brd=NES-NROM-128,sys=NES-NTSC
Final game detection results:
pr=16,ch=8,wr=0,vr=0,ba=0,pa=1|0,brd=NES-NROM-128,sys=NES-NTSC
"Donkey Kong Jr. Math"
Implemented by: class NROM
END NES rom analysis
------
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
katsuster wrote:
Hello, I've updated my TAS faster than 10s.
Would you accept the movie of new version? If yes, I will send it and update the description.
7m32s version:
https://www.nicovideo.jp/watch/sm40566887
Yes, please upload the new movie file to userfiles.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
I still can't get this to sync. I've confirmed my config matches (looks like you removed the B: drive for some reason) and the files in the floppy match as well. I've also deleted flash.bin. Maybe it's the .nvr file, did you do anything to it? I'm just using the one from the Early 90s package renamed barb.pb570.nvr.
One thing that's really throwing me off is that if I use the Early 90s .nvr, the checksum error shows up at around 13 seconds, way before the actual Escape input to clear it. Can somebody make an encode of the bootup so I can see what's going on?
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
Dimon12321 wrote:
Doesn't it correlate with what you and feos responded to me earlier about non-deterministic Windows installation? So, if we use only one specific distribution, then we get it working fine, right?
Yup, we'll tell you the Windows version and the SHA1 of the CD to make sure everyone has the same setup.
How will it all work?
1. From what moment should the actual movie begin? Can put an installed game (folder + .exe) on a PCem hard drive and start a libTAS movie recording where we simply surf the explorer up to the game .exe, launch it and go on with the game as usual?
Or should we begin with mounting a game image with, let's say, Daemon Tools and install it live, then launch it? - Can it be a separate movie?
2. What about game distributions? If we record us installing the game, will it be ok to show a CD key for a game? Specific distributions, that free you from using it or keeping a game disk in the reader to make it work, are actually more stable and convenient.
1. All modifications to the hard drive MUST be done within a libTAS recording, otherwise you introduce more variables that could cause desyncs. You can use installation discs, or we have a reliable method of putting files onto an ISO and you can copy it onto the hard drive within libTAS. And yeah, installation can be a separate verification movie, kind of like with SRAM for other emulators.
2. We definitely wouldn't want CD keys being distributed through movie files. Maybe we could have a discussion about allowing modified versions in this case.
I guess OpenAL installation should be added to the config. Or should it be included into the game installation part?
Visual C++? Even though I can think of 2005+ distributions only, maybe elder versions existed that had to be installed back in the day.
What requires these? And would they not be included with the game?
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
Some other runs that use codes also got submitted, so the plan is to have a wider discussion about that and go from there. Sorry that this is taking a while, but whatever we're going to do will set some precedents so we have to take some time to think about it first.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
We had a pretty major breakthrough for PCem + Windows, and I think we actually found a method that should reproduce a Windows installation reliably for multiple people. For a long time the main hurdle was the fact that Win9x required a CD key to use it at all, and we obviously couldn't provide an installation movie where we end up distributing a serial. However, Windows XP SP3 does not immediately require you to enter a CD key, instead letting you have a 30 day trial period before you have to activate it. It retains full functionality in this trial period, so it should be fine for our purposes (unless your TAS is more than a month long somehow). Obviously we cannot distribute this config like we did the DOS ones, but if we provide the SHA1 of the Windows XP disc and an installation .ltm, then it shouldn't be too hard to produce for yourself.
I have been testing Windows XP in PCem and it surprisingly works great with a few tweaks. This will finally take full advantage of PCem's capabilities. It should be able to run any Windows game from 2000 and earlier, and probably some after it, depending on the system requirements.
Here is what I'm thinking for the config, this is the best hardware PCem has:
Pentium II 450 MHz
256 MB RAM
3DFX Voodoo 3 3000 (16 MB VRAM)
Sound Blaster AWE32
4 GB hard drive (is this enough?)
72x speed CD drive
PS/2 mouse
Some notes about the config:
DirectX 9.0c automatically gets installed with Windows XP SP3.
Sound Blaster does not get automatically configured for some reason, but is easy enough to install manually.
The official Voodoo 3 drivers don't get along too well with XP, but some dedicated people have made 3rd party drivers that I've tested and can confirm fix any 3D issues I was having. It comes with 3DfxTools so you can tweak settings for your game if needed.
Is there anything I'm missing here? And please respond if you think anything else would need to be installed for the config.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
It looks like you can get faster times if you use both the mouse and keyboard: https://www.youtube.com/watch?v=re_0AtObmUU Ehh, looks like this requires tab-selecting, which isn't supported in Ruffle right now.
Also, this isn't the maximum score, apparently you can get up to 188 points.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
I don't think there's any question that this should qualify for Moons - at least I personally would feel comfortable putting it there. The voting is pretty good, the goal is clear, and I thought it was different enough from any%. I already watched it and voted yes.
Now here's something that deviates from our past philosophy: I think we should weigh a category's significance in the speedrunning community more heavily. This category has had over 100 different runners and hundreds of attempted runs over the past couple of years. If we don't believe that makes this category significant enough to publish, then we are the ones that are irrelevant. These kinds of categories are birthed from a specialized community, and as outsiders to that, we should not really be the ones deciding if it's meaningful or not.
As for runs with codes in general, I think there is a place for that in Standard, at least in certain circumstances. With this run I'm not so sure - I don't think anyone wants this to get obsoleted by a cheat run that just goes straight to the last level. Probably best that it stays in Moons as a separate category. But for other submissions, like #7451: andy120195's DS Pokémon Mystery Dungeon: Blue Rescue Team "Wondermail" in 1:36:40.01 and #7485: ThunderAxe31's GB Rolan's Curse "password, game end glitch" in 04:04.80, finding the best possible passwords to use and beating the game as fast as possible actually represents an interesting optimization challenge. I wouldn't want codes that overly trivialize the game and wouldn't actually impress anyone, like noclip or God mode, but I think there's some room to work with here. Something like a "New Game+" goal would be cool, where you can utilize password systems or SRAM to have things unlocked from the start.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
1. If it's an official release and the files are unaltered, then it's perfectly fine to use.
2. I believe the uncracked version has been preserved. For uncracked stuff, anything you find in a zip file is questionable, and it's best to use original floppy disk images if you can. If the game comes with multiple disks you can use the disk swapping hotkeys.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
I got the image and it desynced for me. Can you please provide full instructions for how you set your config up, and also post the contents of your .cfg file? I also noticed there's a large amount of time between bootup and starting the game, that's unusually long for the early 90s config.
Experienced Forum User, Published Author, Reviewer, Expert player
(2393)
Joined: 5/21/2013
Posts: 414
Can you please put how you made the floppy in the description? This information needs to be public. We'll need the MD5 hashes of the files, as outlined here.
