SGDQ Is Live

This is an archive of a topic from NESdev BBS, taken in mid-October 2019 before a server upgrade.
View original topic
SGDQ Is Live
by on (#151632)
Figured I'd mention, in case you're like me and usually forget it's happening until long after it's over. Link for the lazy here.

Feel free to use this thread as a discussion of all things speedrunning. I enjoy both live and tool-assisted runs myself (TASing Dragon View is how I learned 65816 to start with), though I find it hard to stay interested if it's a game I haven't played before. I enjoy seeing how much better than me people can get, if they put in the practise.

Yoshi's Island is on now. One of the few games I'd rather watch someone else dominate than play myself. I got very little enjoyment out of it despite what were no doubt some very impressive game mechanics (all the crying might have something to do with it).

(yes I plugged myself so shoot me, it's the only TAS I'll ever do...)
Re: SGDQ Is Live
by on (#151633)
Well, since we're talking about speed runs, even though this is tool assisted, this is still insane. (look at 8:40)

https://www.youtube.com/watch?v=9ze0RXDU5HQ
Re: SGDQ Is Live
by on (#151634)
That is some pretty epic abuse of physics right there.

My personal favourite moments in TASing bar none are SMW Arbitrary Code Execution by Masterjun, and Mega Man 3 & 4 & 5 & 6 by Baxter & AngerFist, all beaten simultaneously with the same controller input.

The day somebody beats Sonics 1, 2, 3 & Knuckles all with the same controller input, I will have a new favourite.
Re: SGDQ Is Live
by on (#151637)
Khaz wrote:
Mega Man 3 & 4 & 5 & 6 by Baxter & AngerFist, all beaten simultaneously with the same controller input.

:shock: And I thought that the one I showed was cool...
Re: SGDQ Is Live
by on (#151704)
Khaz wrote:
Mega Man 3 & 4 & 5 & 6 by Baxter & AngerFist, all beaten simultaneously with the same controller input.

I just watched the first 30 seconds and it's obvious they do not have the same controller input.
Re: SGDQ Is Live
by on (#151706)
According to the authors, they abuse lots of frame-exact timing to prevent inputs from affecting the wrong game.
Re: SGDQ Is Live
by on (#151713)
How do you explain mega man can go up in one video, left on another and right in yet another then? Perhaps they use simulatenous left+right but then it does not count as an achievement in my opinion.
Re: SGDQ Is Live
by on (#151718)
Bregalad wrote:
Perhaps they use simulatenous left+right but then it does not count as an achievement in my opinion.

Why?
Re: SGDQ Is Live
by on (#151719)
Baxter & AngerFist wrote:
Of course, completing an entire 'left-room' takes longer than it takes for a screen to scroll. Often, we tried to make the game that has to move left face left, while the others face right. This can for instance be done in a frame of lag at other games. When dropping down into a 'left-room', it is good to face left for a frame, right before the screen starts scrolling.

When the screen is scrolling down, the other games are able to progress to the right. After the screen is done scrolling, it is possible to move left in the left-room, and still move right in the other games, with sliding by pressing down+A. This doesn't change the direction of megaman, which makes it possible to move in different directions in different games. A nice example of this comes early in the run, when Mega Man 5 drops down in the second room of starman (the first level). It doesn't always work however. When Mega Man needs to jump in such a room, left will have to be pushed, which is why sometimes the other Mega Mans are affected. Mega Man 6 can avoid this a little, since megaman6 will move left when you press left and right at the same time. The other Mega Man games won't move, so it is easier to movie left with Mega Man 6 than with other Mega Man games. If you stop walking, Mega Man will take a few frames to get in his final standing position. During these frames, it's also sometimes possible to press left without changing megamans direction.

This short time it takes for Mega Man to position himself after running was also used in other occasions. For example when one of the games has to jump to the left, and it was possible to keep another Mega Man which didn't need to move to the left on the ground. If pressing left is skipped for a frame, the Mega Man on the ground will position himself for like ten frames, so he won't move to the left. The Mega Man who is moving to the left will only stop moving for one frame since he is on the air. It was also used at instances where Mega Man had to move on a rushjet to the left, and a lot when multiple boss battles where going on at the same time.

You can read the rest of the description here.
Re: SGDQ Is Live
by on (#151721)
thefox wrote:
Bregalad wrote:
Perhaps they use simulatenous left+right but then it does not count as an achievement in my opinion.

Why?

Because Bregalad keeps very arbitrary standards to make sure that nothing will ever please him. ;P
Re: SGDQ Is Live
by on (#151722)
Uh, what?

Left+right or up+down is impossible to do on a NES controller, thus using such a combination to trick a game into doing something is cheating.
Re: SGDQ Is Live
by on (#151724)
Coming from one who has been through the process, the TASVideos people judge submissions very strictly. They independently verify the results and have high standards (in my opinion sometimes too high) on what they'll allow through. My Dragon View run nearly got rejected because the (ridiculously long) ending cutscene glitches up partway through and doesn't finish. I think I was cut a break on account of mine is the only Dragon View TAS ever submitted.
Bregalad wrote:
Uh, what?

Left+right or up+down is impossible to do on a NES controller, thus using such a combination to trick a game into doing something is cheating.

I thought their rules or guidelines specifically addressed that but I can't find it. This is not the only TAS to abuse that "illegal" input - I believe it's required for the LttP glitch run that's over in 2 minutes for example. It HAS been the subject of some controversy, and I think runs using it may be treated as a separate category.

The reasoning for allowing it, as I understand, is that a dismantled, modified or even just "broken" controller would allow a player to press both buttons at once. The TAS philosophy is supposed to be what a superhuman player could do, using only the game and console. I would reason that breaking the controller to beat the game faster would fall into that category. If your run however requires an external cheat device, a hacked ROM, or anything else which is not explicitly just the stock game+console, that would be cheating.

The few games I've "hacked" so far all blank controller input if that illegal combination is found, so I'd say if a game fails to do so that's fair game to exploit.

EDIT: just to ramble a bit further, I think the dividing line between accepting a TAS that uses a broken controller and accepting one that, for example, requires one to tilt the cartridge in the slot (as both still just require the game and console) is reproducibility. You can argue theoretically that momentarily breaking contact on a couple pins might beat the game, but inputting a classically-forbidden controller combination is something we can (and do) easily verify on real hardware.
Re: SGDQ Is Live
by on (#151728)
Bregalad wrote:
Left+right or up+down is impossible to do on a NES controller, thus using such a combination to trick a game into doing something is cheating.

It seems strange to disqualify it as an achievement based on an arbitrary rule like that. It's also downright ignorant to question the authenticity of the video based on watching 30 seconds of it and not being able to explain what is happening.

On Famicom it's actually possible to press left+right at the same time in a legit way by plugging in two controllers (if the game combines the inputs from both sources). You can see that in action in this video (which happens to be made by rainwarrior): https://www.youtube.com/watch?v=GXuxNjXNKkw But that's really beside the point; TAS videos are not generally concerned about the mechanical limitations of the default controller of the system.

(What if there's a 3rd party controller that allows you to press both direction simultaneously? Does it make it legit?)
Re: SGDQ Is Live
by on (#151740)
thefox wrote:
TAS videos are not generally concerned about the mechanical limitations of the default controller of the system.

Everything about a TAS is impossible to do on a real console, so pressing left + right is hardly a deal breaker when it comes to authenticity.

I have to agree that the video is not as exciting as the description makes it sound though. I mean, it's too perfect, they abused the times when input is ignored so much that you can hardly tell it's the same input. Don't get me wrong, this is a hell of an achievement, and it's very, very impressive, but the video is just not very fun to watch when it looks like just 4 independent TASes. I was expecting weird/funny things like Mega Man running against a wall in one game while actually going somewhere in another, or the 4 of them making the exact same jump, but there's hardly any of that.
Re: SGDQ Is Live
by on (#151741)
thefox wrote:
(What if there's a 3rd party controller that allows you to press both direction simultaneously? Does it make it legit?)

I've recently acquired such a controller called The Game Handler, and I've confirmed it can press all directions in any combination.

Speaking of TAS videos at SGDQ, I'm excited for what I believe is the console verification of the Ikaruga double play run. Even tough it's going to be simply going to be playback of the encoded video.

I also saw this news item pop up at tasvideos: Announcing a speed TAS race at SGDQ 2015!. The game apparently is going to be a "legally redistributable NES ROM file", that "can be completed in under 6 minutes". I wonder what homebrew game this'll be.
Re: SGDQ Is Live
by on (#151744)
43110 wrote:
The game apparently is going to be a "legally redistributable NES ROM file", that "can be completed in under 6 minutes". I wonder what homebrew game this'll be.
I'd guess Driar: http://tasvideos.org/3893S.html (5m 15s)
Re: SGDQ Is Live
by on (#151746)
43110 wrote:
I've recently acquired such a controller called The Game Handler, and I've confirmed it can press all directions in any combination.

(OT, but I wonder if the discoloration in the footage from Immortal shown in the video is caused by it using $0D color for black. To be clear, I know the game uses $0D, the question is whether that caused the color artifacts in the capture.)
Re: SGDQ Is Live
by on (#151749)
thefox wrote:
It seems strange to disqualify it as an achievement based on an arbitrary rule like that. It's also downright ignorant to question the authenticity of the video based on watching 30 seconds of it and not being able to explain what is happening.

Well, no. I dislike false advertising, and I was expecting to see 4 megamans doing the same thing in different levels. What I see instead is one megaman running left, the other running right, and another climbing a latter. It just looks like 4 normal videos side to side to me. Not interesting, I am disappointed and I quit. I see nothing "downright ignorant" about that.
Re: SGDQ Is Live
by on (#151751)
Bregalad wrote:
Well, no. I dislike false advertising, and I was expecting to see 4 megamans doing the same thing in different levels. What I see instead is one megaman running left, the other running right, and another climbing a latter. It just looks like 4 normal videos side to side to me. Not interesting, I am disappointed and I quit. I see nothing "downright ignorant" about that.

Nothing wrong with not finding it interesting and closing the video. The ignorant part was the "and it's obvious they do not have the same controller input."
Re: SGDQ Is Live
by on (#151767)
Bregalad wrote:
Left+right or up+down is impossible to do on a NES controller, thus using such a combination to trick a game into doing something is cheating.

Is this a difference between an NES and a Super NES? If I pushed down hard enough on the Control Pad of my SN ProPad (third-party controller for Super NES), I could get it to light up in the Kirby's Avalanche input test. The same is true of every PlayStation dance pad I've owned, or possibly the Famiclone built into some "plug-and-play" dance pads.
Re: SGDQ Is Live
by on (#151769)
Bregalad wrote:
Well, no. I dislike false advertising

I don't think it was ever advertised as "Megaman doing the same thing in four games", simply "four games beaten with the same input", which is just plain literally true... It may have misled you but I can't think of any other way they could possibly 'advertise' it.
tokumaru wrote:
Everything about a TAS is impossible to do on a real console

Impossible for a human to do...
Re: SGDQ Is Live
by on (#151772)
thefox wrote:
(OT, but I wonder if the discoloration in the footage from Immortal shown in the video is caused by it using $0D color for black. To be clear, I know the game uses $0D, the question is whether that caused the color artifacts in the capture.)

The Immortal's $0D causes horizontal-sync problems with my own capture card, so I think it's quite plausibly the cause of problems in that capture. (Hardware mistaking it part of hsync, incorrectly reading the wrong thing as colour-burst?)
Attachment:
File comment: Horizontal Sync problems caused by colour $0D in The Immortal
immortal_sync_noise.png
immortal_sync_noise.png [ 183.75 KiB | Viewed 2520 times ]
Re: SGDQ Is Live
by on (#151806)
Khaz wrote:
tokumaru wrote:
Everything about a TAS is impossible to do on a real console

Impossible for a human to do...


Yeah, I'm pretty sure there are TAS setups that run directly on console hardware. In fact, AGDQ did this last year with an SNES. TASVideos.org has also discussed "console verification", and there's some active work being done for things like the GBA + GB Player for emulator-to-console stuff.

Not impossible for the real console to do, just impossible for the human sitting at the real console to do :D
Re: SGDQ Is Live
by on (#151816)
tokumaru wrote:
Everything about a TAS is impossible to do on a real console

This, I don't think anybody has yet checked any TAS against real hardware (and hope you aren't dealing with one of those games that rely on actual random sources for their PRNG seeds, as those are practically unprovable even with the appropriate tools).

Bregalad wrote:
I was expecting to see 4 megamans doing the same thing in different levels. What I see instead is one megaman running left, the other running right, and another climbing a latter.

I take it, you never watched the Mario equivalent then.
Re: SGDQ Is Live
by on (#151818)
Sik wrote:
This, I don't think anybody has yet checked any TAS against real hardware


Are my posts invisible today? :( TAS console verification is a thing.
Re: SGDQ Is Live
by on (#151823)
...and that's why I should just ignore threads that have too many unread posts.
Re: SGDQ Is Live
by on (#151828)
Shonumi wrote:
Yeah, I'm pretty sure there are TAS setups that run directly on console hardware.

AFAIK, they bypass the actual controllers and feed data directly into the controller ports, so they're free to press opposite directions, unlike is the case of the common setup. So even though it's using the real hardware, it's still doing something that can be considered impossible by some.
Re: SGDQ Is Live
by on (#151829)
But a TAS that doesn't use opposing input is completely possible on real hardware. Not all TASing involves glitches. A lot are just about perfect timing. That sort of thing used to be really popular in the Sonic community (RIP Sonice Cage Dome...)
Re: SGDQ Is Live
by on (#151831)
I know, I'm just saying that to a human viewer, TASes do a lot of crazy stuff that seem impossible, sometimes including pressing opposite directions, and showing them running on real hardware doesn't make them look more authentic, specially when it comes to pressing opposite directions since these setups bypass the controllers altogether.

I'm not saying they aren't authentic, they obviously are, it's just that this is not an easily observable fact, just like it's nearly impossible to tell that all 4 Mega Man games are being played with the same input.
Re: SGDQ Is Live
by on (#151832)
tokumaru wrote:
AFAIK, they bypass the actual controllers and feed data directly into the controller ports, so they're free to press opposite directions, unlike is the case of the common setup.

Yup! In the case of SNES, this also lets them also use the register bits that are not wired to any buttons, which I believe is required for that SMW ACE example I posted. To me, I don't consider this cheating still. The console itself is built to take that input; the limited controller is an arbitrary obstacle...

It's worth noting that I spoke to one of the people involved with the TASBot a while back, inquiring about console verification on SNES. I'm eager to see my DV run on real hardware, but it might never happen. I've been told that in the case of the SNES the "APU clock jitter" might cause too much random variation in timing, and could theoretically cause MOST SNES games to de-sync.

I guess you could maybe get around it by forcibly injecting clock signals to keep things synchronized, but at that level of tearing-your-console-apart I don't think I'm ever gonna see it.

NES runs seem generally very verifiable though. Some are more complicated though, like Final Fantasy I for example. iirc it takes its RNG from a combination of WRAM and SRAM, neither of which are initialized on startup. To make their TAS sync, they built a special cartridge that you insert first that clears WRAM and SRAM to zero, and then the run can play reliably. (EDIT: Note that, despite this impressive effort, their run was rejected)

There were other solutions possible though, without the added hardware. One approach that would be crazy to implement but possible: They were able to determine the RNG value from how the first battle played out. They could do a separate run for each possibility and, knowing that, they could have programmed the TASBot to intelligently choose which remaining input file to send in order to make it work.

My preferred solution to this would be to try to develop one single input file that will result in a win regardless of the RNG state, similar to the Megamans project. Probably at the cost of a TON of extra time, if it's even possible at all, but I would call that the most "elegant" solution.
Re: SGDQ Is Live
by on (#151833)
tokumaru wrote:
I know, I'm just saying that to a human viewer, TASes do a lot of crazy stuff that seem impossible, sometimes including pressing opposite directions, and showing them running on real hardware doesn't make them look more authentic, specially when it comes to pressing opposite directions since these setups bypass the controllers altogether.

I'm not saying they aren't authentic, they obviously are, it's just that this is not an easily observable fact, just like it's nearly impossible to tell that all 4 Mega Man games are being played with the same input.


Okay, thanks for clarifying. I guess you should have stated earlier that TAS only "seems" impossible rather than suggesting it actually was :P

IMO, that's half the appeal of watching TAS videos, because they look like something no one could accomplish. No one but a robot, that is.
Re: SGDQ Is Live
by on (#151834)
Anything is "impossible" if your rules forbid it. :P It's impossible to do an SMB run without pressing down if you press down.

The use of L+R is just one exploit of many. These multi-game TASes are a huge technical achievement. Personally if I was doing it I would try not to use L+R, but I wouldn't do it at all because it's a tremendous amount of hard work. I'm not going to shit on a great effort like this because they didn't do it exactly my favourite way. This is still a very impressive thing.
Re: SGDQ Is Live
by on (#151835)
I think this is an amazing achievement, it just wasn't as fun to watch as I thought it would be. It's cool to know it's been done, though.
Re: SGDQ Is Live
by on (#151931)
rainwarrior wrote:
Personally if I was doing it I would try not to use L+R, but I wouldn't do it at all because it's a tremendous amount of hard work.

Wasn't the no-friction Lunar Ball TAS literally done by a bot because there was way too much to test? (and it still took like two years to do everything anyway)