Joined: 11/11/2006
Posts: 1235
Location: United Kingdom
Nach wrote:
Hi Toad King,
We should be showing off many more speedruns next time.
One issue with this is that, while our runs are certainly cool (especially when run on a console) they don't really play out any differently to what you can see on the site.
I noticed there were a few 4-player races this year and I was wondering for next year how possible it would be to do something like [871] NES Mega Man 3, 4, 5 & 6 by Baxter & AngerFist in 39:06.92 or perhaps [2636] NES Super Mario Bros., 2, The Lost Levels & 3 by agwawaf in 08:49.78 over four consoles? turning all 4 machines on at once wouldn't be too hard if they were all set to power on and then just powered on at the plug via a power strip or something.
<adelikat> I am annoyed at my irc statements ending up in forums & sigs
Joined: 3/9/2004
Posts: 4588
Location: In his lab studying psychology to find new ways to torture TASers and forumers
Raiscan wrote:
Nach wrote:
We should be showing off many more speedruns next time.
One issue with this is that, while our runs are certainly cool (especially when run on a console) they don't really play out any differently to what you can see on the site.
The point is to expose TASVideos and TASs to those not familiar with it. It's not intended to be interesting for our regulars.
Raiscan wrote:
Or have the bot send the reset signal to all 4 at the same time, even if they were turned on at different times. But I do definitely like the power strip idea if feasable.
Warning: Opinions expressed by Nach or others in this post do not necessarily reflect the views, opinions, or position of Nach himself on the matter(s) being discussed therein.
Joined: 11/11/2006
Posts: 1235
Location: United Kingdom
Nach wrote:
Raiscan wrote:
Nach wrote:
We should be showing off many more speedruns next time.
One issue with this is that, while our runs are certainly cool (especially when run on a console) they don't really play out any differently to what you can see on the site.
The point is to expose TASVideos and TASs to those not familiar with it. It's not intended to be interesting for our regulars.
Of course, I agree. But it seems a shame to cut runs short, a waste to show long-ish runs in their entirety, and a lot of setup time to run lots of little runs in succession. I think showing some of the cool runs like the quad runs is something that would be good to show. That and of course a happy balance of other super glitchy stuff / concepts like this year, and some of the most popular runs on the site.
<adelikat> I am annoyed at my irc statements ending up in forums & sigs
Moderator, Senior Ambassador, Experienced player
(906)
Joined: 9/14/2008
Posts: 1014
Raiscan wrote:
Of course, I agree. But it seems a shame to cut runs short, a waste to show long-ish runs in their entirety, and a lot of setup time to run lots of little runs in succession. I think showing some of the cool runs like the quad runs is something that would be good to show. That and of course a happy balance of other super glitchy stuff / concepts like this year, and some of the most popular runs on the site.
I spoke with CoolMatty and he hinted at a fairly lengthy possible block of time, although I later talked to Mike (the head organizer for AGDQ) and he implied that the most we would likely be able to get would likely be an hour and a half of TAS-related material. All of us involved this year generally agree that we should do more normal runs, and perhaps runs entirely without console verification being done in an emulator with really excellent commentary. I still want to have some kind of mind-blowing run again like we did this year but I can't wear as many hats next time - the sacrifices my family had to make were too great and telling my son that I couldn't play two-player Portal 2 with him because I was busy working on a video game seemed both ironic and sad at the same time. Since we know we are more or less guaranteed a spot we'll be able to take a bit more time to prepare and hopefully it will be a bit less time away from the family next time. But, I'm ranting on accident now - I'm seriously tired. :)
So, in summary, yes - I agree with this. Thanks for the feedback!
Are AGDQ2015 TASes resources released or will be released publicly ?
I checked on TASes forum and on entries, I haven't found it yet.
Maybe I missed it ?
IRC log:
<AnotherGamer> So can someone tell me what's up with the part in the first bot segment where the chat is flooded with repetitive crap
<AnotherGamer> Was that intentional or some screwup
I’d also like to know the answer to this.
Also how did the camera hijack work?
My suggestion is that for next time at least one of the exhibitions should be just a TAS of a game, as others have suggested. Of course the run can be really wild and extraordinary, eg. extremely glitchy. I'm thinking of Megaman-levels of glitchiness. (In fact, Megaman might be a good game to show off.) I think this ought to elicit some awe in those who have never seen such a thing.
Another idea for another demonstration: Perform arbitrary code execution, and then upload and run some kind of demo. I'm talking about a demoscene style demo. Something that's awesome-looking and pushes the hardware to its absolute limits, and seems to do things that look impossible for that hardware to do.
Moderator, Senior Ambassador, Experienced player
(906)
Joined: 9/14/2008
Posts: 1014
ALAKTORN wrote:
Nach wrote:
IRC log:
<AnotherGamer> So can someone tell me what's up with the part in the first bot segment where the chat is flooded with repetitive crap
The replay board has a 32-frame buffer. When the end of the Pokemon Red glitch movie was reached, the bot looped back to the first frame in the buffer and kept replaying those 32 frames repeatedly. This was avoided the second time by padding the end of the movie and chaining everything together using bash magic (thanks to some suggestions from ais523).
ALAKTORN wrote:
Also how did the camera hijack work?
The camera has a serial interface that is documented online, and Vulajin was kind enough to make a Python stub library for interpreting the commands up, down, left, right, in, and out as received by Twitch chat. At least, that was the idea. We had some cabling issues and you may notice from the movie of the second TASBot presentation that things were a bit more sketchy than we would have liked all around, but either way you can find more information in the GitHub repository linked from the ArsTechnica article; I'm at the airport on the way home at the moment and can't easily look it up but it should be fairly easy to find.
I congratulate you on a very fun demonstration. I guess I don't have much else to say.
I'm going to sneak in my recommendation for a Super Metroid TAS for next time right here. :) On a console or otherwise. Getting TASbot to sync for 45 minutes might be quite a tall order, on top of getting a verifiable movie in the first place considering that most of them are snes9x ones right now.
I really enjoyed this tasblock as well as the one from last year, thanks dwangoAC and all the people involved in this.
I have however one small piece of criticism to say: I feel we should take time at the beginning to explain exactly what is a TAS and the basic concepts, so that we are sure that people who have never heard of it can understand what is going on.
I have however one small piece of criticism to say: I feel we should take time at the beginning to explain exactly what is a TAS and the basic concepts, so that we are sure that people who have never heard of it with it can understand what is going on.
That's actually a very good point. The presenters did indeed seem to just assume that all viewers know what a "TAS" is, or what the little robot is doing.
Explaining in detail may be boring to those who already know that stuff, but it's informative to those who don't.
Moderator, Senior Ambassador, Experienced player
(906)
Joined: 9/14/2008
Posts: 1014
Warp wrote:
keylie wrote:
I have however one small piece of criticism to say: I feel we should take time at the beginning to explain exactly what is a TAS and the basic concepts, so that we are sure that people who have never heard of it with it can understand what is going on.
That's actually a very good point. The presenters did indeed seem to just assume that all viewers know what a "TAS" is, or what the little robot is doing.
Explaining in detail may be boring to those who already know that stuff, but it's informative to those who don't.
I couldn't possibly agree more. The major problem was more that in both blocks of time we had we were rushed to the last minute dealing with technical issues (the first time we were simply not done yet, and the second time we shot ourselves in the foot and spent hours trying to figure out why it wasn't working anymore until we discovered that extra logging we had added to debug something else was causing corruption).
Both of these statements are various vague ways of saying that I failed to meet my own schedule which was to have absolutely all of the "framework" done by December 9th and the final, final payload complete by December 31st. That would have allowed us to have at least three solid days where absolutely nothing changed unless it had to, using the time to rehearse exactly what we needed to say, who was going to say it, and how we were going to set up and tear down like a well-tuned NASCAR pit crew.
Circumstances conspired against us with various delays, technical issues, and volunteer time constraints and what we ended up with was absolutely no margin whatsoever with all of us rushing just to have anything at all to show at the last minute. As a result, none of us even considered the smaller details such as how we planned on explaining to the audience what a TAS was. To be honest, our block was so compressed that we would have been hard-pressed to explain everything we needed to explain even if everything had gone smoothly.
To correct the core issues described here, my plan for next year is to do less, better. For the sake of ensuring we have a better run (and for the sake of my family), I need to shed some of the extra duties I took on this year and focus only on organizing the team toward our goals, rehearsing with the final payload, and presenting the resulting runs at AGDQ with clarity, humor, and optimism. That means I'll be pushing harder to have other volunteers take on things like organizing prizes, sorting out donation incentives, creating the actual payload, debugging the daatapath / logic analyzer work, tracking any changes needed for emulation accuracy, managing changes to the replay board, creating any TAS movie files, and creating the final console payload if we do another ACE exhibition.
We have a lot more flexibility this year - we have now firmly cemented ourselves as being worth watching at AGDQ (some press outlets have even billed our block as the highlight of the entire marathon!) and after discussions with Mike, the head organizer, we will likely be allowed more time in the next marathon we choose to attend. We still need to decide if we want to continue our "tradition" of skipping SGDQ and focusing on AGDQ only, but either way we should assume that we will have a block of time and get started on development earlier to give us more time at the end for fine tuning. If we can do those things, I'm positive I can spend the time to clearly convey what TASing is and how we use emulators to create perfect playback, what the controller interface board (TASBot) is, what console verification is, and what an arbitrary code execution exploit is in a clear, concise, and entertaining manner.
That was a *far* longer answer than you probably wanted to read (and far longer than I originally planned to type :) but I think all of the elements I described above are important to meet the goal of better clarity. Thoughts?
Incidentally, since I haven't seen it said yet, thank you, dwangoAC, for spearheading this effort! I know that many other people contributed a lot of time, effort, and resources into this project, but you took the lead to make certain that it went as smoothly as possible. Were there things that could have been improved? Absolutely. Was it nonetheless a fantastic showpiece? Also absolutely.
If you want to keep AGDQ as a showpiece for "major advances" in TASing and experimenting with the medium, maybe SGDQ could be used for simply running some entertaining pre-existing TASes?
Also, a random idea: code up an emulator so a speedrunner could play against a TAS ghost. This would of course have to be game-specific and would work best in a game with regular checkpoints (like door transitions in Super Metroid, say). It'd also be wildly unfair. :) But it'd be a neat way to show off the differences between TAS play and highly-skilled human play.
Pyrel - an open-source rewrite of the Angband roguelike game in Python.
code up an emulator so a speedrunner could play against a TAS ghost.
That sounds amazing. It would best work with simple 2D sidescrollers, so that the human player has a chance at playing perfectly. Ninja Senki would probably be nice to run like that, you could have the TAS ghost reset at every screen transition (so that it doesn’t go too far ahead off screen), and at the end of each transition you could have a timer like in driving games that says “+1.32” or whatever the time lost to the TAS is.
I have no idea how technically feasible this is, but the idea sounds amazing.
We still need to decide if we want to continue our "tradition" of skipping SGDQ and focusing on AGDQ only [...] Thoughts?
Skipping SGDQ (at least this year) seems like the best idea, not only because organizing everything seems like quite the hassle already, but also to prevent the idea from getting stale. Having a full year in between would mean that there's twice the content to take your pick from and do something with, which would hopefully allow each presentation to feel new.
I'm not a hardware expert and thus don't know what kind of work, and how much, it would entail, but perhaps the TASbot could be developed in the direction of being more reliable and less affected by external interference.
Moderator, Senior Ambassador, Experienced player
(906)
Joined: 9/14/2008
Posts: 1014
Warp wrote:
I'm not a hardware expert and thus don't know what kind of work, and how much, it would entail, but perhaps the TASbot could be developed in the direction of being more reliable and less affected by external interference.
We made two major changes during the second block, which seemed to have substantial positive benefits - we put tinfoil between gaffer's tape and wrapped all of the controller cables with it, and the second thing we did is we set up everything on a real table rather than relying on the big metal cart that acted like a giant EMI-inducing antenna. We had no issues with interference in the second block, so I'm fairly convinced we can use the same or similar solutions going forward with better success. Additional ideas include placing shielding around the replay board itself, which is something we experimented with at one point while in the practice room surrounded by TV's. It turns out it wasn't needed and made it difficult to work with the smaller, older replay board but we'll be using some method like that eventually.
Skipping SGDQ (at least this year) seems like the best idea, not only because organizing everything seems like quite the hassle already, but also to prevent the idea from getting stale. Having a full year in between would mean that there's twice the content to take your pick from and do something with, which would hopefully allow each presentation to feel new.
I disagree. Better to have smaller contents twice a year than a big once a year. Apparently GDQs themselves aren't getting stale, so why would TASes get stale? If games could just run on an emulator (something to check if it's OK), it would probably reduce the burden significantly while still providing for good speedruns, and indeed, a much larger selection of games. The TAS block does not always have to "WOW!" every year. The other blocks don't do that and they get by just fine.
It's nice to have a WOW-factor once in a while, but it doesn't have to be every time.
Moderator, Senior Ambassador, Experienced player
(906)
Joined: 9/14/2008
Posts: 1014
EEssentia wrote:
Scepheo wrote:
Skipping SGDQ (at least this year) seems like the best idea, not only because organizing everything seems like quite the hassle already, but also to prevent the idea from getting stale. Having a full year in between would mean that there's twice the content to take your pick from and do something with, which would hopefully allow each presentation to feel new.
I disagree. Better to have smaller contents twice a year than a big once a year. Apparently GDQs themselves aren't getting stale, so why would TASes get stale?
These are all fair points. Attending an additional marathon does require someone with hardware and presentation ability to be present, however, and that introduces some challenges. There are very few people with replay boards at the moment, and while I hope that changes I anticipate that for the time being I will remain he primary keeper of TASBot and the most visible representative (ambassador?) of TASVideos to the broader gaming community. Having said that, I have a wife and kids I care about greatly who were substantially neglected in the months leading up to AGDQ 2015 and I need to ensure I find the right balance. I can certainly reduce the amount of roles I fill and reduce the duration of time I spend at GDQ's, which will substantially help. That leaves only one remaining issue - I need some way to pay for the extra plane tickets, SGDQ registration, hotel, and meals that I'm currently avoiding by only going to AGDQ. On the plus side, SGDQ is statistically probable to be physically closer to me and is likely to be cheaper overall, and for some reason people are willing to help me offset travel costs (for the donors that helped me to get to AGDQ, I again say I am in your debt and thank you for your assistance). All that to say I would love to be able to attend at SGDQ this year and will do so if it is feasible, but it will be for a far shorter duration and with a less involved project.
EEssentia wrote:
It's nice to have a WOW-factor once in a while, but it doesn't have to be every time.
I think almost everyone is on the same page here - we really need to get back to basics, spend a bit more time explaining what we're doing and even show off savestates in an emulator first, and demonstrate to the realtime community how these tools can help them practice a run often without any sacrifices to the authenticity of the timing of a game, dependant on emulation accuracy. I'm sure we'll still have something at least somewhat interesting for a finale but I'm positive we can do a better job of just doing off plain-old TAS's without any requirement for console verification for once.
Thanks for the feedback all - keep it coming!
I would caution against having a "race", at least one in which the human is destined to lose. If anything could fuel a common newcomer's misconception of TASes being intended to compete against Unassisted Runs on equal ground, and thus as "cheating", that would do it. And besides, I don't think it would be more interesting than the TAS on its own. It would basically be like, "here's a normal unassisted run, oh and here's also a video of a TAS happening alongside it to see what one of those looks like. Try to watch both.".
I don't know whether I have an opinion on doing it twice a year. I don't think you can ramp up the impressiveness of TC payloads twice a year, haha. But you could certainly show one or two regular runs at SGDQ, which from the feedback I saw would be well received. Including by me. I suppose that means I support the idea.
These are all fair points. Attending an additional marathon does require someone with hardware and presentation ability to be present, however, and that introduces some challenges. There are very few people with replay boards at the moment, and while I hope that changes I anticipate that for the time being I will remain he primary keeper of TASBot and the most visible representative (ambassador?) of TASVideos to the broader gaming community. Having said that, I have a wife and kids I care about greatly who were substantially neglected in the months leading up to AGDQ 2015 and I need to ensure I find the right balance. I can certainly reduce the amount of roles I fill and reduce the duration of time I spend at GDQ's, which will substantially help. That leaves only one remaining issue - I need some way to pay for the extra plane tickets, SGDQ registration, hotel, and meals that I'm currently avoiding by only going to AGDQ. On the plus side, SGDQ is statistically probable to be physically closer to me and is likely to be cheaper overall, and for some reason people are willing to help me offset travel costs (for the donors that helped me to get to AGDQ, I again say I am in your debt and thank you for your assistance). All that to say I would love to be able to attend at SGDQ this year and will do so if it is feasible, but it will be
for a far shorter duration and with a less involved project.
I understand it takes time to plan, prepare and all of that stuff, and yeah, you guys who did this are amazing. But this much involved stuff is required because we're doing such advanced stuff as pulling off arbitrary code execution on the real hardware, which requires both the knowledge of doing so, knowledge of a hardware and replay unit and the unit that controls everything and, of course, having an actual replay board. Like you say, it's a bottleneck because so few have all these required skills.
But consider the normal runners on GDQs. They show up to show off their games, and while they practice, it's much less involved than doing what TASVideos have been doing the last two marathons. Why can't we just "send" some "normal" TASers and show off some TASing or just playing back some TASes? You know, the stuff that doesn't require so much involvement or planning.
That would significantly reduce the burden on everyone. But once in a while, the community can band together and create these awesome hardware hacking stuff that's much more involved and complicated. But it does not have to be every time. Especially now in the early stages. Maybe as time progresses, as more people get involved, as more people get replay boards and can operate them, TASVideos can pull off some greater shows, but until then, why not WOW the crowd with some awesome Super Mario 64 gameplay or the like?
Moderator, Senior Ambassador, Experienced player
(906)
Joined: 9/14/2008
Posts: 1014
EEssentia wrote:
Why can't we just "send" some "normal" TASers and show off some TASing or just playing back some TASes?
This is, in effect, what I've opened the door up to as of the second block at AGDQ 2015; I do wish I had used my time during that second block to say different things but I'm very glad I had the opportunity to say that we plan to do TAS's on emulators in the future alongside console-verified runs in order to open up the number of possibilities of things we can show. If I don't have to be there with hardware and presentation skills, then yes, it's definitely possible for someone else to go to the marathon to represent TASVideos. It's even fine if that person is a little less confident in their presentation abilities - you have to learn sometime, although I'd recommend said individuals consider joining Toastmasters or presenting in front of a local user group before jumping into the 100,000-plus live viewer crowd. (Also, wear asbestos or never read the chat log afterward because chat tends to say really, really unkind things about your spouse, but I digress.)
To rephrase, I don't have to be the sole representative of TASVideos or be the bottleneck, and in no way will I stop someone else from stepping up if they want to fill the role. I'd definitely appreciate it if said person talked to me about it, of course. :) There I go rambling again but yes, I think this is also completely feasible, especially if we decide to do our next console-verified exhibition at AGDQ in order to give ourselves more time and not stretch the development team as thin this year.
