I was able to organize exhibiting Mega Man, Sonic Advance, and Ikaruga with commentary from respective realtime runners (cleartonic, Kirbymastah, and ShinerCCC - Thanks!)
Arthur 'Mr. Podunkian' Lee, the original author of Streemerz, gave me permission to use his work, joined the IRC channel during the competition, and even made a donation directly prior to the results playback
thefox who ported the original version of Streemerz to the NES gave me permission to redistribute the NES ROM file and even posted helpful memory addresses
TASVideos.org received a huge boost in traffic with continued increased load following the event per DeHackEd
There was a big uptick in new user registrations (sort descending) following SGDQ and a number of new users are continuing to engage on IRC and in the forums
Overall "press" coverage (i.e. bloggers and news outlets) have provided positive coverage of what we did
A number of commenters in various places indicated they greatly appreciated that we stepped back in intensity compared to last time to explain how we go about creating TAS's
While I was onsite at the event I often had TASBot with me and a number of people took selfies with TASBot - whenever I walked around *without* TASBot I might as well have been invisible :)
The not so awesome:
I was jittery and shaky while presenting (not from nervousness of being in front of that many people but more flat out not feeling well, in part due to lack of sleep combined with poor opportunities for good food the day of the block)
I did not do a good job of explaining why Rockman was in an emulator (can't console verify due to DelayStageClear glitch timing) or why Ikaruga was a video (we sort of touched on the disc read speed issues but it could have been disclaimed better at the start); some people didn't understand we console verified Sonic Advance but even then the existence of content that wasn't console verified was received with mixed reviews
Commenters in Reddit and elsewhere were downright abusive in describing how badly I disappointed them by not blowing their minds again (I won't be linking it, it's mostly vitriolic), with many incorrect references to "last year was awesome, this year sucked" even though AGDQ 2015 was only 6 months ago
We didn't have everything worked out on the speed TAS competition judging methods due to some last minute scrambling and as a result the TASJudy script didn't have adequate time to run through all submissions, forcing me to announce the best results we had tabulated thus far (and slighting Mothrayas in the process); we later discovered I made an error in my beliefs of the sync stability between FCEUX and BizHawk and I have apologized to Mothrayas in IRC for the problems; in the end, a real submission was made with the help of multiple authors
We still took far too long setting up despite my best intentions; I was yet again scrambling at the last minute to work through problems and I was not as prepared as I needed to be for the third event running
Overall, in the balance I'm very happy with how this event turned out, but I do have serious regrets for not doing a better job communicating while on stream that viewers should expect console oriented shenanigans at AGDQ and less extreme showings at SGDQ simply due to the amount of time preparing big exploits require
I'd like to again thank Endrift for making console verification of Sonic Advance possible and for being willing to participate on the couch - that run was much better off for it. Additional thanks go to the other commentators including cleartonic, Kirbymastah, and ShinerCCC all of whom made massive contributions in explaining things from an RTA perspective better than I could have. I'd again like to put out a word of thanks to Aktan for awesome encodes and to RGamma, Raiscan, SMK, adelikat, Mothrayas, DeHackEd, and Nach who all helped out in various ways with the website and the speed TASing competition, along with a couple people who wished to remain anonymous - you know who you are, thanks for the help! As always, thank you to those who donated and helped me get to the event (I'll be posting a transparency report once everything settles).
I have every intention of making TASVideos participation happen at AGDQ 2016 so any feedback you have on how this event went would be appreciated below. Thanks!
I didn't notice anything wrong with your presentation. If you weren't feeling well, it really didn't show at all. In other words, you made a very good job.
As for the contents of the TAS segment, I liked that you selected more "pure" TASing for this marathon (in that they were actual speedruns and playthroughs), rather than go with the non-speedrunning gimmicks. However, I think that you went perhaps a bit too much to that side in that you used the emulator for one run and just a video for another. I understand perfectly the technical reasons for this, but I think that the audience would prefer seeing the cute little robot "playing" on an actual console. Just running an emulator on a PC doesn't have that same feeling...
I was thinking, perhaps, and if at all possible, if in the next marathon (or one of the future marathons) you could actually have the robot visibly connected to the console on-camera, for everybody to see. To really drive home that yes, the robot is directly connected to the gamepad port of the actual console. I don't know if this is technically possible because I don't know what kind of setup is needed at the site, but just a random idea.
If people are desiring more "gimmicks", then perhaps have one such "gimmick" and one more "regular" TAS, as an in-between compromise that ought to have the best of both worlds.
Moderator, Senior Ambassador, Experienced player
(898)
Joined: 9/14/2008
Posts: 1007
Warp wrote:
If people are desiring more "gimmicks", then perhaps have one such "gimmick" and one more "regular" TAS, as an in-between compromise that ought to have the best of both worlds.
I'll save the majority of the heavy gimmicks for AGDQ's but in effect this is exactly what I aimed to do - show a couple of solid TAS's without console verification and show a really unique method of TAS'ing a Game Boy Advance game on real hardware. It would have been impossible to show Mega man or Ikaruga console verified and Ikaruga looked *way* better outside of an emulator so it was an obvious choice but I should have done a far better job of explaining why I did this. I will have to consider how to handle this type of situation in future marathons, but for AGDQ 2016 I expect to do console verification and this problem won't exist there so I'm not putting much energy into it for the moment. Thanks for the feedback!
Joined: 8/14/2009
Posts: 4083
Location: The Netherlands
Yeah, the TAS block this time around wasn't as wow-ing as the AGDQ 2014 and AGDQ 2015. Of course, it's hard to top things like SMW Pong/Snake and Pokémon Plays Twitch, but to the average short-attention span viewer of twitch chat or reddit, an explanation/introduction of TASing around in Mega Man is just boring in comparison. Live TASing is just not really an entertaining watch if you're not involved into it, which the average viewer was not. I think the problem was that the viewers were expecting more of the past few TAS Blocks, which were mainly console verification and specially-made entertainment runs for that, whereas this TAS block was really focused on TASing itself and showcasing actual tool-assisted speedruns (and one superplay). I think the previous TAS Blocks gave a bit of a wrong impression to the average viewers as to what typical TASing really is like. Instead of console verification, and specially made entertainment runs, it consisted of an introduction to TASing and playing back videos of published movies. It didn't contain something groundbreaking like the past two TAS Blocks (even though AGDQ 2015 was just half a year ago, people easily forget that), so it did not really live up to a lot of viewers' (unrealistic) expectations. People were just expecting too much, and expected too much something like the AGDQ TAS blocks.
The presentation itself was good, I thought. Everyone did their part in the TAS Block well. The TASing introduction looked like a good introduction as to what the main essence of TASing is, and the run commentaries were good. The TAS block itself was executed well.
Regarding the TAS Race results, I'm aware that it involved a lot of last-ditch efforts to make the process work at all, and insufficient testing/too quick conclusions were the result of that. I'm not really blaming you or anyone else, but it was a bit disappointing not to see my (likely winning) result on the GDQ stream regardless. (Also, I still really want to see the final results). Other than that, as you noted, the TAS Race has been a great success, and was really successful in attracting new members to the community. No matter how "bad" or boring the TAS Block may have been reviewed on reddit or twitch chat, it did its job well.
All in all, I think you shouldn't let the negative feedback upset you too much - reddit and especially twitch chat are known for being harsh communities at times. I can see a bit where they are coming from, but don't let their critique take away from the fact that the TAS Block has been overall a success, and the TAS Race especially has been a big success. Be proud of that.
http://www.youtube.com/Noxxa
<dwangoAC> This is a TAS (...). Not suitable for all audiences. May cause undesirable side-effects. May contain emulator abuse. Emulator may be abusive. This product contains glitches known to the state of California to cause egg defects.
<Masterjun> I'm just a guy arranging bits in a sequence which could potentially amuse other people looking at these bits
<adelikat> In Oregon Trail, I sacrificed my own family to save time. In Star trek, I killed helpless comrades in escape pods to save time. Here, I kill my allies to save time. I think I need help.
Yeah, these first three TAS blocks were like the Godfather trilogy: the first two were incredible, whereas the third, while still great, couldn't live up to those heights.
To be honest, even if we had shown some custom-made ACE run for this event, it wouldn't have been as impressive as the first time, because it's something people have already seen before. The problem is that the showcase has to do something more impressive every year just to get the same amount of mind-blowery. That even the Megaman TAS (which in my opinion is still the greatest TAS ever made) got a relatively lukewarm response makes me doubt that we'll ever be able to reach that level of hype again.
Joined: 8/15/2005
Posts: 1934
Location: Mullsjö, Sweden
What I would like to see is some actual TASing done live. Yes, it might be a tedious process, but I know (After testing it during ESA) that a lot of people find it interesting. I didn't make a great job with the presentation myself at ESA however. But people have enjoyed watching me TAS backstage at ESA/AGDQ, one reason being that I'm very quick with my fingers when TASing, speed TASing FTW. You had a brief test with Mega Man where you showed the concept at least. It would make it more interesting if it was an actual TAS project that you worked on optimising and then replaying the result in real time to show what you managed to accomplish and how long that took.
One downside with this TAS presentation was that there were practically no surprises (Apart from which game was picked for the competition.). As the other three runs were already known, anyone can have looked them up on Youtube long before the actual stream. It doesn't have to be a crazy ACE TAS, but just a new TAS that has yet not been shown to the public.
I'd also like to add, that a potential reason we had so many submissions (and contributing to the fact judging took so long), was me fucking up the submissions page:
1. no spam protection (someone started mass-uploading files (60-100/mn) that required cleanup and not all spam submissions were removed)
2. a bug that lead to hash collisions in the uploaded files due to the hash being calculated on a very small subset (mainly 0-length bytestrings) of possible files necessitating reuploads after it was fixed. Old uploads weren't deleted resulting in duplicate regular submissions (some of them might be quite long)
The technical side needs to improve _a lot_ for the next event. Much was done impromptu and manual maintenance was required until the very end of various cutoffs.
All syllogisms have three parts, therefore this is not a syllogism.
It's perhaps somewhat ironic (and even unfortunate in a sense) that the two first TAS blocks were kind of too awesome for their (or our) own good. As Mothrayas said, they probably gave too many people the wrong impression that TASing is all about hacking and breaking games and making them do funny things (rather than, you know, speedrunning and completing the games). Those expectations are hard to follow and, well, not really what it's all about.
This time you showed what TASing is really about, which I thought was great, but it had to live in the shadow of the two earlier "gimmick" presentations.
Those wrong impressions and expectations are also quite unfortunate in the sense that it's a speedrunning marathon (ie. completing games as fast as possible), not a "let's break this game and make it do funny things" marathon. So it kind of misses the whole point... (But on the other hand, the "gimmick" TASes presented something different, something new, instead of the same thing as everything else during the marathon, which may be part of the reason.)
One idea that came to my mind which would perhaps be "gimmicky" but still preserve the spirit of speedrunning would be to connect TASbot to two consoles at the same time and have the same input control two different games (ie. a multigame TAS; one input, two games). This might be technically infeasible, though.
(Also, for it to work as a presentation, the games and the TAS would need to be selected such that it's very clear visually that what's happening in the games is synchronized, ie. that it really is the same input being given to both games. The playable characters in both games need to move the same way most of the time, etc. It just doesn't look very convincing or interesting if the two games seem to be advancing very independently of each other.)
I agree with Tompa that, in future, we don't have to go for crazy total control stuff every time, but at least have new TASes to show. The problem is there's no guarantee when a TAS will be completed (you never know what's going to hold you up for weeks while you search for that extra frame you're sure you can squeeze out if you just jig the subpixels around a bit), so you can't really assign a TASer to make such-and-such a run to be unveiled at GDQ.
The problem is there's no guarantee when a TAS will be completed (you never know what's going to hold you up for weeks while you search for that extra frame you're sure you can squeeze out if you just jig the subpixels around a bit), so you can't really assign a TASer to make such-and-such a run to be unveiled at GDQ.
Moreover, it's hard to know which TASes will be good to show off at a GDQ event without having seen the completed TAS in action. It'd be a shame if "we" (the TASing community) unveiled a new groundbreaking TAS (e.g. first TAS on an as-yet unTASable platform) at a major event and, because of the way the game mechanics worked, the run was a total snoozefest. I'd say leave it up to whoever wants to run the TAS block at the event to pick good TASes, and if someone wants their TAS to be first shown at the event then they need to convince the presenter that that's a good idea.
Pyrel - an open-source rewrite of the Angband roguelike game in Python.
Moderator, Senior Ambassador, Experienced player
(898)
Joined: 9/14/2008
Posts: 1007
Derakon wrote:
thatguy wrote:
The problem is there's no guarantee when a TAS will be completed (you never know what's going to hold you up for weeks while you search for that extra frame you're sure you can squeeze out if you just jig the subpixels around a bit), so you can't really assign a TASer to make such-and-such a run to be unveiled at GDQ.
...I'd say leave it up to whoever wants to run the TAS block at the event to pick good TASes, and if someone wants their TAS to be first shown at the event then they need to convince the presenter that that's a good idea.
We unveiled new TAS's in both AGDQ events, in the sense that SMW with that payload hadn't been seen at AGDQ 2014 and there was a new, shorter discovery for AGDQ 2015 by Masterjun. But, keep in mind, we have to go through the game submissions process just like every other runner, and it's up to the GDQ selection committee to decide whether or not to accept a given game. That selection process happens months in advance, so we skirted the issue by calling it a mystery game, but that was a bit of last-minute trickery with the folks placing games on the schedule more than anything. All that to say I agree that something new is best and we will definitely be doing that for AGDQ 2015 but there are limits on how much we can deffer the game name.
Joined: 11/11/2006
Posts: 1235
Location: United Kingdom
dwangoAC wrote:
[...] I wanted to provide a place to discuss the involvement TASVideos had at SGDQ 2015 and solicit feedback.
I have a few humble suggestions:
Ensure that whatever people are working on for *GDQ, it's all in a repository somewhere for collaboration or at least access. If someone disappears with all their work on their PC only, it's basically lost :)
Reddit is a very polarized community, so It's important to not be dissuaded by the hivemind; that said, I do agree with one comment that Ikaruga was a pretty obscure game to demo. It would have been nice to have a game destroyed in a funny way; perhaps [1438] SNES International Superstar Soccer Deluxe "playaround" by Marcokarty in 15:24.38? (i.e in addition to games being beaten fast)
I agree that the past two performance of TASBot at AGDQ set an impossibly high standard, and there was a missed opportunity to cement SGDQ as the more run-oriented event for TASes. Perhaps next year at AGDQ it should be explained as quickly as possible, or right at the end that TASbot will be back for runs at SGDQ.
Other than that, I thought it was great, and I think the TASing competition is a wonderful thing that could allow potential newcomers to submit work and see it put in a league table (maybe as a vertically scrolling list as the winning result(s) is/are played?)
<adelikat> I am annoyed at my irc statements ending up in forums & sigs
Joined: 1/15/2015
Posts: 79
Location: United Kingdom
As someone who is just a watcher of TASes, here are my opinions:
I think the Ikaruga run was a bit of a flop. I noticed it didn't get nearly as much applause as the other two games. The fact that it was pre-recorded made it seem a little less real, and it might just be that I don't know the game very well, but it seemed a bit dull. I'm sure if you had played the game, you'd know how hard it was, but because it was played so well, it didn't look very hard.
I also think that a new video should have been shown. I suppose the Streemerz run sort of fit that, but that wasn't a properly made TAS. As has already been pointed out, the games were announced beforehand, and so you could easily have looked them up and watched them yourself.
I don't think it was a bad segment. The Mega Man and Sonic runs were interesting. The commentary was informative. The section at the beginning demonstrating how TASes are made was a good inclusion (if not a bit long).
I think the problem was, ironically, the TAS Block had too much actual TASing. By that, I mean that a lot of the runs from previous GDQs were not standard TASes; their goal was not to complete the game as fast as possible, but to make the game do something it wasn't supposed to. People had heard of the TAS block as 'that time when they got Snake working in Super Mario World, and Twitch in Pokémon Red'.
I think that you need to find the right balance between normal runs, and crowd-pleasing runs.
I definitely think just playing a video of Ikaruga was a bad idea. Console verification is best, but if that’s not possible, you should still at least replay runs on the emulator, and not just that, you could have extra Lua scripts to show interesting data about the run that you wouldn’t be able to see by simply going on YouTube and watching the official encode. Maybe you could even savestate during particular tricks, and go through them with frame-by-frame while explaining what’s happening. Basically, make it an experience that’s something more than what just any person can go on YouTube to see.
As I was writing the post I started getting more ideas and now I think that actually playing back runs on emulator and using Lua/input viewer/save&loadstate and explaining the more important tricks could be a pretty valid alternative to console verification.
Mini-tutorial: I thought this was a good idea, but not explained very well. I know time was a factor, but I think a pre-recorded video would have been better here for this very reason.
TAS showcases: I didn't see the point of showing pre-recorded videos for a live event. For Sonic Advance it made sense because of the cam hack next to the console verification, but not for the other games. Also, the commentator for Ikaruga did a really bad job, and seemed to try to remember more than anything. Overall, I felt that it was average compared to other SGDQ runs.
Competition: Great game choice! The time limit was just right in my opinion, and I really enjoyed fighting against the hard deadline (vs the high probability of cheaters in a honor system)! I did not expect to rank this high (31st for win condition) considering that it was my first TAS. Seeing two of the best runs compete against each other at the end and seeing where I messed up was pretty cool, even if they didn't end up being the real winners. (By the way, for the room where dwangoAC mentioned that they were using different strategies, I actually tried both routes during the competition and I ended up using the same amount of frames.) Aside from the technical difficulties, the only thing that I felt went wrong was that the rules on TASVideos were not entirely accurate, which made me lose some time at the beginning due to the confusion. I hope a similar competition will be organized in the future! :D
We may have seen existing TASes, and most of the speedrunners at the event might have seen then, but I'm quite certain that the vast majority of the hundreds of thousands of viewers have not. Thus I don't think it's that big of a deal to show existing TASes. We shouldn't be judging the interest of the TAS block based on our personal experience, because we are biased.
It would be nice if the TAS block presented things that have never seen before by anybody, and thus surprises everybody, but that can be hard to do every time, and perhaps a bit unrealistic. (If it's eg. just some regular game completion TAS, then a long time would probably need to be spent to make it, and kept under the wraps possibly for months, if it's supposed to be a "surprise" revealed at the marathon. Is it worth just for a few minutes of "surprise"?)
I quite enjoyed the whole thing but didn't expect Ikaruga to be played through mpv there! It's fine to make things differently in these *GDQ events so you're not repeating yourself and I can totally agree with that. To be absolutely honest with myself; I'm still somewhat the same along the lines what ALAKTORN said. Let's hope this thread helps you out!
Moderator, Senior Ambassador, Experienced player
(898)
Joined: 9/14/2008
Posts: 1007
Tompa wrote:
One downside with this TAS presentation was that there were practically no surprises
This was (more or less) deliberate, with the only "surprise" to the general public being that Sonic Advance was console verified, but people are used to that. I anticipate that SGDQ events will have far fewer surprises in general but proper messaging both in the pitch, on the schedule, and at the beginning of the block to set expectations correctly is required. For AGDQ, there should be some serious surprises if at all possible.
Regarding coments from various people about the beginning explanation. I agree, the explanation went on too long. The primary reason for this is I had memorized a sequence where pressing right at frame 142 led to predictable scores but after re-watching the video I see I somehow had select held down and that prevented it from working, which completely threw me off and ended up in something like 7 cycles. Then I was so shaky from not feeling well that I kept hitting wrong buttons and it took me far longer than it should have to show manipulating enemy movement and trying to explain "Hey, if you're a realtime runner and there's a section that's giving you trouble, here's how you can practice it". This could have definitely gone better. I think at least some kind of short intro (whether it's interactive or not) is needed, but it should be closer to 45 seconds long, not 7 minutes long.
I had such a huge fear of repeating the humiliation of a desync for a third event in a row that I played it extra extra safe and opted to use an encode rather than hoping that Dolphin didn't desync. I now regret this decision - had I properly tested Ikaruga with the beta versions of Dolphin and gained some confidence in it working it probably would have been just fine but indeed Ikaruga was a video encode and I didn't properly explain that part (I was about to I can see from re-watching the video, but I was interrupted and never got to that part). We didn't get to explaining why it wasn't possible to console verify until so late into the (admittedly somewhat long) run that most people probably missed it, and my failure to get the timer working didn't help matters. I don't plan on using video encodes again because of the backlash from this, but I stand by my choice for Ikaruga - while a lot of people found it boring it also received a near $2k donation and lots of viewers enjoyed it. It was a bit regrettably long, though.
Thanks to everyone for the feedback - it *is* useful, as I'm starting to see some clear trends in things people care about. Keep it coming if you have more!
The only problem I had with the TAS block is that the commentary during the Ikaruga section concentrated entirely on the run, as if it were a regular speedrun, and did nothing to explain things about TASing in general, such as what is involved in creating a two-person TAS.
I think that showing a slight bit of the process of creating a TAS was a good idea. Personally I didn't find that part too long or boring. (I noticed that the number of frames needed to wait before starting the level was unexpected by you, but that didn't bother me. I thought it was interesting rather than awkward or bothersome.)
Perhaps in the next GDQ marathon, something else could be briefly demonstrated, such as how to find the memory address for a certain value, how to show said value on screen in real-time, and how to use it for something useful. I think some good example of this could be showcased and explained in a minute or two.
(I know this is sounding a bit like a TASing tutorial, but hey, if it's interesting, so what? It can entice people to try it.)
That's actually not a bad idea, now that I ponder over it. Most of the explanations of what TASing is as given during other runs as SGDQ only went so far as to say "Using frame advance and save states". Which while that's the most obvious bit, isn't the be all and end all of the process.
Anymore, what we do is we disassemble the game and study the physics, we search for loopholes and programming errors, and we watch what the game is thinking. Probably the "best" examples of what this can result in are the ACE runs and ACMLM's "A Bored God Plays Tetris". Another thing that we could help clarify is the difference between actual randomness and deterministic RNG - I saw a lot of people on twitch chat complain that it's not random if the same result can be forced during the 100% Chrono Trigger run. If we can include an explanation of that (probably best if we can console verify one of the more controversial Monopoly movies for this purpose) then it might make things easier on other runners as well.
Another thing that we could help clarify is the difference between actual randomness and deterministic RNG - I saw a lot of people on twitch chat complain that it's not random if the same result can be forced during the 100% Chrono Trigger run.
Yeah, I guess that would be good. Would need to be scripted out properly though, I doubt anyone could give a full, understandable explanation of (P)RNGs on the fly.
Warp wrote:
Perhaps in the next GDQ marathon, something else could be briefly demonstrated, such as how to find the memory address for a certain value, how to show said value on screen in real-time, and how to use it for something useful. I think some good example of this could be showcased and explained in a minute or two.
Another thing that we could help clarify is the difference between actual randomness and deterministic RNG - I saw a lot of people on twitch chat complain that it's not random if the same result can be forced during the 100% Chrono Trigger run. If we can include an explanation of that (probably best if we can console verify one of the more controversial Monopoly movies for this purpose) then it might make things easier on other runners as well.
Lets see, how about this:
The first thing you[the audience] have to understand is that we don't know if there is such a thing as randomness outside of Philosophy or information theory; nor is it very relevant in practice. In the real world, what is used is unpredictability: if something is too hard to predict, we can use it for pseudo-randomness. High quality unpredictable data is supposed to be indistinguishable from "true" random data without computational capabilities we don't have, and may never achieve.
There is one problem: high quality unpredictability is slow; typically, a few bits per second if you are using a typical commercial "hardware number generator". So you need a method to extrapolate from small quantities of high quality unpredictable data to large volumes of reasonable quality unpredictable data.
This is done with pseudo-random number generators (PRNG). A PRNG is a small program that generates a crazy sequence of numbers by a deterministic method. Its only claim to unpredictability is that it keeps its internal state secret; without knowing it, you generally can't know what the next output will be given the current output. The initial state of a PRNG is called a seed; modern computers and operating systems use high quality unpreditcable data to seed a PRNG, and periodically reseed it with more unpredictable data.
Video games — especially older ones — generally have very few good sources of unpredictable data; the chief among those is the player. Moreover, older console games generally didn't reseed the PRNG — they either chose a seed based on when the player started the game, or had a fixed seed. Thus, the only real source of unpredictability used was the player. By removing all unpredictability from the player, you remove all unpredictability from these games.
Moderator, Senior Ambassador, Experienced player
(898)
Joined: 9/14/2008
Posts: 1007
marzojr wrote:
...By removing all unpredictability from the player, you remove all unpredictability from these games.
Hm. Might be a bit too technical and verbose.
That text is definitely a bit too deep but yes, I think I could do a better job of explaining the general concept. Where did your quote come from, by the way? Just curious.
