And here I was wondering if the previous run was provably optimal, only for a counterexample to to drop minutes later.
In the submission text, 100thCoin mentions that it's neat that the optimal version of the game is now the same version as in the TAS that inspired the creation of the site. I'd go a little farther, and say that it's poetic that the game and version that inspired the idea of perfect play, is the one that may possibly have just achieved perfection. Fantastic work.
(Well, there's already some perfect TASes on the site, but still.)
Extremely impressive improvement, considering how seemingly close to optimal the last TAS was. Also pleased to see DPCM runs are now fully recognized as the incredible achievements that they are. I've always loved TASes and challenges that optimize for a single goal so strongly that the "gameplay" becomes something entirely new that emerges from the constraints.
Also, 19 frames is so fast that I start to wonder if you could formally prove whether this is in fact the fastest possible completion, similar to what's been done with Dragster. Seems barely possible, but would probably take a very large amount of effort to prove.
Ok, feel free to disregard the above runs. Like I said, not the main point.
As arbitrary code execution speedruns go (as opposed to playarounds, which are a whole different beast), I would rank this as easily in the top 10% on the site for entertainment, maybe at the very top. It sounds like a joke to say that there "isn't a moment of downtime in the run", but it really is true that there's inherent entertainment value in what gets skipped. If the game worked slightly differently, this could have been another ACE run that takes a few random-looking actions in the first level then jumps to credits. Don't get me wrong, I still find those runs entertaining, but at this point it would be one of many. And certainly it would have been accepted without controversy in that case. Instead this run stands alone by improving over all those existing runs to as great a degree as they improved over the traditional full game runs that came before them. And it is precisely that lack of additional setup required that provides the extra factor of interest, and yes, entertainment, above all ACE runs that came before it.
Now, while they may not have been written with a run anything like this in mind, I think it would still be informative to see what happens if we apply the voting guidelines here anyway:
Entertainment
While the entertainment level of a TAS is somewhat subjective, there are a few principles that most of our players and viewers have agreed upon that make for runs with high entertainment.
An entertaining run should:
- Be fast-paced. If it takes five minutes for anything to happen, nobody will want to watch your run. Games with fast characters are well-suited for this.
This run achieves this goal better than any other in existence. Perfect by this metric.
- Be varied. Watching a character do little more than run to the right and jump occasionally is not entertaining. Generally, the more possibilities for what your character can do at any given time, and the more of them that are used, the better.
Ambiguous. No time passes, so variance over time isn't a meaningful concept. Low or middling by this metric depending on your interpretation.
- Do the unexpected. Abuse of glitches and otherwise causing the game to perform in unexpected ways makes for far more interesting gameplay.
I don't think a single person on the planet expected anything like this could be possible before ais's research into controller polling. Perfect by this metric.
- Have lots of action. The more that is going on at the same time the more impressive the result will be. Just don't overdo it to the point where it is impossible to follow the gameplay.
A lot is certainly going on at once, though it is in fact impossible to follow.
(Hypothetically, it would be possible to make an encode with Quicksilver-vision, which is to say, displaying inputs at a constant rate and slowing down the video correspondingly when necessary. The video would still require explanation, but once you understood the general idea it you could then follow the video in a meaningful sense.)
Based on the general philosophy suggested by the above guidelines, I'd consider 2.2 far out of line, let alone the 0.0 entertainment some have rated it as. Entertainment should never be measured in units of time.
Finally, I'd like to appeal to variety. This site has countless runs of various sorts of traditional gameplay, as it certainly should. At this point, it even has a decent collection of "traditional ACE" runs, as incredible as it is that such a thing can even exist. There is nothing else like this, it seems to me that TASVideos could only benefit as a site from not excluding such a complete gamechanger.
As a slight extra factor, I'd like to point out that there is in fact precedence in Moons for TASes that optimize the entire gameplay away but have strong merits of their own.
http://tasvideos.org/1145M.htmlhttp://tasvideos.org/4754S.html
Though, I don't think that's nearly as important a point as the sheer significance and ingenuity of this run. It stands today as the answer to the question of just how far it is possible to go via TAS - the answer is, at least in the case of one game, total control of the console at any time whatsoever via inputs alone. And based on what I've read on the investigations into other games, I expect this SMB3 will still hold that position of "most extreme TAS possible" ten years from now, though time will tell. This run will and should be remembered, the knee-jerk reaction not so much.
Another way of looking at it - this run is one of the exceptional few that can truly claim to have "solved" the game in question (SMB1 any% being the other famous example). There's a certain artistic merit to achieving perfection itself.
Extremely disappointed to see this run - by my view, the most historic TAS ever - getting vaulted. Imagine if you all hadn't known about this from AGDQ already, just how blown away you would have been by the submission. And certainly there's plenty of general interest in a run such as this, the rather considerable news coverage back at the original GDQ reveal speaks to that.
If nothing else, the amount of entertainment per second of runtime is absolutely off the charts.
I've been thinking of doing a run which would be fairly iffy by the usual TASVideos guidelines, but I think would nevertheless be valuable in its own right. Specifically, I'm thinking of an ACE run of the Spaceworld prototype for Pokemon Gold and Silver.
Now, this prototype doesn't actually have an ending programmed, so my proposal is to recreate a mock one using the ACE itself, maybe something along the lines of porting the credits from the finished game. (Optionally, there could be further playing around for entertainment prior to this.)
I do realize that the TASVideos rules normally forbid playing on a prototype game when a final release exists, however in this case I feel it's reasonable to argue that the final game is not just a finished version of the prototype, when 50% of the engine and Pokemon, plus 100% of the maps, were scrapped in between the two. Just about the only thing that would remain the same compared to the final game is the player character's sprite. (Even the console isn't entirely the same, since they switched from the SGB to the GBC mid-development.) Certainly the glitches available to use are very different, as ACE takes 30 minutes in the final game but is looking likely to be possible in under 5 thanks to some glitches inherited from the first generation.
So what I want to know is, could a run like this be considered acceptable for publication, if it proved sufficiently entertaining? Or would the choice of game (a prototype without an ending) disqualify it regardless?
A Virtual Console game is simply another release of a game. The fact that it happens to be implemented via embedding an emulator and ROM is nothing more than an implementation detail. If the Wii emulation is good enough, no reason why it shouldn't be allowed.
Having a payload already setup before the timing of a TAS even starts is, in my opinion, entirely unreasonable. Instead, there's two possible ways to go with this:
1) Require the save file to be set up within the GB Tower. (This doesn't yet have emulator support.)
2) Allow for the submission of a single run that involves playing two different games in sequence. This hasn't been done before, but is not in my opinion much different from playing red+blue simultaneously.
