I wholly support the publication of board game TASes, regardless of what rule or tier changes end up being made to support them.
In general, I've found runs on easy AI difficulties more entertaining than runs on hard. In my perfect world, these would be separate categories and could potentially both have runs published, if the differences were interesting enough to support both. In a situation where neither is particularly entertaining, I'd say we want the fastest run only.
Since goal choice is not always as obvious as it is for other games, I feel like each game should get an allowance for one published run of a non-moon-worthy goal. If someone submits such a run, and then later someone else submits another run of the same game that has a different goal, but still isn't moon-worthy, it could obsolete the old one if people think it's better. This way we end up with a consensus-canonical run of every game, but avoid an absurd proliferation of boring categories from people just trying to get published easily.
A warb degombs the brangy. Your gitch zanks and leils the warb.
There's been a few points voiced in the GB Tennis submission thread, that sound very true to feelings of some people.
http://tasvideos.org/forum/viewtopic.php?p=400464#400464
A Demo tier would sound best as a tier for research and development achievements. Right? So, R&D being a direction, a movie aiming there must have 2 critical qualities: optimality and impressiveness. Right?
Now a question. Is a Fool's Mate in Chess impressive, even if it's optimal? Developing a bot that can TAS something is impressive obviously, or making a research on how to complete some absurdly random game, or whatever. But what about the easiest/fastest way to beat any chess party?
If it doesn't look impressive, it can not be added to such a Demo tier. If we still want it there, we should change the "impressiveness" criterion. Sure, if tricking a simple AI into a Fool's Mate is not impressive, tricking some sophisticated one into it may be, but that would mean we'd need to choose games that fit into Demo tier. Which makes it not too reachable for some board games probably.
Please guys let's think over this paradox carefully.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
Something along the lines of "R&D" could be a good name for the demo category. I prefer "category" instead of "tier" because I don't consider one category to be superior over another (except stars, of course), and I would rather have them noted as different groups. The categories need a slight redefinition to reduce subjectivity. Without giving names to the categories, here is how I would define them:
The speedrun category is for TASes that have speed as the main goal rather than entertainment. (Clearly defined any% and 100%)
The entertainment category is for TASes that have entertainment as the main goal rather than speed, and are deemed entertaining. There should not be any mixing between this and speedrun, which is what the current system does.
The star category is for TASes that are the best on the site. (This will remain the same) This could work as a concurrent category so that it is easy to tell what type of starred TAS it is.
The demo category is for runs that are interesting technical achievements (but not trivial or simplistic), but are not deemed entertaining enough to be in that category, and do not fit in the speedrun category. There are a number of rejected TASes that can fit in this category, most notably the gruefood delight.
As for the chess example about early mates, it depends on how it is accomplished. I could make a 2-player version, but it would not be any real accomplishment, nor would beating up a simplistic engine. But if someone could get a notable engine to get an early mate (or at least relatively quick in terms of number of moves), that would be very impressive, but not entertaining because of the long waiting.
I would say that if a chess (or whichever game) engine can be manipulated to make any arbitrary moves, that's not as interesting as playing an optimal game against a stronger engine that can't be affected so fundamentally.
(I would guess that most of the even slightly competent chess engines cannot be manipulated into making moves you want because, among other things, they use opening libraries, and obviously won't make bad moves at least if the player follows a standard opening. It becomes fuzzier if the player makes a move not in any standard opening, but even then I doubt a decent engine can be manipulated into making egregiously bad moves, at least if it has time to process even a modest amount of plies.)
I think the challenge should be "how fast can this engine be beaten when it's playing at its strongest". With more advanced engines it can become quite challenging because it's not anymore just a question of grinding with savestates. You really need to find the optimal moves.
Perhaps one possible way of making a long chess game more interesting for the viewer is, if the chess game is one of those that show what moves the computer is considering, to add on-screen commentary of those moves, what they look on the board, what they mean, what consequences they have, etc.
I'm against board games in general unless something really impressive can be done (eg Monopoly), but ESPECIALLY against things like chess that have a billion different versions. One run against one AI MIGHT be enough of a curiosity to slip in, but then what? The strategy and play isn't going to change for any other chess game. You're just going to get a bunch of nigh identical runs with slight variations due to different AI quirks or things like the time it takes to flip a piece. And for what? Everyone knows the AI for these games, especially older ones, is poor, and saying "it's on expert" somewhere in the description doesn't make much of a difference to the viewer experience. It might make the results somewhat more difficult to manipulate, but how many people are intimately familiar with the AIs to understand the difference in difficulty of every single chess game out there? Also, manipulating just a couple rare results is something that should go without saying in any TAS. A run that just needs to do it just a couple times with almost no other input is something for a bot to spit out on its own.
If tomorrow, some submitted some 10 odd chess TASes, I doubt it'd go over well. And to illustrate how much redundancy is possible, there are nearly 50 different mahjong games (and almost 30 pachinko!) on the SNES alone. How much variation do you expect is between all of them?
Joined: 4/8/2005
Posts: 1573
Location: Gone for a year, just for varietyyyyyyyyy!!
Sorry for not having time to read the whole conversation. A random idea popped into my head anyway, so I'll share it. Expert gurus can see whether it goes anywhere or not:
Maybe there could be a new rule for board games: It needs to look different than what an actual game on a real board would look like. In other words, if it just looks like a simulation of an ordinary game and there is nothing else, it is not accepted. It needs to look different or do something different than just move pieces and end. This might solve a chunk of cases while being a relatively well defined concept. Maybe do something like that? No need for new tier, because runs that are more than simulations of board game matches should have enough interesting qualities that they fit into the current tiers. I don't know. Good luck!
Joined: 10/28/2013
Posts: 130
Location: United States
Tangent wrote:
One run against one AI MIGHT be enough of a curiosity to slip in, but then what? The strategy and play isn't going to change for any other chess game. You're just going to get a bunch of nigh identical runs with slight variations due to different AI quirks or things like the time it takes to flip a piece.
This is extremely unlikely. The algorithms and opening books for 8- and 16-bit chess games vary dramatically. I can say from experience that even different iterations of the same program (e.g. Chessmaster and New Chessmaster on Game Boy) will have completely different responses to the same move, sometimes as early as move 2 or 3. And tricks or prepared lines that work on one program will fall completely flat on another, believe me.
If you're able to find an input sequence to trigger arbitrary moves by the CPU, then sure, you might get identical games. But the good news is, they'd only be 2-3 moves long anyway. :) If you can force the CPU to make the move of your choosing from the beginning, then Fool's Mate (or a similar line, e.g. 1. e4 d5 2. exd5 Qxd5 3. Ke2?? Qe4#) will end the game almost as soon as it started.
Chess TAS are a very specialist niche, but they're likely to be far subtler and more challenging than you're making them out to be. Whether they'll be entertaining to the casual viewer, well, TBH I don't really care: as I've made clear elsewhere, entertainment is a secondary priority for me.
Joined: 10/28/2013
Posts: 130
Location: United States
Warp wrote:
I would say that if a chess (or whichever game) engine can be manipulated to make any arbitrary moves, that's not as interesting as playing an optimal game against a stronger engine that can't be affected so fundamentally.
Not necessarily, or at least I'd be interested in a TAS that (for instance) finds a way to manipulate RAM directly to deceive the CPU into a bad line. Given what can be pulled off with games like International Superstar Soccer Deluxe, I'd like to think at least one poorly-programmed chess program out there can be smacked around with a storm of unexpected inputs at the right time.
Warp wrote:
(I would guess that most of the even slightly competent chess engines cannot be manipulated into making moves you want because, among other things, they use opening libraries, and obviously won't make bad moves at least if the player follows a standard opening. It becomes fuzzier if the player makes a move not in any standard opening, but even then I doubt a decent engine can be manipulated into making egregiously bad moves, at least if it has time to process even a modest amount of plies.)
If the CPU isn't being directly manipulated through RAM corruption etc., you're not going to get it to move 1. f3 and 2. g4. OTOH, you can absolutely take advantage of bad opening book programming to get the CPU to move instantly for 10-12 moves, only to end up in a lost position. That would actually be ideal, since you fly through the game and then wrap it up quickly. And opening books have certainly been known to contain errors and busted lines of play, though I don't immediately know of any so egregious as to walk the CPU into immediate checkmate.
Beyond that, there are known "anti-computer" styles of play, obviously not effective against current engines, that trick the CPU into accepting sacrifices that a human player can see will lose, but whose consequences just exceed the computer's move depth. I know of one major chess publication in the 1990s that had a column specifically dedicated to making computers look stupid, and Tim Krabbé has also done some investigating in that direction.
If tomorrow, some submitted some 10 odd chess TASes, I doubt it'd go over well. And to illustrate how much redundancy is possible, there are nearly 50 different mahjong games (and almost 30 pachinko!) on the SNES alone. How much variation do you expect is between all of them?
This seems like a more pressing issue than chess differences. I'm not sure how to draw the line for cases like that. :o
I for one would enjoy watching a chess TAS, not sure if the concept is really viable but it's an interesting idea. Whether it's against a really powerful chess engine or even just something from like an old DOS game, it could be good either way.
Although I think it would be a lot harder to make against a powerful engine because you probably couldn't manipulate it as well and you would also obviously have to be quite competent at chess yourself. Watchability might also be less because I think with more advanced engines they actually take a bit of time to calculate their moves, and at such high levels of play unless the viewer is himself a master chess player a lot of the brilliance and complexity of the moves, and hence of the TAS, would be lost on them... unless, perhaps, like others have said, the TAS contained some kind of commentary to explain the logic behind them and the impact each move has on the game.
For those reasons maybe a TAS, not of the game of chess itself, but of any old video game that happens to be about chess would be more in line with TASing. I don't doubt in such cases there would be enough things that an average chess-playing TASer could discover and exploit to create entertaining videos. I can imagine scenarios like quickly capturing all of the computer's pieces without losing any of the players', or only using one specific piece to capture throughout the whole game, or the player losing almost all of their pieces early on only to come back and recapture all of the computer's pieces or just simply win by checkmate despite being severely low and outnumbered in material.
tl;dr: TAS chess when???
Joined: 10/12/2011
Posts: 6335
Location: The land down under.
TheYogWog wrote:
tl;dr: TAS chess when???
BLAH. BLAH. BLAH.
Refer to my post on this thread about a serious Chess TAS that was submitted and could be obsoleted with Godlike RNG (for Black win).
WebNations/Sabih wrote:
+fsvgm777 never censoring anything.
Disables Comments and Ratings for the YouTube account.Something better for yourself and also others.
BLAH. BLAH. BLAH.
Refer to my post on this thread about a serious Chess TAS that was submitted and could be obsoleted with Godlike RNG (for Black win).
Yeah but why not a run that's less speedrun-y and more superplay-y? Like Mortal Kombat, sort of.
Actually, a good Chess playaround would be to manipulate the AI of a weak engine into following the moves of an actual famous game between grandmasters (while entering the moves of the other grandmaster).
Joined: 10/28/2013
Posts: 130
Location: United States
Why don't we look at a simpler premise: can anyone extract the opening book from Chessmaster? If we can find an error -- a line that leads directly to a forced mate, ideally -- then we can exploit it, most of the CPU's moves will be instantaneous, and we won't need to dumb it down in order to get a quick win.
Beating the game on Newcomer/1 difficulty (as in the run posted earlier) establishes a "fastest possible" run, I suppose. But I'd like to see the CPU taken down at its best by exploiting its flaws (and no, the menu command that tells it to move immediately isn't one of them).
Why don't we look at a simpler premise: can anyone extract the opening book from Chessmaster? If we can find an error -- a line that leads directly to a forced mate, ideally -- then we can exploit it, most of the CPU's moves will be instantaneous, and we won't need to dumb it down in order to get a quick win.
Beating the game on Newcomer/1 difficulty (as in the run posted earlier) establishes a "fastest possible" run, I suppose. But I'd like to see the CPU taken down at its best by exploiting its flaws (and no, the menu command that tells it to move immediately isn't one of them).
There's a guideline for hardest difficulty, and facing Chessmaster instead of Stanley the monkey or something certainly follows that
Those bots are usually greedy, so if you can find an opening with a sacrifice (thinking they are ahead in material), they will likely fall for that trap.
I bought Absolute Chess for the DSiWare and despite using an opening book, it'll fail miserably after it ends (because of sacrifices).
Why don't we look at a simpler premise: can anyone extract the opening book from Chessmaster? If we can find an error -- a line that leads directly to a forced mate, ideally -- then we can exploit it, most of the CPU's moves will be instantaneous, and we won't need to dumb it down in order to get a quick win.
Beating the game on Newcomer/1 difficulty (as in the run posted earlier) establishes a "fastest possible" run, I suppose. But I'd like to see the CPU taken down at its best by exploiting its flaws (and no, the menu command that tells it to move immediately isn't one of them).
That still won't solve what to do with the many different chess games that exist out there; do they obsolete each other, or all published despite being quite similar to one another?
Joined: 10/28/2013
Posts: 130
Location: United States
Anty-Lemon wrote:
There's a guideline for hardest difficulty, and facing Chessmaster instead of Stanley the monkey or something certainly follows that
Agreed. Just to be clear I was referring to the run posted earlier in the thread, done on Newcomer/1 difficulty, which is the default. On that difficulty I'm pretty sure Chessmaster plays either the first move that comes into its "head", or in any event searches no more than 1 move (2 ply) ahead.
jlun2 wrote:
That still won't solve what to do with the many different chess games that exist out there; do they obsolete each other, or all published despite being quite similar to one another?
There's basically no chance that the same line will be optimal against different programs, any more than the same set of inputs would work in RPGs or platformers (pipe down, Mega Man!), so this should be a non-issue. I can say from experience that a prepared line doesn't work even against different revisions of the same program, e.g. Chessmaster and New Chessmaster for Game Boy.
It's 100% trivial to defeat any of these old, 8- and 16-bit chess programs; heck, I've done it, even on the hardest difficulty, and I'm not a master-strength player. Anyone can download a free chess engine that'll destroy any vintage chess program, not to mention any of us. :)
It's not a matter of finding a single brilliant line that works for every program -- that's not going to happen -- but rather of figuring out the weaknesses specific to each program's opening book, algorithms, and/or UI. Each program will almost certainly be unique in that department, so the trick IMHO is finding the shortest path to victory in any particular program, either in terms of time or in moves, against the CPU at its max setting.
That's normally a very slow setting, but there's a way around it: as long as the CPU is in its opening book (tables of prepared opening moves), it normally moves instantaneously. An approach that exploits that is likely to be both effective and quick; find a data error, or a line that's been erroneously included that leads to an immediate forced mate, and you've struck gold.
If a game-breaking vulnerability can't be found in the opening book, then memory corruption might let you trick the CPU into playing the "wrong" move, e.g. by getting it to look at the wrong part of its opening book for its next move. Since 1. f4 is a normal opening move (Bird's Opening), find some way to get it to play 2. g4, and you've got Fool's Mate. But that's a long shot.
Joined: 10/28/2013
Posts: 130
Location: United States
Meant to reply to this too:
samurai goroh wrote:
Those bots are usually greedy, so if you can find an opening with a sacrifice (thinking they are ahead in material), they will likely fall for that trap.
I bought Absolute Chess for the DSiWare and despite using an opening book, it'll fail miserably after it ends (because of sacrifices).
Wow, I'm surprised that a recent engine is vulnerable to that! With the most current high-end chess engines, you're starting to see games with very deep sacrifices that seem more strategic than tactical, since they don't lead to immediate mate or regaining the material. That's terrifying and inspiring, all at once. Here's a recent example between Stockfish and Jonny.
But yeah, positional sacrifices are often lethal against older or weaker software. I don't think it'd work for a TAS, though, since converting the advantage would usually take many moves with long CPU cogitation. If the distance to mate exceeds the horizon effect, then the CPU will think about each move along the way.
The trick to keeping the time low is to make the distance between opening book and mate as short as possible, I think.[/b]
Add http://tasvideos.org/forum/viewtopic.php?p=405806 to the list (with its rejection reason "As Jenga is a board game, this game also does not qualify for the Vault.")
I still think we desperately need a category for these games. What's in essence "(entire game genre) does not qualify for publication" feels completely contrary to the spirit of the site, IMO. If it does not qualify for vault, then create a category where it does.
Was there any consensus on what to do with the multiple variations of the same board game? Currently, whenever that was asked, the only game that was pointed to was Chess. What about Mahjong? Or even Jenga for that matter? How many seemingly duplicate submissions of similar games are allowed?
tl;dr - Stop talking about chess and take account onto other board games with a million ports please.
Putting this here because I accidentally (sort of) posted it in a gruefood thread instead. Whoops.
z1mb0bw4y wrote:
Mothrayas wrote:
Patashu wrote:
Jenga is a board game? RIP
This is actually something I did some research into when I judged this movie. It turned out that yes, Jenga is indeed commonly categorized as a board game.
From wikipedia,
"A board game is a game that involves counters or pieces moved or placed on a pre-marked surface or "board", according to a set of rules. Games can be based on pure strategy, chance (e.g. rolling dice), or a mixture of the two, and usually have a goal that a player aims to achieve. Early board games represented a battle between two armies, and most modern board games are still based on defeating opposing players in terms of counters, winning position, or accrual of points (often expressed as in-game currency)."
Tetris is a game in which pre-defined pieces are moved about a pre-defined board and stacked in such a way as to accrue points.
http://tasvideos.org/4423S.htmlhttp://tasvideos.org/1065S.htmlhttp://tasvideos.org/2786S.htmlhttp://tasvideos.org/2459S.htmlhttp://tasvideos.org/2471S.html
I'm sure there are more, but I'm lazy. I find it ABSOLUTELY HILARIOUS that the common argument against board games is "wut r we gona do wit all the chess gaemz?" when we still accept tetris clone after tetris clone after tetris clone.
Rejecting someone's hard work and effort because "it's a board game" is a farce of an excuse.
There's actually been some discussion about that on IRC. The main point why the system we were suggesting there won't work is because the intention of tiers is to promote. Indeed, once we remove the entertainment criterion from tier division, there's nothing to promote, and no way to do so. Right now, even with how faulty the system may be, it's promoting very well:
Newcomer attention -> Stars -> Moons -> Vault -> Gruefood Delight.
It was suggested, that Vault game-based criterion is lowered, Moons entertainment criterion is raised, and a middle tier is added, a to-be-defined tier. It'd be the primary one to send accepted runs to, to allow some automation when deciding what tier to put a run to, deciding between Vault, Middle and Moons depending on rating: http://tasvideos.org/Adelikat/FourTierSystemRules.html
Adelikat, myself, Moth and some other people were in favor of that system. Then adelikat seemed to lost interest.
Adelikat: to answer your recent question on IRC: in board games, people agree to account fastest possible runs as any%, and hardest difficulty as 100%. That way chess would only have 2 runs maximum per game.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
