Kramnik and AlphaZero: How To Rethink Chess (chess.com) 60
Vladimir Borisovich Kramnik, a Russian chess grandmaster, writing at Chess.com: The increasing strength of chess engines, the millions of computer games and the volumes of opening theory available to every player are making top-level chess less imaginative. Decisive games in super-tournaments have declined, while the number of games with what I'd call "creative" content is also on the slide. [...] From my own experience, I know how difficult it has become to force a complex and interesting fight if your opponent wants to play it safe. As soon as one side chooses a relatively sterile line of play, the opponent is forced to follow suit, leading to an unoriginal game and an inevitably drawish outcome. Of course, there are still some fascinating top-level games being played, but to keep chess alive, I believe we must reverse this trend before the game's spirit fades away.
So I started thinking, if the outcome is always the same, perhaps there's something we can do to reinvigorate the game. I spoke with Demis Hassabis, the founder and CEO of the artificial intelligence lab DeepMind. [...] Working with DeepMind researchers Ulrich Paquet and Nenad Tomasev, we used AlphaZero as a petri dish to test different variations and see how the game might unfold. Ultimately, our mission was to find an adjustment to the rules to allow more space for human creativity. [...] My aim was to find a chess variant that would not only have the potential to bring the excitement and decisive victories back to chess, but is also aesthetically pleasing. The goal was to reignite interest and introduce players and audiences to the immense complexity and creativity of the original game of chess.
To begin, we tasked AlphaZero with exploring a variant that prevented either side from castling, trying different opening moves from both sides. The outcome was beyond our expectations! We let AlphaZero learn how to play "no-castling chess" from scratch, allowing the program to incrementally learn how to master the game through a process of trial and error, similar to how it taught itself to play classical chess. After playing millions of games, AlphaZero became a no-castling expert, allowing us to analyze how it plays and assess the overall game balance. The win/loss percentages for both White and Black are similar to classical chess, suggesting that the no-castling variant should be quite playable without favoring a particular player. Preventing the king from retreating to a safe distance means that all of the pieces have to engage in the melee, making the play more dynamic and entertaining, with a number of original patterns.
So I started thinking, if the outcome is always the same, perhaps there's something we can do to reinvigorate the game. I spoke with Demis Hassabis, the founder and CEO of the artificial intelligence lab DeepMind. [...] Working with DeepMind researchers Ulrich Paquet and Nenad Tomasev, we used AlphaZero as a petri dish to test different variations and see how the game might unfold. Ultimately, our mission was to find an adjustment to the rules to allow more space for human creativity. [...] My aim was to find a chess variant that would not only have the potential to bring the excitement and decisive victories back to chess, but is also aesthetically pleasing. The goal was to reignite interest and introduce players and audiences to the immense complexity and creativity of the original game of chess.
To begin, we tasked AlphaZero with exploring a variant that prevented either side from castling, trying different opening moves from both sides. The outcome was beyond our expectations! We let AlphaZero learn how to play "no-castling chess" from scratch, allowing the program to incrementally learn how to master the game through a process of trial and error, similar to how it taught itself to play classical chess. After playing millions of games, AlphaZero became a no-castling expert, allowing us to analyze how it plays and assess the overall game balance. The win/loss percentages for both White and Black are similar to classical chess, suggesting that the no-castling variant should be quite playable without favoring a particular player. Preventing the king from retreating to a safe distance means that all of the pieces have to engage in the melee, making the play more dynamic and entertaining, with a number of original patterns.
Mandate the Bongcloud (Score:3)
If every GM had to use the bongcloud opening every game, it would increase the silliness in games immensely.
Re: (Score:3)
Better yet, just make the following changes:
1. An array of 8 boards making cubical 3D chess, with all movements extended into the 3rd dimension
2. Random spots labeled as portals, allowing teleportation to other boards at random at the same square using an 8 sided dice to pick the board to teleport to.
3. Mirrored sides on pieces with lasers for an additional way to remove pieces from the game.
That ought to randomize things enough to bring the creativity back to strategy.
There are already many chess variations (Score:5, Insightful)
There are already umpteen fun chess variations that humans can play easily and have a lot of fun. The best way to enjoy chess is to play other humans and just throw the computer algorithms in the waste bin.
Re:There are already many chess variations (Score:5, Insightful)
However if you are playing with someone who is really into chess, they have learned strategies which they fall on.
Most of us who play chess for fun, know the rules of the game, but we come up with our own strategy.
It is like a program written by a self taught programmer who sat down and learned all the commands the language has to offer. Vs someone who got some formal learning in programming. They are taught methods and ideas on how to use groups of commands to create solutions that are more optimal then the half hazard coding style from a bunch of learned commands.
If there is a lot of data to search for all the time, I know to order the list, then do a binary search from the middle thus getting my results done in log time. While someone who just learned the language, will often just loop down all the elements to find the value, if they are smart they stop when they find it, or more common they will finish searching all the data and go on.
Yes the self taught programmer may reinvent the binary search them-self, but the formal training that includes strategy will often create more consistent approaches to the problem.
Chess is the same way. There are so many formal strategies that the game isn't really a test for your mind, but a practice in following a pre-learned strategy.
Re: (Score:3, Interesting)
A solved game is solved. Competition level play is pointless once it is solved, and will become boring because it will become repetitive.
Re: (Score:3, Funny)
The best way to enjoy chess is to play other humans and just throw the computer algorithms in the waste bin.
The best way to enjoy chess is to play other humans and just throw the pieces at the other player.
Re: (Score:2)
Man, three full paragraphs of droning on about how to make Chess more fun, and all they came up with was removing one of the special rules. Such amazing creativity!
I think we should make chess more fun by replacing the boring grid with a more sophisticated map full of halls and rooms and obstacles, and also replace the boring dark ages pieces with much more interesting pieces like wizards and dualists and zombies and giant wolves and so on....and change the combat machanics from simple move-to-kill to some
Re: (Score:3)
Man, three full paragraphs of droning on about how to make Chess more fun, and all they came up with was removing one of the special rules. Such amazing creativity!
I think we should make chess more fun by replacing the boring grid with a more sophisticated map full of halls and rooms and obstacles, and also replace the boring dark ages pieces with much more interesting pieces like wizards and dualists and zombies and giant wolves and so on....and change the combat machanics from simple move-to-kill to something more varied with spells to cast and die rolls based on offensive and defensive stats based on what gear the pieces have equipped at the moment.
And then top it all off by adding a sense of progression through the games, and an engaging fanciful narrative too.
THAT would make chess a lot more interesting.
Why bother with that when we already have D&D?
Re: (Score:2)
Whoosh.
Re: There are already many chess variations (Score:2)
I like this oneâ"
Www.yavoch.com
Yavoch: Yet Another Variation On CHess.
Computers are the only creative chess players now (Score:5, Interesting)
After years of stagnating, AlphaZero did a great job at making people re-think the value of position and mobility of pieces vs just the value of material.
Sure, humans have always been good at making short-term material sacrifices for position.
But AlphaZero takes this to a whole new level; valuing bishops on long diagonals that don't have obvious payoffs for 20+ moves.
Because of opening book memorization, Chess has been boring for a long time. Bobby Fischer had some good rants on the topic, leading to Fischer Random Chess - which seems to have some of the benefits OP described.
https://en.wikipedia.org/wiki/... [wikipedia.org]
Re: (Score:2)
This isn't about enjoying chess per se; it's about keeping competitive chess alive. How do you do that when computers can precompute out all the lines of play, and you have no way of preventing competitors from studying those results before a match?
Re: (Score:2)
There are already umpteen fun chess variations that humans can play easily and have a lot of fun. The best way to enjoy chess is to play other humans and just throw the computer algorithms in the waste bin.
How about the MarioKart version of chess? Every 10 moves, randomly appearing item boxes appear to make pieces disappear, appear, or move; make parts of the board temporarily opaque, promote non-last rank pieces, etc. Yes, and make Mario and Bowser pieces on the board.
More seriously, how about bypassing the rote openings by randomly selecting a midgame position? This forces players to rely more on on-the-fly reasoning and less on memorization. For example, a computer could select a few (say five position
Re: (Score:2)
It was called "Fairy Chess" and it's been around almost since the game started, that name being applied more around the closing of the Victorian era. It has a lot of fanciful new pieces in it like the grasshopper, and one of the variants is the 3D chess made popular(well, among certain circles) in ST-TOS. The notion of throwing away the computers isn't really useful since once the genie is out of the bag... I think Kramnik's point was to make a subtle variation and see how that opened up the game without
let's play global thermonuclear war (Score:2)
let's play global thermonuclear war
I'm surprised it took this long. (Score:1)
Re: (Score:2)
There is some difference between Tic-tac-toe and chess other then game complexity.
Tic-Tac-Toe has a fixed set of moves where all possible game play options are available. Chess you can have both sides kings just iterating back and forth forever as a possible game play option. So all possible options cannot be calculated, thus a shortcut in the algorithm is needed. Being that the game could last a long time, part of the gameplay is wearing out your opponents attention. Two equally good players may still be
Re: (Score:2)
Chess you can have both sides kings just iterating back and forth forever as a possible game play option.
Umm, no, there's the three repetitions rule, the 50 moves rule, and the dead position rule, all of which keep a chess game from going forever.
Re: (Score:2)
That's not true. A game where no draw condition exists will never end in a draw.
Here are the possible end states for a two player solved game with two perfect players, A and B:
a) the game will always end in a draw
b) player A will always win
c) player B will always win
d) the game will never end
Re: (Score:2)
Possibility d) is not possible in chess, as a game of chess can be declared a draw after fifty moves with no pawn moves or captures. Since each player can only possibly make 48 pawn moves and 15 captures, it follows that a game that does not end some other way must eventually be a draw by the fifty-move rule.
Re: (Score:2)
So the trick to win, is to not play chess but gladiatorial combat to force your opponent to "resign" from the game.
Re: (Score:2)
So the trick to win, is to not play chess but gladiatorial combat to force your opponent to "resign" from the game.
https://en.wikipedia.org/wiki/... [wikipedia.org]
So in other words... nerfing (Score:2)
Just like computer games, people are seeking to nerf chess pieces - starting with the Castle.
Welcome to the same iterative world e-sports gamers inhabit, Chess Masters.
Re: (Score:2)
Just like computer games, people are seeking to nerf chess pieces - starting with the Castle.
Welcome to the same iterative world e-sports gamers inhabit, Chess Masters.
You're suggesting that changing rules to make a game more interesting started with video games? Professional sports are constantly being iterated.
Re: (Score:3)
Next: queens, rooks, and bishops will be limited to moving a maximum of 4 spaces at a time.
Not going to work (Score:4, Insightful)
The advantages all have to do with this line of play being less understood. Once opening theory is developed all the advantages will go away: the number of "play it safe" games will go up once it is understood how to reach safer positions quickly; the number of decisive games will decline commensurate with the level of understanding of the opening theory; and the beginners will soon enough face seasoned players who will know opening theory by heart and will have no chance.
This is just like dating a woman you just met - new, fresh, and exciting for the first week or so, increasingly boring and even irritating after.
Re: (Score:2)
The experiment showed the contrary, or at least aimed to see if that would be the case.
By letting AlphaGo play normal chess for millions of games, it did converge to a rather boring situation.
By letting AlphaGo play no-castling chess for a similar number of games, it converged to more interesting situations, or at least that's what they say.
AlphaGo is simulating years of evolution here and it seems to end up being more interesting. And tweaking games rules can make a huge difference. You want to make gam
Re: (Score:2)
The experiment was not clearly quantified. I am not sure one can conclude anything from the vague promotional statements by a biased inventor and proponent of these changes. More importantly, it seems clear that with enough study, major patterns will emerge in opening theory and Kramnik at the end of the article acknowledges this. So the most one can claim here is that it will take more simulations and more study than with regular chess to get to the same situation.
Castling (Score:2)
It's been about twenty years since I gave a damn about chess. Yes it got very boring very quickly when creativity vanished.
Removing castling sounds even more boring. Castling was always an exception, since it's not always available.
What about a restriction. Restrictions breed creativity. Maybe just one of the following:
** Can't move the same piece twice in consecutive turns **
** Can't move the same piece more than N times in a game **
** Can't move the same type of piece twice on consecutive turns
Re: (Score:2)
I think their preference is to increase the number of possible moves rather than reduce them.
Re: (Score:2)
That's the silly part. Creativity is always about how you use less.
Re: (Score:2)
Creativity is always about how you use less.
That is why the most creative games have only 1 possible move, right?
Your statement about creativity is true only in things that already have infinite possibilities. It forces people to explore possibilities that they may not have considered before. It works with painting, coding, or architecture. But it does not apply to a system with a fixed set of rules. You can get much creativity if you say "Paint whatever you want, but you must do it only using red." That works because there is an infinite number
Re: (Score:2)
The problem is not human creativity. It's that computers are analyzing out all the possible moves. Reducing the number of possible moves makes that more of a problem.
Opening wook of moves (Score:2)
"We found if, rather than a guaranteed capture, the pieces had to fight with random rolls based on powers, the outcomes could not be predicted. This would dovetail in the modern computer world because the pieces could be animated. A large creature with a hammer could fight a much larger creature, and sometimes the larger one would pick up the smaller one and smash him to the ground, winning. Another, later time, the one with the hammer could smash the larger one on the head, winning. These are just exam
Re: (Score:2)
Gee, that sounds familiar. [wikipedia.org]
Re: (Score:2)
but it would still be advisable to let the wookie win . . .
hawk
960? (Score:2)
What happened to Fischer AKA 960 Chess? I thought that was the Next Big Thing to make the game less computer dominated.
Fischer Random Chess (Score:5, Informative)
Didn't Bobby Fischer foresee this about 20 years ago, being utterly bored of normal chess, and invent Fischer Random where the home rank placement is randomized with certain restrictions?
I mean, it is amazing that a game with so vast a state space has begun to fall thanks to study and computation. Adding something like three orders of magnitude to the complexity should help goose the game, no?
From the Wikipedia article on the subject --
Fischer's goal was to eliminate what he considered the complete dominance of openings preparation in classical chess, replacing it with creativity and talent. His belief about Russians fixing international games also provided motivation. In a situation where the starting position was random it would be impossible to fix every move of the game. Since the "opening book" for 960 possible opening systems would be too difficult to devote to memory, the players must create every move originally. From the first move, both players must devise original strategies and cannot use well-established patterns. Fischer believed that eliminating memorized book moves would level the playing field.
Short term, maybe (Score:2)
# of possible chess positions: 10^43 - 10^50
# of possible go positions: ~ 2*10^170, and AI holds its own with pros.
That's a looong way to go, and randomizing the starting point for the AI shouldn't take long.
Might invigorate contests between meatbags for while, though.
Cites:
https://www.chess.com/forum/vi... [chess.com]
http://tromp.github.io/go/lega... [github.io]
Re: (Score:2)
This is fascinating and I never knew about this variation and the reasoning. Back in the 90's I loved chess and wrote chess games for the early handhelds that were out then, e.g., Handspring Visor, Compaq iPaq, T-Mobile Sidekick, etc. My focus was on the front-end and for the back-end, I used GNU Chess. The trick was to run GNU Chess on a server hosted in "the cloud" (actually a rack server at a local ISP) and have it equipped with a huge "book" of opening moves and games - something that would never have f
Re: (Score:2)
Most games humands have devised are solvable. Chess and Go just have a huge game state that it takes an immense amount of memory and computational power to compute the solution for.
Contrast this with say, Tic-Tac-Toe which is simple enough to solve in a few minutes.
And yes, most games are solvable. Even ones of pure chance in which case the solution is the one that increases your odds of success des
Re: (Score:2)
Why does castling make a game "sterile"? (Score:2)
What is it about castling that would cause a game to be "sterile" and "drawish"??
Re: (Score:2)
It is a quick, easy and effective way to move your king to a well-defended position, blunting your opponent's attacks.
Kramnik has recently retired from competition (Score:2)
Kramnik has already admitted that he wants a career away from chess:
https://chess24.com/en/read/ne... [chess24.com]
Now that he has retired from competition it is time to say that "We have to rethink chess". That means, that _others_ have to rethink chess, not himself. While he was still playing he did not recommend any such drastic measures as the old way was working for him.
The feats of AlphaZero are a good excuse, but when he lost to Deep Fritz back in 2006
https://chess24.com/en/read/ne... [chess24.com]
having missed checkmate-in-one
Solution: Stop taking it so seriously (Score:5, Insightful)
Chess can be fun if it's played between two reasonably intelligent people who know the rules but never cared enough to bother studying the game.
It's just like any other game. The more seriously a person takes it, the more fun is sucked out of it.
Re: (Score:2)
Studying chess helps one appreciate the nuances and there is no end to nuances in chess for a lowly human. Appreciating nuances is fun whether you win or lose! I refer you to youtube 'Magnus Carlsen 1 minute bong cloud attack" for some serious fun that could only be had by serious study.
Re: (Score:2)
Chess can be fun if it's played between two reasonably intelligent people who know the rules but never cared enough to bother studying the game. It's just like any other game. The more seriously a person takes it, the more fun is sucked out of it.
Not taking it seriously enough can be just as much of a downer, like not making the least bit of effort to role play in a D&D session or to play as a team in a team game. Yes, it can get too serious too but if it doesn't have some element of skill, strategy or teamwork and is just luck or button mashing then it becomes boring real quick. Being challenged and mastering the game is a large part of what's gaming is about and you can't really have that without the risk of someone going overboard trying to a
Why not randomize the back row? (Score:2)
Would picking a random back row configurations advantage one color over the other?
Bring on the Monsters! (Score:2)
How has no one mentioned Chess Boxing yet? (Score:2)
https://en.wikipedia.org/wiki/... [wikipedia.org]
Chess is deterministic (Score:2)
The only reason chess i
Re: (Score:2)
Indeed - as I was reading this I was thinking to myself that chess was getting boring because people know enough now to start playing "GTO" chess. I find poker much more interesting too.
In poker, GTO stands for Game Theory Optimal - a theoretical strategy (as in the complete strategy does not exist as yet, though individual situations can be 'solved' ) for playing that cannot be exploited by another player. At its root it is a defensive strategy - while it cannot be exploited, it also cannot exploit anot
But the machine still wins (Score:1)
I find the Kramnik's definition of success at inventing a more creative kind of chess funny: "After playing millions of games, AlphaZero became a no-castling expert, allowing us to analyze how it plays and assess the overall game balance."
The minute the players, human or otherwise, begin feeding games of the newer variety into their "modded" chess algorithms, it's game over, and we're back to square one.
The test of a truly "creative" kind of chess is where the machine can only achieve a win-loss ratio simi
so they want to invigorate interest in chess (Score:2)
by switching to not-chess?
A better tournament tie breaker (Score:3)
Since the 2018 World Chess Championship [wikipedia.org] ended in a draw between Carlsen and Caruana which was then decided by a "rapid chess" perhaps this would make for a more interesting alternative tie breaker. I like the idea of removing castling and changing the dynamics of the game in tournaments when both sides play it safe and boring. I imagine chess purists would disagree saying it isn't chess anymore but something does needs to change or the game will stagnate.
Perhaps something similar to Archon (Score:2)
It was essentially a chess game, except that in order to capture a piece, you had to fight it out, video game style.
More powerful pieces had more health and stronger attacks, but it was not impossible for a lowly Pawn-like piece to defeat a powerful Queen-like piece.
The board itself would go through light/dark cycles, giving the appropriate sides bonuses in fights.
For example, when the board is on the dark cycle, the dark pieces were