Chess variants made with help from alpha zero article

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
mmt
Posts: 274
Joined: Sun Aug 25, 2019 6:33 am
Full name: .

Re: Chess variants made with help from alpha zero article

Post by mmt » Sat Sep 12, 2020 6:48 pm

I forgot about checkers. There are also board games like Catan, Monopoly, Risk. Not really a board game but poker is also big - ties are rare.

jp
Posts: 1392
Joined: Mon Apr 23, 2018 5:54 am

Re: Chess variants made with help from alpha zero article

Post by jp » Sun Sep 13, 2020 3:14 am

mmt wrote:
Sat Sep 12, 2020 6:48 pm
I forgot about checkers. There are also board games like Catan, Monopoly, Risk.
I think it's insulting to chess to compare it with Catan, Monopoly, Risk, etc. Who is the World Monopoly Champion?

The only game on the previous list that truly has no draws is Backgammon -- and that fact is not obvious.

mmt
Posts: 274
Joined: Sun Aug 25, 2019 6:33 am
Full name: .

Re: Chess variants made with help from alpha zero article

Post by mmt » Sun Sep 13, 2020 6:23 am

jp wrote:
Sun Sep 13, 2020 3:14 am
I think it's insulting to chess to compare it with Catan, Monopoly, Risk, etc. Who is the World Monopoly Champion?
A popular board game is a popular board game and randomness, complex rules, and more than 2 players make these games complex in a different way. You can just as well say that chess is not a good game because it's deterministic, has only 2 players, and has no hidden states. Or from the AI point of view, you can say that computers have exceeded humans' abilities very early in chess so it's no longer interesting.
jp wrote:
Sun Sep 13, 2020 3:14 am
The only game on the previous list that truly has no draws is Backgammon -- and that fact is not obvious.
If the draws are rare enough then it doesn't really matter.

jp
Posts: 1392
Joined: Mon Apr 23, 2018 5:54 am

Re: Chess variants made with help from alpha zero article

Post by jp » Sun Sep 13, 2020 8:26 am

mmt wrote:
Sun Sep 13, 2020 6:23 am
A popular board game is a popular board game
I don't think popularity should be the deciding factor, but FWIW the popular opinion is certainly that these games (Monopoly, etc.) are not to be taken seriously.

mmt wrote:
Sun Sep 13, 2020 6:23 am
Or from the AI point of view, you can say that computers have exceeded humans' abilities very early in chess so it's no longer interesting.
Chess ceased to be interesting from the AI point of view half a century ago when it became obvious producing strong moves had nothing to do with the goals of AI.

But the public still buys the myth linking AI and chess engines.

mmt wrote:
Sun Sep 13, 2020 6:23 am
If the draws are rare enough then it doesn't really matter.
I don't know what the draw rates are in Go (7 komi), XIangqi, checkers, etc. If the rate is too low, I'd say that's a bad feature!

Raphexon
Posts: 299
Joined: Sun Mar 17, 2019 11:00 am
Full name: Henk Drost

Re: Chess variants made with help from alpha zero article

Post by Raphexon » Sun Sep 13, 2020 9:40 am

jp wrote:
Sun Sep 13, 2020 8:26 am
mmt wrote:
Sun Sep 13, 2020 6:23 am
A popular board game is a popular board game
I don't think popularity should be the deciding factor, but FWIW the popular opinion is certainly that these games (Monopoly, etc.) are not to be taken seriously.

mmt wrote:
Sun Sep 13, 2020 6:23 am
Or from the AI point of view, you can say that computers have exceeded humans' abilities very early in chess so it's no longer interesting.
Chess ceased to be interesting from the AI point of view half a century ago when it became obvious producing strong moves had nothing to do with the goals of AI.

But the public still buys the myth linking AI and chess engines.

mmt wrote:
Sun Sep 13, 2020 6:23 am
If the draws are rare enough then it doesn't really matter.
I don't know what the draw rates are in Go (7 komi), XIangqi, checkers, etc. If the rate is too low, I'd say that's a bad feature!
Xiangqi and checkers are similarly drawish to chess.

Go with 7 komi is maybe 1-2% draw rate.

mmt
Posts: 274
Joined: Sun Aug 25, 2019 6:33 am
Full name: .

Re: Chess variants made with help from alpha zero article

Post by mmt » Mon Sep 14, 2020 5:19 am

jp wrote:
Sun Sep 13, 2020 8:26 am
Chess ceased to be interesting from the AI point of view half a century ago when it became obvious producing strong moves had nothing to do with the goals of AI.

But the public still buys the myth linking AI and chess engines.
I wouldn't go this far now that AlphaZero has appeared. We don't really have a clue what might lead to AGI but reinforcement learning of neural nets and starting from just the rules of the game is certainly interesting in machine learning. Once some advancement already exists it seems less impressive in retrospective.

Tord
Posts: 5
Joined: Tue Feb 27, 2018 10:29 am

Re: Chess variants made with help from alpha zero article

Post by Tord » Mon Sep 14, 2020 11:06 am

As a lover of chess variants myself, I greatly enjoyed this paper. The selection of variants is not exactly the one I would have made, but they do all have the advantage of being playable with a normal chess set: No fairy pieces or differently shaped boards are required. My favorite of those variants would probably be torpedo chess.

Concerning a few other chess variants discussed in this thread and elsewhere:

Grand chess: I love the bishop+knight and rook+knight pieces. Being able to discover entirely new tactical motifs and mating patterns and studying new classes of endgames are so much fun. I don't like the complicated promotion rules. The empty back rank and the pawns starting on the third rank give attacks and king safety a very different feel compared to classical chess, I'm not sure how I feel about this. It's a little bit unfortunate that the rook+knight pieces often get exchanged very quickly after a symmetrical development in the opening. Perhaps the game could be improved by using a rotationally symmetric rather than a mirror symmetric starting position. Like other large chess variants, grand chess is unfortunately somewhat unergonomic. It would be necessary to either have a physically larger chess board or physically smaller pieces, both of which is less than ideal, especially in fast games.

Capablanca chess: Most of what I said about grand chess applies here as well. Compared to grand chess, king safety feels more chess-like, which I think is an improvement. I also prefer the simpler and more traditional castling rules. The non-square board feels a little unaesthetic.

Seirawan chess: Probably my favorite 8x8 variant. I'd love to see this get the same kind of adaptation as Chess960. It has obvious similarities to Capablanca chess, but the standard board size is both better looking and more ergonomic. The main downside is that the board feels a little too crowded before the first few pieces get exchanged off.

Crazyhouse: Fun, but gives white a too big opening advantage, and chess pieces are just a little too powerful for a chess variant with drops (shogi with its larger board and more slow-moving pieces work better with drops). Impractical to play with regular chess equipment, especially at fast time controls.

Chess960: My least favorite variant, and unfortunately the only one that has reached any kind of adaptation. With both the board and the pieces being exactly like in regular chess, it just doesn't add anything to the game.

Hexagonal chess: The two very similar variants by Władysław Gliński and Dave McCooey are both very attractive (a little known fact is that Glaurung –- the precursor of Stockfish –– started it's life as an engine that could play both normal chess and Gliński's hexagonal chess, using almost exactly the same code). I love the geometry of the board, discovering the new abilities the traditional chess pieces gain through the shape of the board, and discovering all the new tactics, mating patterns and endgames. McCooey's variant has more chess-like pawn captures, but I don't like only seven pawns per side. In both variants, the board feels a little too empty. I have tried to find a good opening setup with Capablanca's pieces (the rook+knight and the bishop+knight) added, but I was never able to come up with something I liked. Like the large board variants discussed above, 91 cell hexagonal boards are a little too large to be ergonomical.

User avatar
gbtami
Posts: 378
Joined: Wed Sep 26, 2012 11:29 am
Location: Hungary
Contact:

Re: Chess variants made with help from alpha zero article

Post by gbtami » Mon Sep 14, 2020 12:02 pm

Tord wrote:
Mon Sep 14, 2020 11:06 am
As a lover of chess variants myself, I greatly enjoyed this paper. The selection of variants is not exactly the one I would have made, but they do all have the advantage of being playable with a normal chess set: No fairy pieces or differently shaped boards are required. My favorite of those variants would probably be torpedo chess.

Concerning a few other chess variants discussed in this thread and elsewhere:

Grand chess: I love the bishop+knight and rook+knight pieces. Being able to discover entirely new tactical motifs and mating patterns and studying new classes of endgames are so much fun. I don't like the complicated promotion rules. The empty back rank and the pawns starting on the third rank give attacks and king safety a very different feel compared to classical chess, I'm not sure how I feel about this. It's a little bit unfortunate that the rook+knight pieces often get exchanged very quickly after a symmetrical development in the opening. Perhaps the game could be improved by using a rotationally symmetric rather than a mirror symmetric starting position. Like other large chess variants, grand chess is unfortunately somewhat unergonomic. It would be necessary to either have a physically larger chess board or physically smaller pieces, both of which is less than ideal, especially in fast games.

Capablanca chess: Most of what I said about grand chess applies here as well. Compared to grand chess, king safety feels more chess-like, which I think is an improvement. I also prefer the simpler and more traditional castling rules. The non-square board feels a little unaesthetic.

Seirawan chess: Probably my favorite 8x8 variant. I'd love to see this get the same kind of adaptation as Chess960. It has obvious similarities to Capablanca chess, but the standard board size is both better looking and more ergonomic. The main downside is that the board feels a little too crowded before the first few pieces get exchanged off.

Crazyhouse: Fun, but gives white a too big opening advantage, and chess pieces are just a little too powerful for a chess variant with drops (shogi with its larger board and more slow-moving pieces work better with drops). Impractical to play with regular chess equipment, especially at fast time controls.

Chess960: My least favorite variant, and unfortunately the only one that has reached any kind of adaptation. With both the board and the pieces being exactly like in regular chess, it just doesn't add anything to the game.

Hexagonal chess: The two very similar variants by Władysław Gliński and Dave McCooey are both very attractive (a little known fact is that Glaurung –- the precursor of Stockfish –– started it's life as an engine that could play both normal chess and Gliński's hexagonal chess, using almost exactly the same code). I love the geometry of the board, discovering the new abilities the traditional chess pieces gain through the shape of the board, and discovering all the new tactics, mating patterns and endgames. McCooey's variant has more chess-like pawn captures, but I don't like only seven pawns per side. In both variants, the board feels a little too empty. I have tried to find a good opening setup with Capablanca's pieces (the rook+knight and the bishop+knight) added, but I was never able to come up with something I liked. Like the large board variants discussed above, 91 cell hexagonal boards are a little too large to be ergonomical.
Hi Tord! try https://www.pychess.org It has Grand, Capablanca, Seirawan, Crazyhouse, Chess960 and some of the combination of them also, not to mention all the big regianal variants (and more!). The site is based on https://github.com/ianfab/Fairy-Stockfish
It would be cool if you can contribute Hexagonal chess to Fairy :wink:

Raphexon
Posts: 299
Joined: Sun Mar 17, 2019 11:00 am
Full name: Henk Drost

Re: Chess variants made with help from alpha zero article

Post by Raphexon » Mon Sep 14, 2020 12:23 pm

The start position of Capablanca is relatively garbage.
Testing it with Fairy-Stockfish gave me a fairly hefty opening advantage at 60+0.6.
White vs Black: +952 -438 =610 [0.628] 2000
There are also unprotected pawns in the beginning.


I've also tested Gothic Chess. (Capablanca but different start position)
Which was a lot more balanced: White vs black = +360 -291 =338 [0.529]
More balanced than standard chess too.

User avatar
Laskos
Posts: 10619
Joined: Wed Jul 26, 2006 8:21 pm
Full name: Kai Laskos

Re: Chess variants made with help from alpha zero article

Post by Laskos » Mon Sep 14, 2020 1:49 pm

Raphexon wrote:
Mon Sep 14, 2020 12:23 pm
The start position of Capablanca is relatively garbage.
Testing it with Fairy-Stockfish gave me a fairly hefty opening advantage at 60+0.6.
White vs Black: +952 -438 =610 [0.628] 2000
There are also unprotected pawns in the beginning.


I've also tested Gothic Chess. (Capablanca but different start position)
Which was a lot more balanced: White vs black = +360 -291 =338 [0.529]
More balanced than standard chess too.
What openings did you use?

Post Reply