Big Chess

[Will Taylor]

Hi all,

I'm interested in knowing whether it would be possible to program a decent engine to play FICGS Big Chess. The game is similar in most aspects to normal chess, with the most significant changes being a much bigger, 16x16 board, and 32 pieces for each side. I would have thought the large board would make it really difficult, but as I'm not a programmer I don't know and would be interested to hear your opinions.

Thanks,

Will

P.S. The starting setup can be seen at http://www.ficgs.com/help.html (scroll down some way).

[hgm]

It would not be more difficult than building an engine for standard Chess. It would not search as deep, of course, because of the larger branching factor. WinBoard would require some changes in the move parser to handle double-digit rank numbers before it could be used as a GUI.

[Will Taylor]

Thanks for the reply. How big a problem do you think not being able to search as deeply would be? Would it really cripple the engine's strength, or could it still be a GM killer?

[hgm]

No idea what strength would pass as a GM in this game... Engines also come in all kinds. A 2000-Elo engine can be built in a week, a 3000-Elo engine will take significantly longer (if you ever get there). A 1500-Elo engine could be built in a day...

[Greg Strong]

Thanks for the reply. How big a problem do you think not being able to search as deeply would be? Would it really cripple the engine's strength, or could it still be a GM killer?

The change would be huge. A typical Chess midgame position has about 35 legal moves. In Big Chess it would be over a hundred. Then, consider that the size of the search tree goes up exponentially with greater depth. So the total reachable depth would be massively reduced. On the other hand, you're compairing the computer's strength to a human's, and people have similar problems trying to see far ahead in situations with a multitude of options ...

I can say that search depth goes down significantly even with small increases. My program ChessV can't search anywhere near as deeply even in Christian Freeling's Grand Chess (10 x 10 board) as in standard Chess. The program can, in theory, be programmed to play variants up to 16 x 16, but the degredation of search depth is so pronounced that I've never added support for anything larger than 10x10, although I do plan to add Wildebeest Chess (11 x 10) someday soon given how much faster computer have become.

[Alex Lobanov]

I have a test engine for bigchess.
Im used standard methods, without optimizations. Search depth 7-10 plies for middlegame.

[smrf]

10x8 chess based on Capablanca's extended piece set might be an approach, where some competitors will keep the whole thing more interesting.

[zamar]

The change would be huge. A typical Chess midgame position has about 35 legal moves. In Big Chess it would be over a hundred.

I believe that aggressive LMR can solve most of the problems. In bigchess board there must be much more "garbage moves" than in usual chess board. Order non-captures with history or gain table heuristic and reduce very aggressively at the end of the list.

The engine not to be named was the first to demonstrate that this works extremely well with usual chess and I'd expect this approach to work even better with Bigchess.

Of course to tune everything right could be work of months or years.

[FICGS]

I have a test engine for bigchess.
Im used standard methods, without optimizations. Search depth 7-10 plies for middlegame.

Hi all,

As you may know, I introduced Big Chess at FICGS as a game that could be similar to the Go (weiqi) game by its complexity... Go engines are now quite strong but still cannot beat strong players (and are always completely crushed by the best human players)...

The making of a strong Big Chess engine is a great challenge according to me... of course most moves are often totally useless but in most positions many moves look equal, strategy takes a bigger place and one need to understand the positions (e.g. queen traps in 20 moves or so, endings and so on), that's the point. And of course the piece values are quite different from Chess and vary during the game...

I'll be happy to organize Big Chess test matchs at FICGS between the best players & big chess engines ! Just email me if you're interested.

[Don]

I think this would be the perfect game to program on a 256 bit computer using bit boards.

Hi all,

I'm interested in knowing whether it would be possible to program a decent engine to play FICGS Big Chess. The game is similar in most aspects to normal chess, with the most significant changes being a much bigger, 16x16 board, and 32 pieces for each side. I would have thought the large board would make it really difficult, but as I'm not a programmer I don't know and would be interested to hear your opinions.

Thanks,

Will

P.S. The starting setup can be seen at http://www.ficgs.com/help.html (scroll down some way).