liground icon indicating copy to clipboard operation
liground copied to clipboard

Enhance LiGround with the Xiangqi Stockfish engine

Open lex312 opened this issue 3 years ago • 2 comments

https://github.com/maksimKorzh/xiangqi-stockfish-template This is the one and only Xiangqi Stockfish engine which is at least 200 elo stronger than the Fairy-Stockfish Xiangqi engine.

A compile for windows would be great. One compatible exe and one exe for modern computers. For macOS Big Sur and Apple M1 chips of course too.

lex312 avatar May 15 '21 22:05 lex312

I have talked to some Xiangqi friends and they would like to see these engines:

  • Cyclone is the strongest engine with it's own neural network (the first and the only Xiangqi engine with a neural network, since 01.05.2021) http://www.ccyclone.com/
  • Xiexie Master 2.8.0 http://xiexiemaster.com/download.php
  • SA Chess http://www.sachess.com/purchase_sa_chess/

Maybe we can take some ideas from those guis too.

lex312 avatar May 17 '21 19:05 lex312

Hello @lex312 , the engine you suggested are either commercial or freeware engines. LiGround will not include freeware and commercial engines as default engines. However, you are free to add any commercial Xinagqi engine you have purchased as long as it uses the UCI protocol. USI protocol support may be added later into LiGround.

Cyclone is the strongest engine with it's own neural network (the first and the only Xiangqi engine with a neural network, since 01.05.2021) http://www.ccyclone.com/

Cyclone is neither the first nor the the only Xiangqi engine using a neural network. I suspect Cyclone uses NNUE networks now, as proposed by Nasu Yu for shogi and later integrated into Stockfish for classical chess.

Before Cyclone, the engine by NeymarL was using AlphaZero like neural networks for Xiangqi:

  • https://github.com/NeymarL/ChineseChess-AlphaZero

Recently, Xiangqi support was also added to CrazyAra:

  • https://github.com/QueensGambit/CrazyAra/pull/93

However, I haven't found the time yet to publish a release for Xiangqi, because the loss condition when repeatedly checking in Xiangqi has not been implemented yet:

  • https://github.com/QueensGambit/CrazyAra/issues/101

QueensGambit avatar May 18 '21 08:05 QueensGambit