Animal Chess

Animal Fight Chess Game（斗兽棋） written in rust, using Alpha-Beta-Pruning algorithm, and implement AlphaZero algorithm for training.

How To Play

To win the game, one player must successfully move any animal into the Den（兽穴） of the opponent. See rules:

• http://ancientchess.com/page/play-doushouqi.htm
• https://en.wikipedia.org/wiki/Jungle_(board_game).

How To Run

It need rust nightly version.

Prerequisite

$ rustup default nightly

Play

$ git clone https://github.com/netcan/AnimalChess.git
$ cd AnimalChess
$ cargo run --release

Developing

$ sudo apt-get install python3-dev python-dev
$ pip3 install virtualenv ipython
$ cd pymodule
$ virtualenv venv
$ source venv/bin/activate
(venv) $ pip install -r requirements.txt
(venv) $ maturin develop
(venv) $ ipython
In [1]: import animal_chess_pymodule

In [2]: board = animal_chess_pymodule.Board('l5t/1d3c1/r1p1w1e/7/7/7/E1W1P1R/1C3D1/T5L w - - 0 1')

In [3]: board
Out[3]:
l #@# t
 d # c
r p w e
 ~~ ~~
 ~~ ~~
 ~~ ~~
E W P R
 C # D
T #@# L

In [4]: board.generate_all_steps()
Out[4]:
[24688,
 24673,
 24656,
 25202,
 25187,
 25185,
 25716,
 25701,
 25699,
 26230,
 26198,
 26213,
 29057,
 29042,
 29025,
 29040,
 30085,
 30070,
 30053,
 30068,
 32897,
 32880,
 34422,
 34437]

In [5]: board.move_chess(24688)

In [6]: board
Out[6]:
l #@# t
 d # c
r p w e
 ~~ ~~
 ~~ ~~
 ~~ ~~
  W P R
EC # D
T #@# L

Todo

• Seperate game.rs to gui.rs and board.rs
• Add Monte Carlo Tree Search Algorithm
• benmark board operation
• export module for python3
• add some test
• benmark pymodule
• encode/decode move
• generate fen
• repr board
• patch sdl2 to support load_texture from buffer