alyshmahell / gym-chess

A simple chess environment for openai/gym

gym-chess

A simple chess environment for gym. It computes all available moves, including castling, en-passant, pawn promotions and 3-fold repetition draws.

8	♜	♞	♝	♛	♚	♝	♞	♜
7	♟	♟	♟	♟	♟	♟	♟	♟
6
5
4
3
2	♙	♙	♙	♙	♙	♙	♙	♙
1	♖	♘	♗	♕	♔	♗	♘	♖
	a	b	c	d	e	f	g	h

Setup

Install the environment:

pip3 install git+<URL_FOR_THIS_REPO>  

Play against yourself or against a random bot.

import gym
import gym_chess

env = gym.make('ChessVsRandomBot-v0')
env = gym.make('ChessVsSelf-v0')

Play

You can either get available moves and convert them into the action space, or get the available actions directly.

import random

# moves
moves = env.get_possible_moves(state, player)
m = random.choice(moves)
action = env.move_to_actions(m)

# actions
actions = env.get_possible_actions(state, player)
action = random.choice(actions)

A move is a dictionary that holds the id of the piece (from 1 to 16 for white and -1 to -16 for black), its current position and its new position.

move = {
    'piece_id': <int>,
    'pos': [<int>,<int>],
    'new_pos': [<int>,<int>]
}

NB: when a move is converted to an action and vice-versa, the 'pos' attribute is dropped (it doesn't provide any useful information since the current position of a given piece is recorded in the current state).

Pass an action to the environment and retrieve the new state, reward, done and info:

state, reward, done, __ = env.step(action)

Reset the environment:

initial_state = env.reset()

Visualise the chess board

Visualise the current state of the chess game:

env.render()

Show the moves that a piece can make given a set of moves (by any/all pieces on the board)

# Player 1 moves
piece = 6 # white queen
env.render_moves(state, piece, moves, mode='human')

You can also retrieve the list of squares that pieces are attacking and defending by specifying the "attack" option:

attacking_moves = env.get_possible_moves(state, player, attack=True)