boardgame2 0.0.5

2-player zero-sum board game extension for OpenAI Gym.

boardgame2

boardgame2 is an extension of OpenAI Gym that implements multiple two-player zero-sum 2-dimension board games, such as TicTacToe, Gomuko, and Reversi.

Environments

• Reversi-v0
• KInARow-v0, as well as Gomuku-v0 and TicTacToe-v0
• Go-v0 (Experimental)

Install

pip install --upgrade boardgame2

Usage

Create a Game

import gym
import boardgame2

env = gym.make('TicTacToe-v0')
env = gym.make('Gomuku-v0')
env = gym.make('KInARow-v0', board_shape=5, target_length=4)
env = gym.make('KInARow-v0', board_shape=(3, 5), target_length=4)
env = gym.make('Reversi-v0')
env = gym.make('Reversi-v0', board_shape=6)
env = gym.make('Go-v0')

Play a Game

import gym
import boardgame2

env = gym.make('TicTacToe-v0')
print('observation space = {}'.format(env.observation_space))
print('action space = {}'.format(env.action_space))

observation = env.reset()
while True:
    action = env.action_space.sample()
    observation, reward, done, info = env.step(action)
    if done:
        break
env.close()

API

Constants

boardgame2.BLACK

The 1st player.

boardgame2.WHITE

The 2nd player.

boardgame2.EMPTY

Neither of the two players.

Functions

boardgame2.strfboard

strfboard(board:np.array, render_characters:str='+ox', end:str='\n') -> str

Format a board as a string

boardgame2.is_index

is_index(board:np.array, location:np.array) -> bool

Check whether a location is a valid index for the board

boardgame2.extend_board

extend_board(board:np.array)

Get the rotations of the board. Only valid for square board.

Classes

boardgame2.BoardGameEnv

The base class of all board game environment.

__init__(board_shape, illegal_action_mode:str='resign', render_characters:str='+ox', allow_pass:bool=True) -> boardgame2.BoardGameEnv

Constructor. board_shape can be either an int or (int, int).

seed(seed=None) -> NoneType

See gym.Env.seed().

reset() -> tuple

See gym.Env.reset(). observation is in the form of (np.array, int).

step(action:np.array) -> tuple, float, bool, dict

See gym.Env.step().

render(mode:str='human')

See gym.Env.render().

is_valid(state:tuple, action:np.array) -> bool

Check whether the action is valid for current state.

get_valid(state:tuple) -> np.array

Get all valid locations for the current state.

has_valid(state:tuple) -> bool

Check whether there are valid locations for current state.

get_winner(state:tuple)

Check whether the game has ended. If so, who is the winner.

get_next_state(state:tuple, action:np.array) -> tuple

Get the next state.

next_step(state:tuple, action:np.array) -> tuple, float, bool, dict

Get the next observation, reward, done, and info. Similar to gym.Env.step().

PASS

Intend to skip, as a predefined constant action.

RESIGN

Intend to resign, as a predefined constant action.

boardgame2.KInARowEnv (registered as KInARow-v0, as well as Gomuku-v0 and TicTacToe-v0)

__init__(board_shape, target_length:int=3, illegal_action_mode:str='pass', render_characters:str='+ox') -> boardgame2.KInARowEnv

boardgame2.ReversiEnv (registered as Reversi-v0)

__init__(board_shape, render_characters:str='+ox') -> boardgame2.ReversiEnv

boardgame2.GoEnv (registered as Go-v0, Not fully implemented)

__init__(board_shape, komi:float=0., allow_suicide:bool=False, illegal_action_mode:str='pass', render_characters:str='+ox') -> boardgame2.GoEnv