Skip to content

Zulko/easyAI

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

easyAI

EasyAI (full documentation here) is a pure-Python artificial intelligence framework for two-players abstract games such as Tic Tac Toe, Connect 4, Reversi, etc. It makes it easy to define the mechanisms of a game, and play against the computer or solve the game. Under the hood, the AI is a Negamax algorithm with alpha-beta pruning and transposition tables as described on Wikipedia.

Installation

If you have pip installed, type this in a terminal

sudo pip install easyAI

Otherwise, download the source code (for instance on Github), unzip everything into one folder and in this folder, in a terminal, type

sudo python setup.py install

Additionally you will need to install Numpy to be able to run some of the examples.

A quick example

Let us define the rules of a game and start a match against the AI:

from easyAI import TwoPlayerGame, Human_Player, AI_Player, Negamax

class GameOfBones( TwoPlayerGame ):
    """ In turn, the players remove one, two or three bones from a
    pile of bones. The player who removes the last bone loses. """

    def __init__(self, players=None):
        self.players = players
        self.pile = 20 # start with 20 bones in the pile
        self.current_player = 1 # player 1 starts

    def possible_moves(self): return ['1','2','3']
    def make_move(self,move): self.pile -= int(move) # remove bones.
    def win(self): return self.pile<=0 # opponent took the last bone ?
    def is_over(self): return self.win() # Game stops when someone wins.
    def show(self): print ("%d bones left in the pile" % self.pile)
    def scoring(self): return 100 if game.win() else 0 # For the AI

# Start a match (and store the history of moves when it ends)
ai = Negamax(13) # The AI will think 13 moves in advance
game = GameOfBones( [ Human_Player(), AI_Player(ai) ] )
history = game.play()

Result:

20 bones left in the pile

Player 1 what do you play ? 3

Move #1: player 1 plays 3 :
17 bones left in the pile

Move #2: player 2 plays 1 :
16 bones left in the pile

Player 1 what do you play ?

Solving the game

Let us now solve the game:

from easyAI import solve_with_iterative_deepening
r,d,m = solve_with_iterative_deepening(
    game=GameOfBones(),
    ai_depths=range(2,20),
    win_score=100
)

We obtain r=1, meaning that if both players play perfectly, the first player to play can always win (-1 would have meant always lose), d=10, which means that the wins will be in ten moves (i.e. 5 moves per player) or less, and m='3', which indicates that the first player's first move should be '3'.

These computations can be speed up using a transposition table which will store the situations encountered and the best moves for each:

tt = TranspositionTable()
GameOfBones.ttentry = lambda game : game.pile # key for the table
r,d,m = solve_with_iterative_deepening(
    game=GameOfBones(),
    ai_depths=range(2,20),
    win_score=100,
    tt=tt
)

After these lines are run the variable tt contains a transposition table storing the possible situations (here, the possible sizes of the pile) and the optimal moves to perform. With tt you can play perfectly without thinking:

game = GameOfBones( [  AI_Player( tt ), Human_Player() ] )
game.play() # you will always lose this game :)

Contribute !

EasyAI is an open source software originally written by Zulko and released under the MIT licence. Contributions welcome! Some ideas: AI algos for incomplete information games, better game solving strategies, (efficient) use of databases to store moves, AI algorithms using parallelisation.

For troubleshooting and bug reports, the best for now is to ask on Github.

How releases work

Every time a MR gets merged into master, an automatic release happens:

  • If the last commit's message starts with [FEATURE], a feature release happens (1.3.3 -> 1.4.0)
  • If the last commit's message starts with [MAJOR], a major release happens (1.3.3 -> 2.0.0)
  • If the last commit's message starts with [SKIP], no release happens.
  • Otherwise, a patch release happens (1.3.3 -> 1.3.4)

Maintainers