4

Motivation:

I am trying to make a basic AI agent that can play chess against an opponent. The goal is to see how good it can become through the use of machine learning later on and also learn the fine details in chess that are hidden from us when we just play it, such as evaluation parameters.


Code:

Here is what I have so far:

import chess, chess.pgn, time, math, io
import numpy as np 

from selenium import webdriver
from selenium.webdriver.common.keys import Keys
from selenium.webdriver.common.action_chains import ActionChains
from selenium.webdriver.support.ui import Select

piece_values = {'P': 10, 'N': 30, 'B': 30, 'R': 50, 'Q': 90, 'K': 100, 'p': -10, 'n': -30, 'b': -30, 'r': -50, 'q': -90, 'k': -100}

# These are all flipped
position_values = {
        'P' : np.array([ [0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0],
                        [5.0,  5.0,  5.0,  5.0,  5.0,  5.0,  5.0,  5.0],
                        [1.0,  1.0,  2.0,  3.0,  3.0,  2.0,  1.0,  1.0],
                        [0.5,  0.5,  1.0,  2.5,  2.5,  1.0,  0.5,  0.5],
                        [0.0,  0.0,  0.0,  2.0,  2.0,  0.0,  0.0,  0.0],
                        [0.5, -0.5, -1.0,  0.0,  0.0, -1.0, -0.5,  0.5],
                        [0.5,  1.0, 1.0,  -2.0, -2.0,  1.0,  1.0,  0.5],
                        [0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0] ]),

        'N' : np.array([[-5.0, -4.0, -3.0, -3.0, -3.0, -3.0, -4.0, -5.0],
                       [-4.0, -2.0,  0.0,  0.0,  0.0,  0.0, -2.0, -4.0],
                       [-3.0,  0.0,  1.0,  1.5,  1.5,  1.0,  0.0, -3.0],
                       [-3.0,  0.5,  1.5,  2.0,  2.0,  1.5,  0.5, -3.0],
                       [-3.0,  0.0,  1.5,  2.0,  2.0,  1.5,  0.0, -3.0],
                       [-3.0,  0.5,  1.0,  1.5,  1.5,  1.0,  0.5, -3.0],
                       [-4.0, -2.0,  0.0,  0.5,  0.5,  0.0, -2.0, -4.0],
                       [-5.0, -4.0, -3.0, -3.0, -3.0, -3.0, -4.0, -5.0] ]),

        'B' : np.array([[-2.0, -1.0, -1.0, -1.0, -1.0, -1.0, -1.0, -2.0],
                       [-1.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0, -1.0],
                       [-1.0,  0.0,  0.5,  1.0,  1.0,  0.5,  0.0, -1.0],
                       [-1.0,  0.5,  0.5,  1.0,  1.0,  0.5,  0.5, -1.0],
                       [-1.0,  0.0,  1.0,  1.0,  1.0,  1.0,  0.0, -1.0],
                       [-1.0,  1.0,  1.0,  1.0,  1.0,  1.0,  1.0, -1.0],
                       [-1.0,  0.5,  0.0,  0.0,  0.0,  0.0,  0.5, -1.0],
                       [-2.0, -1.0, -1.0, -1.0, -1.0, -1.0, -1.0, -2.0] ]),

        'R' : np.array([[ 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,  0.0],
                       [ 0.5, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0,  0.5],
                       [-0.5, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.5],
                       [-0.5, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.5],
                       [-0.5, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.5],
                       [-0.5, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.5],
                       [-0.5, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.5],
                       [ 0.0, 0.0, 0.0, 0.5, 0.5, 0.0, 0.0,  0.0]]),

        'Q' : np.array([[-2.0, -1.0, -1.0, -0.5, -0.5, -1.0, -1.0, -2.0],
                       [-1.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0, -1.0],
                       [-1.0,  0.0,  0.5,  0.5,  0.5,  0.5,  0.0, -1.0],
                       [-0.5,  0.0,  0.5,  0.5,  0.5,  0.5,  0.0, -0.5],
                       [-0.5,  0.0,  0.5,  0.5,  0.5,  0.5,  0.0, -0.5],
                       [-1.0,  0.5,  0.5,  0.5,  0.5,  0.5,  0.0, -1.0],
                       [-1.0,  0.0,  0.5,  0.0,  0.0,  0.0,  0.0, -1.0],
                       [-2.0, -1.0, -1.0, -0.5, -0.5, -1.0, -1.0, -2.0]]),

        'K' : np.array([[ -3.0, -4.0, -4.0, -5.0, -5.0, -4.0, -4.0, -3.0],
                       [ -3.0, -4.0, -4.0, -5.0, -5.0, -4.0, -4.0, -3.0],
                       [ -3.0, -4.0, -4.0, -5.0, -5.0, -4.0, -4.0, -3.0],
                       [ -3.0, -4.0, -4.0, -5.0, -5.0, -4.0, -4.0, -3.0],
                       [ -2.0, -3.0, -3.0, -4.0, -4.0, -3.0, -3.0, -2.0],
                       [ -1.0, -2.0, -2.0, -2.0, -2.0, -2.0, -2.0, -1.0],
                       [  2.0,  2.0,  0.0,  0.0,  0.0,  0.0,  2.0,  2.0 ],
                       [  2.0,  3.0,  1.0,  0.0,  0.0,  1.0,  3.0,  2.0 ]])}

class LichessBot:
    def __init__(self, fen):
        self.fen = fen
        self.bot = webdriver.Firefox(executable_path=r'geckodriver.exe')

    def initialize(self):
        bot = self.bot
        bot.get('https://lichess.org/editor/rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR_w_KQkq_-')
        time.sleep(3)
        analysis = bot.find_element_by_css_selector(".actions > a:nth-child(2)").click()
        time.sleep(1)

    def gameSelect(self, fen):
        bot = self.bot

        fen_area = bot.find_element_by_class_name("analyse__underboard__fen")
        bot.execute_script('arguments[0].setAttribute("value", arguments[1]);', fen_area, fen)

        # Refresh the page to enter new fen number properly every time
        fen_new = bot.find_element_by_class_name("analyse__underboard__fen").get_attribute('value').replace(' ', '_')
        bot.get('https://lichess.org/analysis/standard/{}'.format(fen_new))

    def gameReturn(self):
        bot = self.bot

        fen_return = bot.find_element_by_class_name("analyse__underboard__fen").get_attribute('value')
        time.sleep(1)
        return fen_return

def positionEvaluation(position, piece_values=piece_values, position_values=position_values):
    # Position of pieces is not taken into account for their strength
    if position_values == 'None':
        total_eval = 0
        pieces = list(position.piece_map().values())

        for piece in pieces:
            total_eval += piece_values[str(piece)]

        return total_eval

    else:
        positionTotalEval = 0
        pieces = position.piece_map()

        for j in pieces:
            file = chess.square_file(j)
            rank = chess.square_rank(j)

            piece_type = str(pieces[j])
            positionArray = position_values[piece_type.upper()]

            if piece_type.isupper():
                flippedPositionArray = np.flip(positionArray, axis=0)
                positionTotalEval += piece_values[piece_type] + flippedPositionArray[rank, file]

            else:
                positionTotalEval += piece_values[piece_type] - positionArray[rank, file]

        return positionTotalEval

def minimax(position, depth, alpha, beta, maximizingPlayer, bestMove = 'h1h3'):
    if depth == 0 or position.is_game_over():
        return positionEvaluation(position, piece_values, position_values), bestMove

    if maximizingPlayer:
        maxEval = -np.inf
        for child in [str(i).replace("Move.from_uci(\'", '').replace('\')', '') for i in list(position.legal_moves)]:
            position.push(chess.Move.from_uci(child))
            eval_position = minimax(position, depth-1, alpha, beta, False)[0]
            position.pop()
            maxEval = np.maximum(maxEval, eval_position)
            alpha = np.maximum(alpha, eval_position)
            if beta <= alpha:
                break
        return maxEval

    else:
        minEval = np.inf
        minMove = np.inf
        for child in [str(i).replace("Move.from_uci(\'", '').replace('\')', '') for i in list(position.legal_moves)]:
            position.push(chess.Move.from_uci(child))
            eval_position = minimax(position, depth-1, alpha, beta, True)
            position.pop()
            minEval = np.minimum(minEval, eval_position)
            if minEval < minMove:
                minMove = minEval
                bestMin = child

            beta = np.minimum(beta, eval_position)
            if beta <= alpha:
                break

        return minEval, bestMin

# # To check evaluation
# board = chess.Board()
# print(positionEvaluation(board))
# quit()

# Initialize and set up position
lichess = LichessBot('rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -')
lichess.initialize()

board = chess.Board()
fen = board.fen()
lichess.gameSelect(fen)

while not board.is_game_over():
    if board.turn == True:
        print('\n[INFO] Your Turn\n=========================')
        fen_new = fen
        while fen_new == fen:
            fen_new = lichess.gameReturn()
        board = chess.Board(fen_new)

    else:
        print('[INFO] AI\'s Turn\n')
        minimaxEval, bestMove = minimax(board, 4, -np.inf, np.inf, False)
        print("AI Evaluation: {}\nAI Best Move: {}".format(minimaxEval, bestMove))
        board.push(chess.Move.from_uci(bestMove))
        print("{}\n=========================".format(board))
        fen = board.fen()
        lichess.gameSelect(fen)

This is what the code does:

  • Open firefox terminal and go to lichess.org

  • Enter the analysis mode for a starting chess position

  • Wait for human player to make a move

  • Send the FEN to the python program to make that move

  • Apply minimax algorithm with corresponding depth and position values to evaluate the position and decide the best move

  • Make this move in the python program

  • Get the FEN of the current position

  • Play the best move on the board by pasting FEN into the analysis on lichess


Question:

Right now this only lets me play as the white pieces (computer algorithm works on the black pieces). My question, although it seems basic, is how to make it so that at the start I have the choice of which side to choose? It seems like the minimax algorithm is baised towards computer playing with the black pieces and any attempt I make to adjust this failed to work.


Output:

Here is what a typical output on the console would look like while the game is going on. Nothing special happens when the game ends, I plan to include a summary of the game and outcome later on.

Console output during a game

As can be seen, I make sure to double check that the moves are correctly registered by printing the board setup position in the console output after every move.


Final Note:

I am aware the evaluation metric and maybe even the efficiency of the algorithm might not be the best but these will be adjusted once all the fine details, such as the one posted in the question, are answered.

3

There are two simple adjustments that need to be made for this to work properly with the computer playing the white pieces. I'm not totally familiar with the interfaces you're using so bear with me.

while not board.is_game_over():
    if board.turn == True:

This if statement is used to determine at which point the computer starts evaluating moves and at which point it waits for user input. By changing this to

while not board.is_game_over():
    if board.turn == False:

We can make the computer evaluate moves on white's turn instead of black's. Obviously it would be best to handle this with a value that wasn't hardcoded so that based on some user input we could determine which side the user wanted to play. I'll leave that as an exercise for the reader (since you're using selenium perhaps it's possible to detect when the player flips the board around).

Now onto the minimax algorithm. The minimax function itself is doing what it is meant to do, however your evaluation function is locked into evaluating black's position on the board. This means that even though the computer is controlling white's pieces, it is still trying to make black have the best position. This is due to the piece values and the position tables that you are using.

The simple fix for this is something like the following:

def minimax(position, depth, alpha, beta, maximizingPlayer, bestMove = 'h1h3'):
    if depth == 0 or position.is_game_over():
        if (computer == "BLACK"):
            return positionEvaluation(position, piece_values, position_values), bestMove
        else:
            return -1*positionEvaluation(position, piece_values, position_values), bestMove

Where "computer" is some variable you've defined that stores the current player that the computer is evaluating.

We simply invert the evaluation when white is being controlled and the minimax algorithm gives the expected results.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.