So clearly there's something wrong with your minimax algorithm. That means there's no reason for your PerfectPlayer to not actually be perfect: It should never loose (and against a moves-randomly opponent, it should win quite often). Although ~70% of the games are won by it, it does lose and tie a small fraction of games to the randomized choice opponent. # If the simulated score from a specific move is the best option, do itĪlso, please note that the minimax user does indeed lose at time. # Undo the move (remember, it's only the simulation)

Sim_score = self.minimax(state, opponent) 'score': 1 * (state.empty_square_vals() + 1) if opponent = max_player else -1 * (state.empty_square_vals() + 1) # If there are tokens on the board, then find the optimal move based on that token's positionĮlse: choice = self.minimax(game, self.letter) If len(game.open_moves()) = 9: choice = random.choice(game.open_moves()) # If all moves are open, then do any random move # Get the pefect move for the game - where the MAGIC happens! # Perfect Tic Tac Toe AI using the Minimax Algorithm # Set the player's letter (noughts or crosses)ĭef _init_(self, letter): self.letter = letter # Recognise the player as an object, with a specific letter (X or O) The code I have for both the algorithm player ( PerfectPlayer), as well as the normal computer player ( ComputerPlayer), is below. Third test: Perfect Player X won 68 time(s), Player O won 25 time(s), and there were 7 tie(s). Second test: Perfect Player X won 70 time(s), Player O won 21 time(s), and there were 9 tie(s). Here are a few tests that show results of the Perfect Player, which uses the minimax algorithm, against a normal computer player, which randomly chooses spots to play during the game:įirst test: Perfect Player X won 67 time(s), Player O won 25 time(s), and there were 8 tie(s).

As of now, It has approximately 70% chance of winning each game, but I am trying to increase that number. I am working on a Tic Tac Toe AI in Python that uses the Minimax Algorithm to beat either a computer or a normal player.

Test my tic tac toe algorithm