tictactoe-ai

Tic Tac Toe game that can be played between bot vs bot, human vs bot or human vs human.

There are two implementations for bot players: RandomPlayer and QPlayer, the last plays using a Q-Learning algorithm.

Results

All the results are from 1000 matches of 100 games each, the start player was selected randomly. Where is involved a QPlayer it started from scratch to learn.

RandomPlayer vs RandomPlayer.
RandomPlayer vs QPlayer.
QPlayer vs QPlayer.
As an exception in this case QPlayer instances used pre-trained sets, so they had previous knowledge when started to play.

Playing vs AI

If you want to play vs a bot, it will be a QPlayer and will load a pre-trained set of states conveniently.

Requirements

The project was developed using python 3.8.5 and the requirements are listed in requirements.txt.

Running the code

usage: main.py [-h] [-x X] [-o O]

optional arguments:
  -h, --help  show this help message and exit
  -x X        Player X type, human('h'), random('r') or q-learning('q')
  -o O        Player O type, human('h'), random('r') or q-learning('q')

Example

Next instruction will start a simulation between two instances of QPlayers and then will plot the results.