🧠 Tic-Tac-Toe: Minimax vs Reinforcement Learning

This project implements a Tic-Tac-Toe game in Python where a Reinforcement Learning (Q-Learning) agent competes against a Minimax algorithm. It’s based on Sentdex’s tutorial and expanded to allow:

• Head-to-head matches between agents
• Training and evaluation of the Q-Learning model
• Optional human vs AI gameplay

🚀 Features

• Q-Learning agent trained from scratch
• Minimax opponent with perfect play
• Configurable training episodes
• Win/draw/loss performance tracking
• Play-by-play printouts for auto-play mode

📁 File

• tic-tac-toe-minimax-vs-RL.py – main script with everything in one place: training, gameplay, and visualization

🎯 Project Goal

This project is intended as a simple, educational case to explore how reinforcement learning (Q-learning) can be implemented. It is not optimized to create a perfect AI player — the goal is to demonstrate how to build and train a basic RL agent using a simple game like Tic-Tac-Toe.

🧠 How It Works

• The Q-Learning agent learns from trial and error, improving its strategy over thousands of games.
• The Minimax player evaluates all possible future states to make the optimal move.
• You can pit them against each other, play yourself, or watch training performance evolve.

▶️ Usage

Run the script using:

python tic-tac-toe-minimax-vs-RL.py

You'll be prompted to choose:

• Which agent to train or test
• Number of training games
• Whether to play yourself or watch autoplay

📊 Example Output

Training in progress...
Episode 1000 | Win rate: 85% | Draw rate: 10% | Loss rate: 5%

📚 Inspired By

Tutorial series by Sentdex, which provides excellent resources on Python and AI.

📄 License

MIT License — feel free to use, modify, and share!