Othello

About the Project

This project is a web-based implementation of an AlphaZero-inspired AI designed specifically for the game of Othello (Reversi). Unlike traditional game AI that relies on hard-coded heuristics or massive databases of human moves, this AI learns the game through self-play reinforcement learning.

Technical Highlights

• Deep Learning Framework: Built with TensorFlow.js, allowing the neural network to run directly in your browser with hardware acceleration.
• Neural Network Architecture: Utilizes a dual-head "Policy" and "Value" network to evaluate board positions and predict the most promising moves.
• Search Algorithm: Employs Monte Carlo Tree Search (MCTS) to look ahead, balancing exploration of new strategies with exploitation of known winning patterns.
• Performance: Fully optimized with WebGL, ensuring that even complex AI calculations remain smooth and responsive on your machine.

How to Use

1. Run the Game: Click the "Run Game" button to initialize the engine and load the neural network weights.
2. Observe or Play: You can watch the AI play against itself to see its learned strategies, or challenge the AI to test your own skills.
3. Monitor Progress: Keep an eye on the real-time training metrics (Loss and Policy Accuracy) to see how the "brain" of the AI is evolving.

Future Roadmap

This is a passion project focused on the intersection of deep learning and classic board games. If there is enough interest from the community, I plan to expand this project with the following features:

• High-Quality Weights: Releasing significantly larger and more robust weight files for a much "stronger" AI opponent.
• Analysis Mode (Game Review): A professional toolset allowing users to set up custom board positions, analyze AI evaluations, and branch off into different "what-if" scenarios.
• Pattern Study: Visualization tools to show which areas of the board the AI prioritizes during different stages of the game.

If you enjoy this implementation or find the technical approach interesting, please leave a comment or follow the project for future updates!