This project implements a game-playing AI agent for the board game Havannah.
Havannah is a two-player abstract strategy game invented by Christian Freeling. Players take turns placing their stones on hexagonal cells of a board, aiming to achieve one of the following winning conditions:
- Ring: Form a loop of stones enclosing one or more cells.
- Bridge: Connect any two corners of the board.
- Fork: Connect any three edges of the board.
- Smart Decision Making: Implements search algorithms MCTS with RAVE and UCB to optimize move selection.
- Heuristic Evaluation: A custom evaluation function assesses board positions to enhance decision quality.
- Scalability: Supports various board sizes, from beginner-friendly small boards to full-size competitive ones.
- Interactive Play: Play against the AI in a dynamic interface or simulate AI vs. AI matches to observe its strategic prowess.
Have a look at the setup file for detailed instructions on setting up the environment and usage.
- Machine Learning Integration: Train the AI using reinforcement learning to improve over time.
- Dynamic Heuristics: Adapt heuristics based on player behavior for more challenging games.
This project is licensed under MIT; look at the LICENSE file for details.
This work was completed as part of COL333 Assignment 2 at IIT Delhi, exploring search algorithms for game-playing agents under academic supervision.