Gomoku AI Game

An intelligent game-playing agent for Five-in-a-Row (Gomoku) featuring advanced AI algorithms and heuristic evaluation functions.

Overview

This project implements a Gomoku game where players can compete against an AI opponent that uses sophisticated search algorithms including Alpha-Beta pruning with custom evaluation functions. The game follows traditional Gomoku rules on a 10x10 board.

Game Rules

• Players alternate placing stones (Black and White) on empty intersections
• Black plays first and must place their first stone in the center
• White's first stone can be placed anywhere except the center
• Black's second stone must be at least 3 intersections away from the first
• First player to form 5 stones in a row (horizontally, vertically, or diagonally) wins
• Game ends in draw if board is full with no winner

Features

• Advanced AI Algorithms: Alpha-Beta search with pruning for optimal performance
• Dual Evaluation Functions:
  • Standard evaluation with basic position scoring
  • Advanced evaluation with threat detection and pattern recognition
• Configurable Depth: Adjustable search depth for different difficulty levels
• Interactive Gameplay: Human vs AI with intuitive coordinate input system
• Performance Optimized: Uses caching and efficient board representation

Tech Stack

• Python 3.x
• NumPy: For mathematical operations and board evaluation
• AIMA Framework: Based on algorithms from "Artificial Intelligence: A Modern Approach"

Getting Started

Prerequisites

pip install numpy

Running the Game

python gomoku.py

Usage

Making Moves

When prompted for your move:

• Enter coordinates as x y (e.g., 5 3)
• Use q to quit the game
• Coordinates range from 0 to 9 for both x and y axes

AI Configuration

The game offers two AI opponents with different evaluation strategies:

Standard AI (alpha_beta_cutoff_search)

• Uses basic position evaluation
• Default depth: 4
• Faster execution, good for casual play

Advanced AI (h_alphabeta_search)

• Uses sophisticated threat detection
• Pattern recognition and sequence evaluation
• Default depth: 3
• More challenging opponent

Customizing AI Behavior

To modify the AI settings in main():

# Use standard AI
play_game(game, {"W": player(alpha_beta_cutoff_search)}, verbose=True)

# Use advanced AI  
play_game(game, {"W": player(h_alphabeta_search)}, verbose=True)

To change search depth, modify the depth parameters:

• alpha_beta_cutoff_search: Change d=4 parameter
• h_alphabeta_search: Change cutoff_depth(3) parameter

AI Features

Evaluation Functions

Standard Evaluation:

• Counts consecutive stones in all directions
• Applies positional weights
• Basic threat assessment

Advanced Evaluation:

• Multi-directional pattern analysis
• Winning sequence detection (5+ stones = 100,000 points)
• Threat prioritization (4 stones = 10,000 points)
• Strategic positioning (2-3 stones = 100-1,000 points)

Search Optimization

• Alpha-Beta Pruning: Eliminates unnecessary search branches
• Depth-Limited Search: Configurable search depth for performance tuning
• Caching: LRU cache for previously evaluated positions
• Move Ordering: Prioritizes promising moves first

Project Structure

Gomoku/
├── gomoku.py          # Main game implementation
├── README.md          # Project documentation
└── LICENSE           # MIT License

Example Gameplay

Your turn. Current board:
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . B . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .

Available moves: Enter the coordinates as 'x y'
Or enter 'q' to quit
Your move? 4 4

Contributing

This project was developed with algorithms and concepts from the AIMA (Artificial Intelligence: A Modern Approach) textbook. Contributions are welcome for:

• Additional evaluation heuristics
• Performance optimizations
• UI improvements
• Extended game variants

License

MIT License - see LICENSE file for details.

Author

Autrin Hakimi