Repository: github.com/funkybooboo/rylee

An advanced chess engine designed to mimic human playing behavior using AI techniques and machine learning. Unlike traditional engines optimized for perfect play, Rylee captures the nuances of human decision-making to create a more realistic and educational opponent.

Goals

  • Human-like Play: Simulate natural mistakes and strategic patterns common to human players
  • Adaptive Difficulty: Adjust gameplay style based on opponent’s skill level
  • Educational Value: Provide learning players with realistic, relatable opponents

Implementation Details

Rylee utilizes advanced AI techniques to evaluate board states. Unlike traditional engines that rely solely on Minimax and Alpha-Beta pruning for optimal play, Rylee incorporates:

  • Probabilistic Models: Choose moves that a human of a certain rating might select
  • Pattern Recognition: Identify common human strategic patterns and mistakes
  • Machine Learning: Train on human game databases to capture realistic play styles
  • Rating-Based Behavior: Adjust decision-making based on target player strength

Technology Stack

  • Language: Python
  • AI/ML: PyTorch, reinforcement learning
  • Chess Logic: Custom board representation and move generation
  • Training Data: Human game databases for realistic behavior modeling