♟️ Chess Agent Bot — Deep Reinforcement Learning

An intelligent chess-playing AI built using Deep Q-Learning (DQL) that learns to play chess through self-play and reinforcement learning.
The agent improves over time by evaluating board states, generating legal moves, and optimizing long-term rewards.

---

🚀 Project Overview

This project focuses on building a reinforcement learning–based chess engine rather than relying on classical minimax or hard-coded heuristics.

The agent:

• Learns from self-play
• Understands complete chess rules
• Improves decision-making through reward optimization
• Interacts with a Pygame-based chess environment

---

🧠 Key Features

• ♞ Deep Q-Learning (DQL) for move selection
• ♜ Complete chess rule encoding (legal moves, captures, checks, etc.)
• 🔁 Self-play training environment
• 🎯 Custom reward function for strategic learning
• 📈 Performance improvement after training
• 🎮 Pygame-based GUI for visualization

---

🧩 Tech Stack

Category	Tools
Programming Language	Python
Machine Learning	Deep Q-Learning (DQL)
Libraries	NumPy, TensorFlow
Game Engine	Pygame
Environment	Custom Chess Environment
Version Control	Git & GitHub

---

🎯 How the Agent Learns

1. Observes the current board state
2. Generates all legal moves
3. Selects an action using ε-greedy policy
4. Receives a reward
5. Updates Q-values using the Bellman Equation
6. Improves strategy over thousands of games

---

🏆 Reward Strategy (Simplified)

Action	Reward
Capture opponent piece	Positive
Checkmate	High Positive
Illegal move	Negative
Losing piece	Negative
Winning game	High Positive

---

👤 Author

• pardhu01010