Chain Duel Simulation

A Python simulation model for analyzing AI strategies in Chain Duel, inspired by Chain Duel Online.

Introduction

This project offers three modes for snake-like duels:

• Interactive mode with grid display:
  • Play against AI (use arrow keys)
  • Watch two AIs compete
• Simulation mode for batch analysis

Setup (Ubuntu)

# Create and activate virtual environment
python3 -m venv env
source env/bin/activate

# Install dependencies
pip install -r requirements.txt

Start the game:

python main.py

##Setup (Windows)

• Download and install Python
• In your Python prompt: cd C:[location of your file] python main.py

Start the game:

• select your game mode
• to modify the number of simulations, modify the source code files

Interactive Modes

• Player vs AI: Control green snake with arrow keys while AI controls orange
• AI vs AI: Watch two AI strategies compete in real-time
• Controls: R to restart, ESC to quit

Simulation Mode

Runs multiple games without visualization for statistical analysis. Results saved in simulations/sim_TIMESTAMP/:

• results.txt: Detailed game data
• stats.txt: Performance metrics

Batch Analysis

1. Run full batch simulation across all strategies and cases:

   python run_batch_simulations.py

   This generates batch_results.csv and STRATEGIES_AND_CASES.md in the project root.
2. Analyze and visualize the results:

   python analyze_results.py

   Outputs are saved under analysis_outputs/:
   • RESULTS_SUMMARY.md: Markdown report with
     • Overall average win‑rate table and bar chart
     • Per‑case win‑rate matrices and heatmaps
     • Strategy win‑rate comparison across the 4 initial cases (table & line chart)
   • PNG charts:
     • overall_avg_win_rate.png
     • heatmap_<case>.png (one per case)
     • comparison_win_rate_cases.png (strategy vs cases)

Project Structure

├── main.py          # Main entry point
├── src/
│   ├── core/        # Game mechanics
│   ├── strategies/  # AI implementations
│   ├── ui/         # Game interface
│   └── simulation/ # Batch simulation

TODO

Scoring System

• Evaluate point deduction on first food collection
• Implement competitive scoring mechanics
• Add multiplication factors based on snake length

AI Strategy Improvements

• Implement direct path to first food (Player 1)
• Add hairpin maneuver for first food (Player 2)
• Develop minimal avoidance with food seeking
• Food-unreachable behaviors:
  • Center positioning
  • Board half switching
  • Quadrant selection (based on food history)

License

MIT License - Feel free to use this code for any purpose while maintaining the license notice.

Reference

Based on the Chain Duel game: https://chainduel.net/ & https://github.com/Asi0Flammeus/chainduel_simulation/