ChessBot

ChessBot is a Python-based chess application offering both Command-Line (CLI) and Graphical (GUI) interfaces. It allows you to play or observe matches in various modes, generate training datasets, train models, and configure AI algorithms.

Features

• Python 3.9 Compatible: Developed and tested on Python 3.9.
• CLI and GUI Modes:
  • CLI: Simple, text-based interaction for quick play or AI vs AI observation.
  • GUI: A visual interface powered by PyQt5, with draggable pieces, dialogs for AI configuration, and Tab-key navigation.
• Game Modes:
  • Human vs AI: Play as White against a configurable AI (drag and drop moves in GUI, or UCI commands in CLI).
  • AI vs AI: Set two AIs against each other and watch them battle.
• AI Configuration: Choose algorithms like Heuristic, NegamaxAlphaBeta, IDPVS, or MCTSPolicyGuided. Adjust parameters (depth, parallelization, simulations) easily.
• Data Generation & Model Training: Generate datasets (policy/value or simple value), then train neural networks to improve AI decisions.

Requirements

• Python Version: Python 3.9 recommended.
• Dependencies: Listed in requirements.txt.

Install dependencies:

pip install -r requirements.txt

Generating Datasets

ChessBot can create datasets for training neural networks:

1. Policy/Value Dataset:

   • Script: generate_policy_value_dataset.py
   • Requires a PGN file of games and a Stockfish engine path.
   • Adjust parameters in the script (like NUM_POSITIONS, DEPTH).
   • Run:

     python generate_policy_value_dataset.py

   • Produces policy_value_dataset.npz with:
     • inputs: Board features
     • policy: Move probability targets
     • value: Position values in [-1,1]

2. Simple Dataset (Value-only):

   • Script: generate_simple_dataset.py
   • Produces a simpler dataset with only inputs and scalar evaluations.
   • Run:

     python generate_simple_dataset.py

   • Results in simple_dataset.npz.

Ensure you have set STOCKFISH_PATH and PGN_FILE inside the scripts before running.

Training Models

After dataset generation:

• Train Policy-Value Net:

  • Script: train_policy_value_net.py
  • Needs policy_value_dataset.npz.
  • Run:

    python train_policy_value_net.py

  • Saves policy_value_model.pt to model/.

• Train Simple Value Net:

  • Script: train_simple_chess_net.py
  • Needs simple_dataset.npz.
  • Run:

    python train_simple_chess_net.py

  • Saves simple_model.pt to model/.

Tune epochs, batch size, learning rate in the scripts as desired.

Running ChessBot

Once ready, or even without training, you can run:

python main.py

You’ll see a menu:

=== ChessBot Interface Selection ===
1. Command-Line Interface (CLI)
2. Graphical User Interface (GUI)
3. Exit
====================================
Enter your choice:

• "1" for CLI:

  • Pick Human vs AI or AI vs AI.
  • Human vs AI: Configure AI, then enter moves in UCI format (e2e4).
  • AI vs AI: Configure two AIs and watch the moves in text form.

• "2" for GUI:

  • A dialog asks for Human vs AI or AI vs AI.
  • Human vs AI:
    • Configure the AI in a dialog.
    • A chessboard window appears.
    • Drag pieces to move, AI responds automatically.
  • AI vs AI:
    • Configure both AIs.
    • Watch them play automatically.

• "3" to Exit.

Keyboard Navigation in GUI

Dialogs support Tab navigation:

• Press Tab to move between buttons and input fields.
• Press Enter to select the focused element. This ensures the GUI is accessible even without a mouse.

Troubleshooting

• GUI not starting? Ensure PyQt5 is installed and display is available.
• Missing dependencies? Reinstall:

  pip install -r requirements.txt

• Performance issues? Lower AI depth or disable parallelization.
• No Python 3.9? It may work on other 3.x versions, but 3.9 is recommended.