---

Shinigami Chess Engine (V.1.18.2 – Gen 2 Edition)

Shinigami V.1.18.2 – Gen 2 Edition (Latest)
A professional chess engine with full tree parallelization, advanced NNUE, LLM-powered move explanations, self-adapting features, and a brutal/funny trash talk system.

---

🚀 Latest Updates (V.1.18.2)

• Genetic Feature Engineering: DEAP-powered auto-tuning for piece values and PSTs mid-game.
• Dynamic Opening Book Pruning: Opening book now self-prunes weak lines and adapts to new opponents.
• Enhanced Evaluation:
  • King safety, pawn structure, mobility, outpost, bishop pair, rook/queen evaluation refined.
• Full UCI Protocol: Full UCI compliance.
• Logging: Cleaner, more informative logs for games, search, and training.
• Opponent Learning: Engine adapts to your style in real time.
• Debug: Fixed a Lot of Bugs and made the code more cleaner and organized.

---

Features

• Full Tree Parallelization: Multicore search for epic depth and speed.
• Advanced NNUE (HalfKAv2): Modern neural evaluation (user-trainable).
• CNN Policy Network: Move ordering powered by a convolutional neural net.
• Syzygy Tablebase Support: Endgame perfection (≤7 pieces).
• Genetic Feature Engineering: DEAP-powered auto-tuning for piece values and tables.
• Self-Play & Training: Generates and learns from games, retrains NNUE and policy networks.
• Dynamic Opening Book: Adapts with self-play and opponent moves, prunes weak lines.
• Opponent Learning: Adapts to your style, learning openings and move qualities.
• Puzzle Generator: Built-in tactical puzzles and board tasks.
• GUI & Console: Tkinter GUI or classic terminal mode.
• UCI Protocol: Plug into chess GUIs (Arena, CuteChess, etc.).
• Trash Talk Engine: Witty, customizable, optionally brutal commentary.
• LLM Explanations: In-character move explanations via OpenAI API (optional).
• Extreme Difficulty Modes: From beginner to "The Big Bang" (with cosmic warnings...).

---

Installation

Clone the Repository:

git clone https://github.com/Tonmoy-KS/Shinigami-Chess-Engine
cd Shinigami-Chess-Engine

Install Dependencies:

pip install -r requirements.txt

Requires Python 3.8+.

Optional Assets

• NNUE Weights: Download nnue_weights.bin (HalfKAv2 compatible) and place in project root or specify with --nnue-file.
• Syzygy Tablebases: Download and place .rtb files in ./tablebases or specify via --syzygy-path.
• Polyglot Opening Book: Place .bin file in project root or specify via command-line.

---

Usage

Console Mode:

python3 Main_Code_1

GUI Mode:

python3 Main_Code_1 --gui

Self-Play & Training:

python3 Main_Code_1 --self-play 100

Custom Paths & Cores:

python3 Main_Code_1 --cores 4 --nnue-file /path/to/my_nnue.bin --syzygy-path /path/to/my_tablebases

UCI Protocol (for chess GUIs):

python3 Main_Code_1 --uci

---

Logging

Game statistics, self-play, and learning data are logged to shinigami_engine.log.

---

LLM-Powered Move Explanations

• Uses OpenAI API (set OPENAI_API_KEY environment variable).
• Optional, can be enabled/disabled at runtime.
• Provides witty, in-character commentary on engine moves.

---

Extreme Difficulty Modes

Official

• easy: Beginner
• medium: Club-level
• hard: Strong player
• god-of-death: near Grandmaster
• puzzle: Tactical puzzle mode

Experimental/Jokes

• masochist: Insane mode
• dialing-satan-s-number: Meme mode
• the-big-bang: God mode

---

FAQ

1. What makes Shinigami different from other chess engines?
   A: Genetic feature engineering, trash talk, and LLM-powered move explanations made it a great UX software and different from other Great engines that are just Silent Calculators.

2. How do I enable LLM Explanations?
   A: Set your OpenAI API key and OpenAI model and enable at startup,we also have a LLM model but it's currently not available.

3. How do I enable/disable The Trash Talk Dictionary?
   A: Enabled by default. You can customize or disable in the source code.

4. What are “The Big Bang” and “Dialing Satan’s Number” modes?
   A: Extreme/joke modes. “The Big Bang” is blocked for safety; “Dialing Satan’s Number” is a meme. just to test the limits of Computation.

5. Does Shinigami learn/adapt openings from games?
   A: Yes, via dynamic pruning and opponent learning.

6. How do I train NNUE and Policy Network?
   A: Use self-play, or use your own datasets via Training a PyTorch model. so official Model currently.

7. Is Shinigami UCI-compliant for GUIs?
   A: Yes.

8. Can I contribute neural weights, feature sets, or puzzles?
   A: contributions welcome.

9. What about system requirements for extreme modes?
   A: Use first five modes are Official modes that are Stable; the other modes are just experiments.

10. Can I customize the playing style (aggressive, positional)?
    A: Not currently. planned for future updates.

---

License

License: MIT
Credit: Tonmoy-KS
Do not claim as your own; fork, mod, and contribute instead!

---

Major Contributers

1. Inskitz [https://github.com/InSkitz]
2. Dakter-Tech [https://github.com/dakther-tech]
3. Shanjid-KS [https://github.com/shanjid-KS]

---

How to Contribute

Pull requests, feature ideas, and code reviews are welcome! For major contributions, open an issue first.

---

Contact

GitHub: Tonmoy-KS

---

Happy reaping in 64 squares! — bye

---