An automated framework for detecting potential cheating in chess games through statistical analysis of Portable Game Notation (PGN) files, developed as part of an MTech dissertation in AI & ML at BITS Pilani.
This project addresses the growing concern of chess cheating in online platforms by analyzing gameplay patterns using centipawn loss (CPL) metrics and machine learning techniques. The system identifies statistical anomalies that may indicate computer assistance during gameplay.
- PGN File Analysis: Parse and analyze chess games in standard PGN format
- Centipawn Loss Calculation: Evaluate move quality against optimal engine recommendations
- Statistical Pattern Detection: Identify suspicious gameplay patterns and anomalies
- Data Visualization: Generate comprehensive graphs and charts for pattern analysis
- Machine Learning Classification: 95% accuracy achieved using Neural Network models
- Multi-factor Analysis: Consider game length, move timing, and contextual factors
- Python 3.x
- Chess Libraries:
python-chess,chess.pgn - Engine Analysis: Stockfish 16 (depth 20)
- Data Processing:
pandas,numpy - Visualization:
matplotlib,seaborn - Machine Learning:
scikit-learn,tensorflow
- Consistently low CPL across complex positions
- Unnatural CPL stability throughout games
- High alignment with top engine choices in critical positions
- Performance incongruent with player rating
- Unusual game length patterns
- Inconsistent time usage patterns
- Performance spikes in specific periods
- Drastic improvement without corresponding rating increase
- Low Confidence: Only secondary indicators present
- Medium Confidence: One primary + multiple secondary indicators
- High Confidence: Multiple primary indicators present
- Clone the repository
git clone https://github.com/0xafraidoftime/Chess-Cheat-Detection.git
cd Chess-Cheat-Detection- Install required dependencies
pip install python-chess pandas matplotlib seaborn scikit-learn tensorflow stockfish- Download Stockfish Engine
- Download Stockfish 16 from official website
- Ensure it's accessible in your system PATH
chess-cheat-detection/
├── chess_engine.py # Basic PGN parsing functionality
├── chess_engine2.py # Enhanced analysis with CPL extraction
├── CPL plot.py # Visualization and plotting utilities
├── engine3.py # Additional analysis features
├── game length distribution.py # Game length analysis
├── data/
│ ├── a.pgn # Sample PGN files
│ ├── game.pgn
│ ├── new_game.pgn
│ └── annotated_game2.pgn
├── analysis/
│ ├── *.csv # Generated analysis results
│ └── pgn_analysis.csv # Summary statistics
└── README.md
from chess_engine import parse_pgn
# Parse a PGN file
moves = parse_pgn("your_game.pgn")
print("Extracted moves:", moves)# Run CPL analysis and generate visualizations
python "CPL plot.py"# Analyze game length patterns
python "game length distribution.py"from chess_engine2 import analyze_game
# Analyze a game for potential cheating
analyze_game("suspicious_game.pgn")| Algorithm | Accuracy | Precision | Recall | F1-Score |
|---|---|---|---|---|
| Decision Tree | 85% | 88% | 83% | 85% |
| Random Forest | 92% | 90% | 93% | 91% |
| Neural Network | 95% | 94% | 96% | 95% |
- CPL Analysis: Effectively differentiates fair from engine-assisted play
- Game Context: Complexity, phase, and time control are crucial for accurate detection
- Multi-factorial Approach: Reduces false positives significantly
- Pattern Recognition: Identifies bimodal distribution in game lengths for suspicious accounts
- Requires annotated PGN files or Stockfish integration for evaluation
- May not detect sophisticated cheating (intentional suboptimal moves)
- Dataset size impacts ML model accuracy
- Computational intensity for real-time analysis
- Potential false positives with very strong human players
- Real-time analysis capability for tournament monitoring
- Cloud-based implementation for scalability
- Mobile application for arbiters and tournament directors
- Integration with chess platform APIs
- Advanced neural network architectures (LSTM, Transformer)
- Incorporation of psychological factors and playing style analysis
- Application to other strategic games (Go, Shogi)
- Self-improving detection through continuous learning
This project was developed as part of an MTech dissertation in AI & ML at BITS Pilani under the supervision of Milin Shah (VP & Technology Manager, Bank of America). The research addresses the increasing prevalence of chess cheating in online environments, particularly since 2020.
Contributions are welcome! Please feel free to submit issues, feature requests, or pull requests to improve the detection algorithms and expand the project's capabilities.
If you use this work in your research, please cite:
Pal, A. (2025). Cheat Detection in Chess: Analyzing PGN Files.
MTech Dissertation, BITS Pilani. Supervisor: Milin Shah.
This project is open-source and available under the MIT License. See LICENSE file for details.
- Supervisor: Milin Shah, Bank of America
- Institution: BITS Pilani Department of AI & ML
- Chess Community Contributors: Manan Pahwa (Purdue University), Moin Memon (Bank of America)
- Technical Support: Stockfish Development Team, Python-chess Library Contributors
- Data Sources: Chess.com Research Dataset, Lichess Open Database
Ankita Pal
BITS ID: 2022AC05327
Email: [email protected]
This project aims to maintain competitive integrity in chess while providing transparent, objective methods for cheat detection that can assist tournament arbiters and online platforms.