Chess Move Generator

JavaScript chess move generator with strictly legal move generation

English | Español

🎯 Key Features

• ✅ Strictly legal move generation - No pseudo-moves requiring post-validation
• ✅ Integrated check, checkmate, and stalemate detection - During generation, not as a post-processing step
• ✅ Two implementations: x88 and Bitboards
• ✅ Automatic tactical analysis - Each move includes information about winning captures, hanging pieces, safe squares
• ✅ Complete UCI engine - Compatible with standard chess interfaces
• ✅ Interactive web interface - Visual demo with drag & drop board
• ✅ Web Workers - Calculations without blocking the UI

🚀 Quick Demo

Try it live!

Or open engine.html in your browser locally.

📦 Installation

NPM

npm install chess-movegen-js

Usage

const { Board } = require('chess-movegen-js');

const board = new Board();
board.loadFEN('rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1');
board.generateMoves();

console.log(`Legal moves: ${board.moves.length}`); // 20

📖 Usage Examples

Basic Initialization

// Create a board
const board = new Board();

// Load a FEN position
board.loadFEN('rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1');

// Generate all legal moves
board.generateMoves();

// View the moves
console.log(board.moves);

Move Analysis

Moves include tactical information:

board.generateMoves();

board.moves.forEach(move => {
    const moveStr = board.getMoveStr(move);
    console.log(moveStr);
    
    // Tactical information in move.mask:
    // - mask_check: Gives check
    // - mask_safe: Safe square
    // - mask_hanging: Piece would be hanging
    // - mask_freecapture: Undefended capture
    // - mask_winningcapture: Winning capture
});

Make and Unmake Moves

// Make a move
const move = board.moves[0];
board.makemove(move);

// Unmake
board.undomove();

Perft (Move Generation Testing)

// Count nodes at depth 5
const nodes = board.perft(5);
console.log(`Nodes: ${nodes}`); // 4,865,609 from initial position

// Divide (show nodes per move)
board.divide(4);

🏗️ Project Structure

movegen/
├── js/
│   ├── x88.js           # x88 representation generator (1842 lines)
│   ├── bitboard.js      # Bitboard generator
│   ├── magic-tables.js  # Magic tables for bitboard
│   └── engine.js        # UCI engine with Web Worker
├── assets/              # css and js assets for the demo
├── img/                 # Graphic resources
├── engine.html          # Main interactive demo
├── ANALISIS.md          # Detailed technical analysis
└── README.md            # This file

🎮 UCI Engine

The project includes a complete UCI engine running in a Web Worker:

// Create engine
const w = new Worker("js/engine.js");

// UCI communication
w.postMessage('uci');
w.postMessage('position fen rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1');
w.postMessage('perft 6');

// Listen for responses
w.onmessage = function(event) {
    console.log(event.data);
};

Supported UCI Commands

• uci - Initialize engine
• isready - Check availability
• ucinewgame - New game
• position [fen|startpos] [moves ...] - Set position
• move <move> - Make move (e.g., e2e4)
• undo - Unmake move
• perft <depth> - Move generation test

⚡ Performance

Perft from initial position (Node.js v20+, no debug):

Depth	Nodes	Time	NPS
1	20	<1ms	~25k
2	400	~1ms	~268k
3	8,902	~10ms	~864k
4	197,281	~83ms	2.4M
5	4,865,609	~871ms	5.6M
6	119,060,324	~17s	7.0M

In browser (may vary by browser and hardware):

• Chrome/Edge: ~3-5M NPS
• Firefox: ~2-4M NPS

Note: These results are with optimized code (no debug() calls). Production performance is excellent for pure JavaScript.

🎨 Technical Highlights

1. Pin Detection

Pinned pieces are detected during generation. Illegal moves are never generated:

// pinDirection[side][square] indicates if a piece is pinned
// and in which direction

2. Tactical Enrichment

Each move contains flags indicating:

• If it gives check or checkmate
• If the piece would be hanging
• If it's a winning capture
• If it's a safe square

3. Special Cases

• ✅ En passant capture with horizontal pins
• ✅ Discovered checks (including in castling)
• ✅ Mate-in-one detection
• ✅ Multiple promotions

🔬 Implementations

x88 (Recommended for learning)

• 128-position array (16×8)
• Ultra-fast validation: if (sq & 0x88) continue
• More readable and easier to understand code
• File: js/x88.js

Bitboards (Experimental)

• 64-bit bitboard representation
• Faster in theory, more complex
• File: js/bitboard.js

📚 Documentation

For a complete technical analysis of the code, see ANALISIS.md.

🧪 Testing

The project uses Perft to validate move generation:

// From browser console in engine.html
w.postMessage('perft 5');

// Or in code
const board = new Board();
board.loadFEN('rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1');
console.log(board.perft(5)); // Should be 4,865,609

🧪 Running Tests

The project includes a comprehensive Perft test suite to validate move generation correctness and measure performance.

Quick Start

# Run quick test suite (depths 1-4, ~1 minute)
node tests/perft-test.js --quick

# Test specific position
node tests/perft-test.js --position 0 --depth 5

# Test only x88 generator up to depth 6
node tests/perft-test.js --generator x88 --depth 6

Available Options

node tests/perft-test.js [options]

Options:
  --generator <x88|bb|both>   Select generator to test (default: both)
  --position <n>              Test only position n (default: all)
  --depth <n>                 Test up to depth n (default: 6)
  --quick                     Quick test mode (depths 1-4)
  --help                      Show help message

Test Positions

The test suite includes 7 standard positions from Chess Programming Wiki:

Position	Description	Max Depth Tested
0	Initial position	10
1	Kiwipete (complex middle game)	6
2	En passant edge cases	8
3	Promotions	6
4	Promotions (mirrored)	6
5	Complex tactical position	5
6	Symmetrical position	6

Expected Output

Chess Move Generator - Perft Test Suite

Configuration:
  Generator: x88
  Positions: 7
  Max depth: 4

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Testing: x88 Generator
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Initial Position
FEN: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
  ✓ Depth 1: 20 nodes [1ms, 17,891 NPS]
  ✓ Depth 2: 400 nodes [2ms, 167,560 NPS]
  ✓ Depth 3: 8,902 nodes [9ms, 973,343 NPS]
  ✓ Depth 4: 197,281 nodes [80ms, 2,480,626 NPS]

Kiwipete
FEN: r3k2r/p1ppqpb1/bn2pnp1/3PN3/1p2P3/2N2Q1p/PPPBBPPP/R3K2R w KQkq - 0 1
  ✓ Depth 1: 48 nodes [0ms, 302,457 NPS]
  ✓ Depth 2: 2,039 nodes [2ms, 1,220,008 NPS]
  ✓ Depth 3: 97,862 nodes [32ms, 3,068,986 NPS]
  ✓ Depth 4: 4,085,603 nodes [548ms, 7,461,185 NPS]

...

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Summary
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Total tests: 28
Passed: 28
Failed: 0
Pass rate: 100.0%

Performance Summary:
  Depth 1: 883,697 NPS avg (368 nodes in 2ms)
  Depth 2: 2,126,798 NPS avg (13,446 nodes in 10ms)
  Depth 3: 3,360,700 NPS avg (561,558 nodes in 170ms)
  Depth 4: 4,722,406 NPS avg (22,337,738 nodes in 4.51s)

Performance Benchmarks

Measured on Node.js v20+ with x88 generator (optimized, no debug):

Depth	Nodes	Time	NPS
1	20	<1ms	~25k
2	400	~1ms	~268k
3	8,902	~10ms	~864k
4	197,281	~83ms	2.4M
5	4,865,609	~871ms	5.6M
6	119,060,324	~17s	7.0M

Quick test suite (all 7 positions, depths 1-4): ~1.4 seconds

💡 Tip: Performance is significantly faster with debug logging disabled. Make sure to comment out this.debug() calls in production.

Troubleshooting

If tests fail or show errors:

1. Ensure Node.js is installed: The tests require Node.js v14 or higher
2. Check all files are present: Make sure tests/ directory exists with all test files
3. Verify x88.js modifications: The file should have Node.js compatibility added

Known Test Positions

// Kiwipete position
board.loadFEN('r3k2r/p1ppqpb1/bn2pnp1/3PN3/1p2P3/2N2Q1p/PPPBBPPP/R3K2R w KQkq - 0 1');
console.log(board.perft(5)); // 193,690,690

// Complex en passant capture
board.loadFEN('8/2p5/3p4/KP5r/1R3p1k/8/4P1P1/8 w - - 0 1');

🎯 Next Steps

• Automated tests with Perft suite ✅
• Publish as NPM package ✅
• Repetition check with Zobrist hashing
• WebAssembly optimization
• Add position evaluation
• Implement full alpha-beta search

⚙️ Continuous Integration (GitHub Actions)

• Badge: agregado arriba en este README.
• Qué ejecuta: el workflow CI corre los tests rápidos (npm test) automáticamente en push y pull_request. El job de tests completos de bitboard (npm run test:bb) está configurado como ejecución manual para evitar ejecuciones largas en cada push.

Cómo lanzar el test completo desde GitHub UI:

1. Ve a la pestaña Actions del repositorio.
2. Selecciona el workflow CI.
3. Pulsa Run workflow, elige la rama (main o master) y ejecuta.

Usando la CLI gh (GitHub CLI) puedes lanzar el workflow así:

gh workflow run ci.yml --ref main

Nota: el workflow CI incluye matrix de versiones Node para los tests rápidos y reserva un job manual (test-bitboard-full) con mayor timeout para las pruebas intensivas.

🤝 Contributing

Contributions are welcome. Please:

1. Fork the project
2. Create a branch for your feature (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📖 Resources and References

• Chess Programming Wiki
• 0x88 Board Representation
• Perft Results
• UCI Protocol

📝 License

This project is under the MIT License - see the LICENSE file for details.

👤 Author

Mario Raúl Carbonell Martínez

---

⭐ If you find this project useful, consider giving it a star on GitHub!