Ban Chess Engine

An AI engine for the Ban Chess variant, built on top of the ban-chess.ts library.

Overview

This engine uses traditional chess AI techniques (minimax with alpha-beta pruning) adapted for Ban Chess. The key insight is that the ban-chess.ts library's ply-based system allows us to treat bans and moves uniformly - each ply is simply an action in the game tree.

Features

• Standard minimax algorithm with alpha-beta pruning
• Iterative deepening for better time management
• Position evaluation tailored for Ban Chess:
  • Material balance
  • Mobility (number of legal actions)
  • King safety with ban vulnerability assessment
  • Center control
• Simple CLI interface to play against the engine

Installation

cd ban-chess-engine
npm install
npm run build

Usage

Playing against the engine

npm run play

Using the engine programmatically

import { BanChess } from 'ban-chess.ts';
import { BanChessEngine } from 'ban-chess-engine';

const game = new BanChess();
const engine = new BanChessEngine({
  maxDepth: 6,        // Search depth in plies
  timeLimit: 5000     // Time limit in milliseconds
});

// Get best action for current position
const result = engine.findBestAction(game);
if (result.action) {
  game.play(result.action);
  console.log(`Engine played with score: ${result.score}`);
}

Architecture

The engine leverages the ply-based system from ban-chess.ts:

• Ply 1: Black bans
• Ply 2: White moves
• Ply 3: White bans
• Ply 4: Black moves
• etc.

This means we can use standard minimax without special handling for the ban/move alternation. Each ply is treated as a single node in the search tree.

Evaluation Function

The position evaluation considers:

1. Material: Standard piece values
2. Mobility: Number of available actions (bans or moves)
3. King Ban Safety: Kings near edges are more vulnerable to ban-induced checkmates
4. Center Control: Controlling central squares

Performance

The engine searches approximately 6-8 plies deep in typical middlegame positions within a 3-second time limit. The actual depth depends on:

• Position complexity
• Number of legal actions
• Alpha-beta pruning effectiveness

Future Improvements

• Opening book specifically for Ban Chess patterns
• Endgame tablebases for ban positions
• Transposition tables to avoid re-evaluating positions
• Improved move ordering for better pruning
• Machine learning for evaluation tuning