JavaScriptMinimaxAlpha-Beta PruningDocker

Chess Engine

A fully functional chess game with an AI opponent — built in vanilla JavaScript with zero external dependencies. Everything from the board rendering to the move generation to the AI search was written from scratch.

Features

  • Full chess rule enforcement: castling, en passant, pawn promotion (with piece selection), check, checkmate, stalemate
  • Draw detection: 50-move rule, threefold repetition
  • Three AI difficulty levels using configurable search depth
  • Algebraic notation move history (e.g. e4, Nf3, O-O)
  • FEN string support for importing/exporting board positions
  • Undo move functionality
  • Docker support for local serving

How the AI Works

The AI uses minimax search with alpha-beta pruning, which prunes branches that can't affect the final decision — dramatically reducing the number of positions evaluated.

On top of that:

  • Quiescence search extends the search at leaf nodes when captures are available, preventing the AI from stopping mid-sequence and misreading a position
  • MVV-LVA move ordering (Most Valuable Victim, Least Valuable Attacker) sorts captures to try the best moves first, maximizing pruning efficiency
  • Piece-square tables bias the evaluation function toward good piece placement — knights in the center score higher than knights on the rim, for example

At difficulty 3 (depth 4), the AI plays solid, aggressive chess and will punish obvious mistakes.

Tech Stack

  • Language: Vanilla JavaScript (ES6+)
  • UI: DOM manipulation, CSS Grid for the board
  • AI: Minimax + alpha-beta pruning, quiescence search
  • Serving: Docker + static HTTP server
  • Dependencies: None