A fully playable terminal-based chess game built from scratch in C.
Features human vs AI gameplay, full move validation, a tactical evaluation system, and a Minimax AI with alpha-beta pruning. Designed to run in the terminal with fast, optimized performance and support for all major chess rules.
- Human vs AI chess game in the terminal
- Minimax algorithm with alpha-beta pruning
- Evaluation engine with weighted heuristics:
- Material balance
- Piece-square tables
- King safety, castling bonus, fork bonuses
- Full move legality validation
- ✅ Castling (both sides)
- ✅ En passant
- ✅ Pawn promotion
- ✅ Check, checkmate, stalemate detection
- Opening book for early AI moves
- Dynamic attack maps and pinned piece logic
- Highly optimized for fast search at depth 4
#define VALUE_KING 9999 // arbitrarily large to prioritize king safety
#define VALUE_QUEEN 9
#define VALUE_ROOK 5
#define VALUE_BISHOP 3
#define VALUE_KNIGHT 3
#define VALUE_PAWN 1
#define VALUE_PASSED_PAWN 2
#define BONUS_FORK 3
#define BONUS_KING_PRESSURE 2
#define VALUE_CASTLED_KING 50
#define PENALTY_DOUBLED_PAWN 50
#define PENALTY_ISOLATED_PAWN 20
#define BONUS_BISHOP_PAIR 50
#define BONUS_CONNECTED_ROOKS 25
#define MAX_DEPTH 4
#define INFINITY 10000The engine uses a mix of:
- Tactical bonuses (e.g., forks, pressure)
- Positional heuristics (e.g., doubled pawns, castling)
- Piece-square tables for better piece placement
clang -o chess chess.c
./chessRuns entirely in your terminal. No dependencies.
Move: 2e4e
AI plays: f7f6
A B C D E F G H
8|r|n|b|q|k|b|n|r|
7|p|p|p|p|p| |p|p|
6| | | | | |p| | |
5| | | | | | | | |
4| | | | |P| | | |
3| | | | | | | | |
2|P|P|P|P| |P|P|P|
1|R|N|B|Q|K|B|N|R|
0A= starting square1B= destination square
Example:
- Input your moves in the format
2e4e(start square + end square)- e.g.
2e4emoves the piece at e2 to e4
- e.g.
- Promotion will prompt a piece selection (Q/R/B/N)
- The game alternates between player (White) and AI (Black)
Move: 2e3e
Recent improvements to:
- Alpha-beta pruning logic
- Move legality checks (faster pinned piece detection)
- Memory footprint (no dynamic memory allocation needed)
Search depth of 4 runs smoothly with sub-second response time on modern CPUs.
All code is written in a single C file with:
- Structs for board state and moves
- Modular functions for evaluation, move generation, legality, AI logic, and rendering
📌 Ideal for understanding how a chess engine works at a low level.
This project was designed and implemented solely by me.
All code was written manually, without AI code generation.
The full commit history documents the step-by-step development process.
Most planned features are finished, but here are some possible upgrades:
Work in progress — major rules and features still under development.
- Add a GUI (e.g., SDL, ncurses)
- Implement PGN/FEN loading and saving
- Add support for threefold repetition and 50-move rule
- Integrate a more advanced evaluation model (e.g., neural network or centipawn analysis)
- Support UCI protocol for third-party GUI engines
- Improvements to the AI:
- Right now the AI still struggles with threat detection especially with passed pawns.
- On hold right now.
I’ve always been drawn to low-level programming, especially the clarity and control you get when working in C. It’s often called a “simple” language, but that simplicity is what makes it so challenging — you’re exposed to every detail. I wanted to explore game development and AI from the ground up, so I set out to build a full chess engine from scratch, including the AI. I wrote the minimax algorithm myself and implemented all the move validation, board evaluation, and edge cases manually. Throughout the project, I found myself thinking more critically about structure, readability, and long-term maintainability — and that mindset really helped sharpen my fundamentals. It also gave me deeper insight into search trees, evaluation heuristics, and optimization, while reinforcing what I enjoy most: building something complex from the ground up.
Developed by Xu Chen (xuchen11091). Built using C and lots of debug patience. AI inspired by traditional chess engine design principles and heuristics used in classic engines.
MIT License (or specify if otherwise).
Feel free to reach out or fork this repo if you'd like to contribute or suggest improvements.