_____  ____  __  __ _    _ _____     
|  __ \|___ \|  \/  | |  | | ____|    
| |__) | __) | \  / | |  | | |____  __
|  _  / |__ <| |\/| | |  | |___ \ \/ /
| | \ \ ___) | |  | | |__| |___) >  < 
|_|  \_\____/|_|  |_|\____/|____/_/\_\
 

ChaosEngine ♟️

A deliberately chaotic UCI chess engine in a single C++ file.

It plays legal chess — castling, en passant, promotion, check detection, the whole thing — but with a wild streak. A Chaos slider mixes pure-random moves with a tiny material search, so you can dial it anywhere from "vaguely sensible" to "throws pieces into the void for fun".

It will not beat Stockfish. That is not the point. The point is that it loses in interesting ways.

Build

Requires a C++17 compiler (g++ or clang++).

g++ -O2 -std=c++17 -o chaos chaos.cpp

Run

ChaosEngine speaks the UCI protocol, so it drops into any chess GUI (Cute Chess, Arena, BanksiaGUI) or you can poke it directly in the terminal:

./chaos

uci
position startpos moves e2e4 e7e5
go

It will reply with bestmove <move>. Quit with quit.

The Chaos option

The engine exposes one UCI option, Chaos, ranging from 0 to 100 (default 60):

Value	Behaviour
0	Always plays the best move from a shallow 2-ply search
60	Spicy — material-aware, but with random jitter so it varies
100	Total madness — plays almost any legal move

Set it in a GUI's engine settings, or over UCI:

setoption name Chaos value 100

Play it against Stockfish

Using cutechess-cli (high chaos vs. a deliberately weakened Stockfish so it's not a complete massacre):

cutechess-cli \
  -engine name=Chaos cmd=./chaos proto=uci option.Chaos=100 \
  -engine name=Stockfish cmd=stockfish proto=uci option.UCI_LimitStrength=true option.UCI_Elo=1350 \
  -each tc=10+0.1 \
  -games 10 \
  -pgnout chaos_vs_sf.pgn

Or pit two chaos levels against each other to see how much the tiny search actually helps:

cutechess-cli \
  -engine name=Wild cmd=./chaos proto=uci option.Chaos=100 \
  -engine name=Tame cmd=./chaos proto=uci option.Chaos=10 \
  -each tc=5+0.1 -games 10

How it works

A single file, roughly in this order:

• Board model — a 64-char array (a8=0 … h1=63, standard FEN ordering), plus side-to-move, castling rights and en-passant target.
• Move generation — pseudo-legal moves for every piece, then filtered for king safety (and castling-through-check).
• Evaluation — a tiny material count, nothing fancy.
• Search — a shallow negamax with alpha-beta, used only to keep low-chaos play from total self-destruction.
• Move selection — pickMove() rolls the dice: with probability Chaos% it yeets a random legal move, otherwise it runs the mini-search with a little centipawn jitter so it never plays the same game twice.

Note: ChaosEngine ignores the clock — it moves instantly regardless of time control. That's fine for a random engine. Proper time management would be the first step toward actually making it stronger.

Possible next steps

If you ever want it to win a game:

• Piece-square tables (so it knows knights belong in the centre)
• Deeper iterative-deepening search with real time management
• Quiescence search (stop hanging pieces in capture sequences)
• Transposition tables

---

License

MIT — do whatever you want with it.

---

_{Built with 💜 by R3MU5x}