Sutom

This little project is about playing the sutom game. It is similar to wordle but the first letter is known.

I often play this game and wanted to implement an algorithm to play it 'well'. This as strong ties to information theory which is a topic I always loved studying and thinking about.

My goal was to think about an information theoretic algorithm by thinking 'from first principles' and recollecting what I know about information theory.

For instance, I did not want to re-watch 3b1b video on his wordle implementation. I will treat myself later with that one. I have already watched it at the time, so everything should already be somewhere in my mind!

Usage

I use uv to manage this project. To get started, simply run:

uv sync // create .venv and install dependancies
source .venv/bin/activate // activate the virtual environment

python main.py

Main script

The main script currently looks something like this:

if __name__ == "__main__":
    ground_truth_word = "amour"
    play(ground_truth_word, player_kind=PlayerKind.AI, max_iter=10)

if you want to play the game, change the player_kind argument to PlayerKind.HUMAN.

The 'AI' player implements a simple sampling strategy I describe here.