out ofdistribution

A model guesses your next move.Keep it guessing.

Your choices and the model’s guesses Your last twelve moves appear here, newest at the bottom. Dots mark your choice. Rings mark the guess. A red line joins them when the guess is wrong.

Tap a side or use the ← → keys.The guess is made before you choose.

© 2026 Worth Building

out of distribution

In machine learning, an input is out of distribution when it differs from the data a model was trained on.

This model looks for sequences of up to four moves. Once it has seen what follows a sequence at least three times, it guesses the more common choice. It tries the longest sequence first.

If no sequence gives it a choice, it uses your overall left and right counts after six moves. Otherwise it guesses at random. Each guess is made before you choose.

A wrong guess does not establish that a move is out of distribution. It tells you that this model was wrong on this turn.

Your choices stay in this tab. Starting over clears the model’s learned patterns and your score.