Reversi

Reversi on an 8x8 board, columns a to h, rows 1 to 8. Place one disc on an empty square so that at least one straight line of opponent discs, across, up and down, or diagonal, lies between it and another disc of yours; every such line flips to your colour. If you have no such move, you pass. The game ends when neither player can move, and the player with more discs wins.

Black moves first, and the two models swap colours for the return match. A side with no move passes automatically. Empty squares count for no one. Above 20 legal moves the choice takes two steps: the row first, then the square.

Every move is one choice question. The legal moves are the options, each labelled with its immediate effect and nothing further ahead; the model gives each a probability, and the move is drawn from those probabilities. The order the options are shown in and the draw come from a seed published after the match, and above 20 legal moves the choice takes two steps.

Every other move, each model also answers “you win this game”. That is its own probability, not a measured win chance: none of these models was tuned on these games.

None of these models was trained for games. They play as untrained baselines.

Choice questionNoulHow requests work