How to play

You and a teammate sit at the table. From a standard 52-card deck, each of you gets 5 cards. You can't see each other's cards:

Card values. Number cards (2–10) count as their number; J, Q, K count as 11, 12, 13. The ace is the highest card, counting as 14 — and an ace is not a number card.

Three "condition cards" are drawn at random from a set of 20, and shown to both of you:

Your shared question: looking at all 10 cards combined, is at least one of the 3 conditions satisfied? Each of you forms a belief about the probability.

You both then independently place a token on the board, in the column that matches your belief. The board's middle row tells you what each column pays out if the answer to the question is yes/no:

Once you've both placed, you see each other's token. Your beliefs might change by seeing your teammate's token. You then place a second token (round 2), reflecting your updated belief:

The cards are revealed, the conditions checked. Each token scores by its column: if any condition is met you get the left number; otherwise the right one. The middle column gives 7 either way.

Tokens score independently. The team total is your two scores plus your teammate's two, with a maximum of 40 points possible per game.

The sidebar tracks your scores. The main scoring method is to compare your play to how two perfect Bayesians would have played. More precisely: we compute the expected value of your play and the expected value of the Bayesians' play (with expectations taken with respect to the Bayesians' beliefs) and subtract them from each other. Perfect play thus corresponds to a score of zero -- try to get as close as possible! After a game ends, you will see the exact beliefs Bayesians playing the game would have had, providing feedback to improve your play.