A game where the "winning" move, played by pure logic, pays almost nothing
Two players alternate deciding whether to take a growing shared pot of money or pass it to the other player, who then faces the same choice with a bigger pot, for a fixed number of rounds. Game theory's standard logic says the first player should grab the pot immediately, since the second player would rationally take it on the very last round anyway, and that logic cascades backward to the start. Yet almost nobody actually plays this way. What happens when real people try it?
Reveal the answer
In real experiments, most players cooperate for several rounds before someone defects, walking away with far more money than "rational" backward induction predicts either player should get. The gap between the game-theoretic prediction and actual human behavior, first tested experimentally by Richard McKelvey and Thomas Palfrey in 1992, is one of the most studied puzzles in behavioral economics, suggesting trust can outperform pure self-interested logic.