Match madness : Probability matching in the prediction of the NCAA basketball tournament

Every year, billions of dollars are spent gambling on the outcomes of the NCAA men s basketball tournament. The current work examines how individuals make predictions for one of the most popular forms of betting, tournament pools, in which individuals must correctly predict as many games in the tournament as possible. We demonstrate that individuals predict more upsets (i.e., wins by a higher-seeded team) than would be considered rational by a normative choice model, and that individuals are no better than chance at doing so. These predictions fit a pattern of probability matching, in which individuals predict upsets at a rate equal to past frequency. This pattern emerges because individuals believe the outcomes of the games are non-random and therefore predictable