Prediction of bids for two-outcome gambles in a casino setting

Abstract

An experimental game was conducted in a Las Vegas casino. Ninety two-outcome gambles were presented in turn, and the player's monetary evaluation of each gamble was determined by the Marschak bidding procedure. Ordinal tests supported the monotonicity and the cancellation properties that are necessary for any from the family of expectation models. A number of different parametric expectation models were tested by a least-squares method. The bids were well predicted by the expected value (EV) of the gambles. More elaborate predictors, in which parameters were estimated to represent the individual player's subjective probabilities and/or utilities, merely capitalized on noise; cross-validation showed that no real improvement could be obtained over the EV predictor. Furthermore, no single feature of the bet nor any linear combination of them did as well as EV.