Jeffrey Zeidel

In strategic environments, it is often the case that player choices and random events combine to determine outcomes. We conduct a laboratory experiment using a game with a unique Nash equilibrium in mixed strategies, played repeatedly in fixed pairs, where payoffs are determined by a random draw with probabilities that depend on the chosen action profile. Draw probabilities are fixed and are common knowledge, and players receive full feedback about choices and draw realization after every round. We find that subjects are reinforced by both lucky outcomes and expected payoffs conditional on choices. As luck becomes more important to payoffs, play diverges further from equilibrium. We use the data to estimate and evaluate several dynamic learning models, and find that a selective attribution of credit model, in which subjects credit their own choices for lucky outcomes, and ignore unlucky outcomes, explains the data best.