This paper shows why, in a world with differing priors, rational agents tend to attribute their own success more to skill and their failure more to bad luck than an outsider. It further shows why each agent in a group might think he or she is the best, why an agent might overestimate the control he has over the outcome, and why two agents' estimated contributions often add up to more than 100%. Underlying all these phenomena is a simple and robust mechanism that endogenously generates overoptimism about one's own actions. The paper also shows how these biases hinder learning and discusses some implications for organizations.