We propose a model for the seat inventory control game between two airlines. We do not make probabilistic assumptions on demand, and use worst-case analysis focused on the regret criterion. We prove the existence and uniqueness of a pure-strategy Nash equilibrium for two fare classes. We prove that when two airlines are equivalent in fare structure, capacity and market share, more seats are allocated to lower fare at the route level in the centralized problem compared to the competitive problem. We quantify the average and worst-case performance of competitive and centralized solutions using computational experiments.