Adaptive Learning in Uncertain and Sequential Competition

In competitive markets, companies often lack access to their rivals’ sales, costs, and strategies. Can they still learn to make optimal decisions? In a new study, Li and Mehrotra show that the answer is yes. Their research demonstrates that even without competitor data, firms can adaptively learn to make near-optimal choices using only their own operational information. More strikingly, when all players follow such self-driven learning, the entire market converges to a Nash equilibrium—the stable state predicted by economic theory—without explicit coordination. The study establishes theoretical guarantees for both convergence rates and regret performance and illustrates the framework in inventory management and dynamic pricing settings. These findings provide a foundation for data-driven decision making in competitive and uncertain environments and offer insights into how markets naturally self-organize.