Learning algorithms where each player independently minimizes regret, converging to equilibrium without explicit coordination.