Adversarial Multi-Armed Bandit with Variable Plays This code is for paper: Adversarial Online Learning with Variable Plays in the Evasion-and-Pursuit