Scheduling policies are a key part of HPC resource management. These policies outline where and when to start which compute job on the platform. In this presentation, we study various aspects of the user-centric view of the system. Is there a best scheduling policy in general for objectives such as the average waiting time or average slowdown? Can a good policy be learned on a given system? Is simulation of the system helpful, when applicable? Can a policy be learned on the fly, in an online setting, and how? We will answer these questions via a comprehensive experimental study of priority- based policies using the EASY-backfilling framework.