Large-Scale Deep Learning to Augment Production RL Workloads at Riot Games

Explore how Riot Games leverages large-scale deep reinforcement learning to create game bots with varying skill levels, providing valuable insights for designers and enhancing player experiences. Discover the innovative approach used in Team Fight Tactics, where large neural networks predict outcomes instead of relying on game servers. Learn about the supervised learning process and how Ray Data, Ray Train, and Ray Tune were utilized to simplify and scale the project. Gain insights into controlling and tuning game servers to maximize training efficiency for bots learning gameplay. This 31-minute conference talk from Anyscale's Ray Summit 2022 offers a deep dive into cutting-edge applications of machine learning in game development.