Clustering Players' Strategies for Generation of Non-Player Characters in Mythic Legends

Explore a 29-minute conference talk from Data Science Conference Europe 2023 that demonstrates how machine learning accelerates the development of non-player characters (NPCs) in gaming. Learn how clustering techniques were applied to analyze players' army formations in Mythic Legends, an auto-chess inspired strategy role-playing mobile game, to create more human-like NPCs. Discover the challenges of defining appropriate distance functions for clustering and how they were overcome through a pragmatic approach involving multiple clustering steps with partial distance functions. Understand how a greedy algorithm was implemented to manage combinatorial explosion in the process. Delivered by Rok Piltaver in Belgrade, this presentation offers valuable insights into practical applications of machine learning in game development, specifically focusing on improving NPC behavior generation while reducing development and quality assurance efforts.