Voxel-Based Spatio-Temporal Visualization of Gameplay Traces with Anomaly Detection

As video games take place across space and time, space-time visualizations are promising for conveying gameplay data, as they allow all the data to be shown at once and avoid, e.g., timeline scrubbing. However, space-time visualizations can suffer from clutter and occlusion of data. To address these issues, we explore AI techniques to visualize space-time gameplay traces. Clutter is reduced by aggregating the data of individual traces into voxels. To highlight areas of interest and reduce their occlusion, anomaly detection using isolation forests is used to visually emphasize voxels of interest, changing color and opacity. We present two case studies, based on StarCraft: Brood War and a custom platformer game, demonstrating the impact of the anomaly detection approach.