Visualization-based Iterative Segmentation to Augment Video Game Analytics

Within the field of games, visualization of player log data is becoming an important method for its utility in providing an intuitive and informative way to understand players' experience, which is thus often used by game analytics personnel and game user researchers. Moreover, even players themselves show increased interest in using analytics to quantify and self-improve their performances. Among other types of visualizations, node-edge graphs have proven to be capable of revealing the process of individual and aggregated players, allowing analysts to discover play patterns that can inform game design. However, visualization of player traces often has several disadvantages. First, displaying players' process data does not trivially scale, as tendentially high variance often leads to complex and abstruse graphs. Second, when aggregating all players, individual variations are often overlooked. For example, data from minorities (e.g., casual players or players who played the game very differently than others) are often treated as outliers or noise. In this paper, we present an iterative segmentation approach that allows analysts to interact with the visualization and group players into different subcategories through meta-data or behavioral patterns. Using this approach, analysts can bypass complicated visualizations while protecting significant unique information.