ActorLens: Visual Analytics for High-level Actor Identification in MOBA Games

Abstract

Multiplayer Online Battle Arenas (MOBAs) have garnered a substantial player base worldwide. Nevertheless, the presence of noxious players, commonly referred to as "actors", can significantly compromise game fairness by exhibiting negative behaviors that diminish their team's competitive edge. Furthermore, high-level actors tend to engage in more egregious conduct to evade detection, thereby causing harm to the game community and necessitating their identification. To tackle this urgent concern, a partnership was formed with a team of game specialists from a prominent company to facilitate the identification and labeling of high-level actors in MOBA games. We first characterize the problem and abstract data and events from the game scene to formulate design requirements. Subsequently, ActorLens, a visual analytics system, was developed to exclude low-level actors, detect potential high-level actors, and assist users in labeling players. ActorLens furnishes an overview of players' status, summarizes behavioral patterns across three player cohorts (namely, focused players, historical matches of focused players, and matches of other players who played the same hero), and synthesizes key match events. By incorporating multiple views of information, users can proficiently recognize and label high-level actors in MOBA games. We conducted case studies and user studies to demonstrate the efficacy of the system.