Modeling Visually-Guided Aim-and-Shoot behavior in First-Person Shoters

In first-person shooters, players aim by aligning the crosshair onto a target and shoot at the optimal moment. Since winning a match is largely determined by such aim-and-shoot skills, players demand quantitative evaluation of the skill and analysis of hidden factors in performance. In response, we build a simulation model of the cognitive mechanisms underlying aim-and-shoot behavior based on the computational rationality framework. Unlike typical aimed movements in HCI, such as pointing, the aim-and-shoot offers a unique task scenario: as players move the mouse with their hand, the first-person view camera rotates, which in turn directly affects the target’s visible position on the screen. To realistically simulate such complex mechanisms, we model players’ perceptual, decision-making, and motor processes more sophisticatedly than any existing model. Model fitting based on amortized inference showed that our model could successfully reproduce the behavior of 20 FPS players (10 professionals) on several key measures, outperforming a baseline. Additionally, model fit parameters revealed that professionals had distinct cognitive or motivational characteristics.