Supporting Aim Assistance Algorithms through a Rapidly Trainable, Personalized Model of Players' Spatial and Temporal Aiming Ability

Abstract

Multiplayer digital games can use aim assistance to help people with different levels of aiming ability to play together. To dynamically provide each player with the right amount of assistance, an aim assistance algorithm needs a model of the player's ability that can be measured and updated during gameplay. The model must be based on difficulty parameters such as target speed, size, and duration, that can be adjusted in-game to change aiming difficulty, and must account for player's spatial and temporal aiming abilities. To satisfy these requirements, we present the novel dynamic spatiotemporal model of a player's aiming ability, based on difficulty parameters that can be manipulated in a game. In a crowdsourced experiment with 72 participants, the model was found to accurately predict how close to a target a player can aim and to converge rapidly with a small set of observations of aiming tasks.