Prediction Models for Skill-based Video Games

Researchers: Tiffany Clark and Elizabeth Holloway

Introduction
We preformed a cognitive video game study, where we compared novice and expert video game players under an attentional demand tasks. We used the Nintendo classic, Excitebike, on both a GamePad controller and online emulator. We assumed that experts would complete races faster and with few obstacle errors. But we wanted to see if we could reduce expert game play to that of a novice by have participants complete an additional attentional demand task while in game play.

Process
We had both expert and novice participant groups participate in a control series, where they played the game without interruption. Then we had both groups race the same tracks, but this time with an attentional demand task. While in game play, players had to complete an active listening task. They listened to a list of randomly generated numbers, spoken aloud by the researcher, and had to signal if they heard two consecutive odd numbers. 

Methods
Through screen recordings, we were able to observe game play in order to collect data sets on time to completion, obstacle errors and game strategy. We also conducted participant interviews after both the control series and the attentional demand series.

Results
During the control series, experts had faster mean race times than novices, used more game features, and had less obstacle errors. However, during the attentional demand task series, they sacrificed game features and obstacle errors in order to still achieve a better race time. Expert race time still increased significantly, but not to that of a novice.

Using the model created by our data to represent player skill, we wrote a mathematical equation to predict race completion time based on the number of game features used. In other words, the overall skill of a player can accurately predict how quickly they will complete a track with a given number of obstacles.

-0.116*(Number of strategic features used) + 1.64 = Total time