Telemetric data analysis of Real-time strategy video games; a new 'drosophila' for the cognitive sciences
My talk will describe my lab’s current research exploring human learning in the context of real-time strategy (RTS) games. RTS games are an ideal domain to study expertise for several reasons. First, they are a domain of legitimate expertise. Professional players train full time, practicing 6-9 hours a day and typically have played RTS games for over a decade. Top players can earn $250 000 (US) annually. Tournaments are televised live and professional teams are sponsored by major corporations. Second, the game itself requires a rich set of perceptual, attentional, motor, and decision-making skills, and thus offer many opportunities for insights. Third, each game produces an enormous amount of behavioral data: an average game of chess, for instance consists of 40 moves per player, while the average RTS game in our study consists of 1635 actions per player. Finally, by studying a domain in which expert performance is almost entirely computer-based we can non-invasively obtain accurate measurements of cognitive-motor performance within the domain of expertise itself. But, solving the challenges of data collection leads to datasets so large and complex that extracting useful findings requires creative data analysis techniques. Our research program thus illustrates the importance of statistical tools and training for the 21st century social sciences.