Identifying expertise : data exploration in Tetris

Abstract

The identification of expertise in a complex task is trivial when all data is in after the fact. To achieve a better understanding of the nature of expertise my first goal, in this work, is to identify the behaviors most predictive of different levels of expertise (novice to expert) in the video game Tetris. The second goal is to examine how little data is required to accurately predict levels of expertise. The present study analyzes potential behavioral indicators of expertise in Tetris at three levels (global, local, and immediate) and under four levels of data quantity. Presented are statistical models using the established metrics of behavior to predict overall output performance. Results indicate mild success in predicting performance, offering a starting point for analyzing other complex and dynamic tasks with explicit criteria for successful performance.