Figure Skating Scores: Prediction and Assessing Bias

Abstract

Figure skating has not yet benefited from the statistics craze surrounding America's most popular sports. The literature that does analyze figure skating largely deals with the scoring system abandoned in 2004. This thesis sets out to analyze more recent competitive results and provide a foundation for the analysis of judged sports. For the first time, we assemble all figure skating competition data under the new scoring system in a format amenable to analysis. We build two types of models to predict future competitions by applying linear regression and hierarchical model techniques to fit distributions to previous results. We assess judging bias and show evidence for a small nationalistic bias that rarely impacts rankings.