Implementing Random Forests for Hockey Hall of Fame Induction Prediction: Applications to Language-Based Discrimination

• Activity Number: 258
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Sports
• Author(s): Brian M. Mills and Steven Salaga*+
• Companies: University of Michigan and University of Michigan
• Keywords: Random Forest ; Hockey ; Classification ; Multidimensional Scaling ; Logisitic Regression ; Sports

Abstract

Discrimination based on race, ethnicity, age, religion or other factors can significantly affect the labor market and earnings for those individuals whom may be discriminated against. Previous work in the labor market of professional hockey has found evidence of discrimination based on language, specifically against French speaking professional hockey players in English hockey leagues. We extend this research to the subjective voting involved in Hockey Hall of Fame inductions. We first predict current National Hockey League player induction into the Hockey Hall of Fame using the very competitive classification technique, Random Forests, using training data from historical NHL statistical records and previous inductions. Secondly, we extend on previous labor investigations to consider the possibility of induction exclusions of French speaking professional hockey players. We conclude with multidimensional scaling visualizations and rescale the Random Forest votes to induction probabilities using a logistic regression. Preliminary results show little evidence for language-based discrimination and promising prediction error for classification in sports.