Probabilistic Forecasting: The National Hockey League Totals Market

Abstract

Probabilistic forecasting within the sports setting has become an increasingly popular topic given the rise in participation in sports wagering markets. This paper was concerned with assessing the forecastability of the National Hockey League (NHL) totals market, where the combined number of goals is the sole determinant of wager outcome. Compared to other professional sporting leagues in North America, the NHL has received considerably less research attention and totals betting is arguably the least popular form of sport gambling, making this market prime for research consideration. A logistic regression model was utilized to calculate predicted probabilities and determine whether publicly available wagering information could be used to forecast betting outcomes to a degree that would yield profitable returns. Using betting data from nine seasons, results indicated that the NHL totals market featured a level of forecastability for specific betting strategies, albeit with limited opportunities for profitable returns