The Application of Hurdle Models to Accurately Model 0-0 Draws in Predictive Models of Football Match Outcomes

A novel application of so-called hurdle models is presented in the context of association football match prediction models, which has the flexibility of being able to capture either the deflation or inflation of the probability of 0-0 scores. The deployment of the model in real time on week by week basis to develop one-round ahead predicted forecasts is discussed in the context of a Bayesian framework. The problem of forming appropriate informative prior distributions for parameters that need to satisfy sum to zero constraints is also developed. Using match results data from the Scottish Premier League, the use of the hurdle model approach is shown to improve the predictive accuracy of the probability of 0-0 scores and can lead to a potentially profitable betting strategy when used for betting on the 0-0 score market.