Joint modelling of goals and bookings in association football

Abstract

Summary: A multivariate counting process formulation is developed for the quantification of association football event interdependences which permits dynamic prediction as events unfold. We model data from English Premier League and Championship games from the 2009-2010 and 2010-2011 football seasons and assess predictive capacity by using a model-based betting strategy, applied prospectively to available live spread betting prices. Both the scoreline and the bookings status were predictive of match outcome. In particular, the award of a red card led to increased goal rates for the non-penalized team and the home team scoring rate decreased once they were ahead. Overall the betting strategy profited with gains made in the bookings markets.