Bayesian hierarchical models for analyzing and forecasting football results

We develop a Bayesian model for the analysis and forecasting of football match results with a hierarchical structure based on conditionally independent Poisson distributions. The higher levels of the hierarchy reveal dependency through scoring intensities, which are modelled dynamically and stochastically over time. We verify the out-of-sample performance of our model through a betting strategy that is applied to the match outcomes of the 2013–2014 and 2014–2015 seasons of the “Serie A”, the Italian major football league. Exploiting different staking strategies, we show that our model is able to produce a substantial positive return over the bookmakers average odds.