Application of Poisson and Dixon-Coles models on football match outcome prediction and research of a positive return over investment in betting market

Abstract

Data analysis has become the main driver of successful decision making in our nowadays world. From startups to big businesses application of statistics over constantly accumulating data has proven to be the key for growth in many industries. Currently, alongside business organizations and high-tech firms, governmental institutions, medical industry and many more rely on insights derived from big data. Usage of proper statistical models over data can increase a firm's profitability, identify a medical test's accuracy, support banks recognize fraud transactions and many more. One of the platforms where application of data analysis has grabbed a great deal of attention is over the most popular sport on earth - Football. Application of statistical models in order to predict football match results has been the center of attention for many people, from topp scientists to bookmakers already for quite some time. Certain techniques have been proposed to find potential statistical models that could be helpful in predicting match score outcomes. And with growing betting industry many have tried to beat bookies with the help of statistical models developed for making prediction for match results. In this paper, indirect approaches, namely Poisson and Dixon-Coles models will be applied to predict match score results. The reason why those models are referred as indirect is due to the fact that regression outputs through those models are goals, rather than direct match outcomes. We will try to beat punctuality of decisions derived from one's "gut feeling", an ambiguous term we will formalize in this paper, through using indirect approaches for match outcome modelling. And at the end, it is found that betting strategy formulated with the use of predictions through such models can yield a positive return through betting in the Premier League over the season 2018-2019.