Testing whether a gambling system is profitable

Abstract

The project considers the analysis of data obtained from a sports gambling system. The practical inferential question that arises is whether the proposed system is profitable. To address the inferential question, models are proposed that take into account that different odds are posted on different matches. The translation of the practical problem is seen to be problematic in both the classical and Bayesian framework. We find that it is not so clear as to how the practical problem should be translated into statistical hypotheses involving parameters. In addition, computational difficulties arise for null and alternative hypotheses that form only a small subset of the parameter space. The project does not completely solve the applied problem but rather serves as an initial exploration into the problem with an emphasis on inference and computation.