Estimativa de probabilidades esportivas via modelos de Poisson e de Dixon-Coles

Abstract

This work explores the Poisson and Dixon-Coles mathematical models used to predict the probabilities of match results between soccer teams in a championship. To understand these models, we provide a consistent statistical foundation, addressing fundamental topics such as Probability Mass Function, Normal Distribution, Bernoulli Distribution, Poisson Distribution, and Likelihood Function, which form the basis of Linear Models and Generalized Linear Models. We implemented the Poisson and Dixon-Coles models in Python. We chose to use the DixonColes model to create estimates for matches in the 2025 Brazilian Men’s Soccer Championship, Series A. This allows us to compare these estimates with the actual results of a match and with the odds offered by bookmakers, enabling us to make relevant observations about these comparisons. The sequence of suggested activities aims to contribute to the development of decision-making skills and the prevention of risky behavior, as well as to help elementary school math teachers address the topic in the classroom in a coherent manner, in order to show the harmful reality of gambling.