External authors:

• Alejandro Alvarez ( University of Buenos Aires )
• Guillermo Duran ( Universidad de Chile , Inst Sistemas Complejos Ingn , Consejo Nacional de Investigaciones Cientificas y Tecnicas (CONICET) , University of Buenos Aires )
• Manuel Duran ( University of Buenos Aires )
• Pablo Galaz ( Universidad de Chile )
• Ivan Monardo ( University of Buenos Aires )
• Denis Saure ( Universidad de Chile , Inst Sistemas Complejos Ingn )

Abstract:

This article documents the authors' experience developing an Argentinean website in tribute to Diego Maradona (301060.exactas.uba.ar) that leverages the popularity of football in South America (and the world) to illustrate the application of data science models in sports analytics. In particular, we demonstrate their use in computing probabilities associated with various events (winning matches, advancing rounds, and becoming champions) of the FIFA World Cup Qatar 2022. Building on Dixon and Cole's 1997 seminal model, we develop a competing Poisson model that incorporates for each participating team its attack and defense strengths as well as home-advantage effects. The calibration of the model considers match importance levels and emphasizes the recency of a team's performance. Evaluations of the model's results on various prediction accuracy and error metrics indicate that its performance equals or betters the traditional Poisson model and is similar to established betting sites. Our website featuring the model received over 30,000 visits from 11,000 users across 10 countries during the 2022 World Cup and garnered significant media coverage in Argentina. This successful endeavor underlines the potential of mathematics for predicting football match outcomes but also showcases its potential for countless practical applications and its ability to capture the attention and interest of a wide audience.