The Poisson Regression and Quasi-Poisson Regression Analysis on FIFA World Cup Games

Abstract

Fédération Internationale de Football Association or FIFA has gained increasing popularity over time, drawing the attention of a substantial fan base, including enthusiasts in Malaysia who eagerly follow the FIFA World Cup. Previous researchers have sought a deeper understanding of the tournament's dynamics, then using it as a platform to make predictions, assess team performances, and enrich our comprehension of the highest echelons of international football competition.This study aims to predict the number of goals utilisingPoisson and quasi-Poisson regression, and to identify the optimal model through the assessment of AIC, BIC, standard error, and p-values. Next is to determine the most suitable model to be used as predicting the number of goals, with the selection criteria being the lowest AIC, BIC, standard error, and p-values. The dataset used is ‘FIFA World Cup Matches 2022’ originally obtained from the official FIFA website. The findings reveal that covariates such as goal inside the penalty area, goal outside the penalty area, left inside channel, right channel, attempted defensive line breaks, completed defensive line breaks, yellow cards, passes, and own goals were consistently associated with an increased probability of number of goals at at least 0.1 significance level. Overall, the quasi-Poisson regression model proves to be a more appropriate choice than the Poisson regression model for addressing overdispersed data.