A carregar...
Publicação
Modelos Bayesianos Aplicados à Previsão de Resultados de Futebol
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 808.75 KB | Adobe PDF |
Orientador(es)
Resumo(s)
Em qualquer tipo de desporto a competividade entre jogadores, equipas e espetadores é algo
inerente à actividade. O futebol, sendo um dos mais praticados e atraindo enormes multidões pelo
mundo inteiro, o interesse na previsão de resultados e a introdução de modelos estatísticos para esse
fim era uma inevitabilidade, seja por motivos de curiosidade ou incentivos monetários, como no caso
do mercado de apostas. Com este objetivo, são propostos modelos Bayesianos hierárquicos para
modelar o desempenho das equipas participantes de modo a ter uma ideia de como elas se comparam
entre si. São considerados dois tipos de abordagens: a primeira onde se modelam os golos marcados
pelas duas equipas, para um dado jogo e a segunda que modela os resultados finais do jogo na
forma de um vetor (vitória, derrota, empate). Para além da previsão, pretende-se ainda comparar
estas abordagens e usar as previsões realizadas para verificar se existe superioridade por parte de
uma das abordagens e dentro dessa qual o modelo mais adequado. No total são aplicados quatro modelos bayesianos a dados provenientes da temporada 2014-2015 da Primeira Liga Portuguesa de Futebol.
In any type of sport the competitivity between players, teams and spectators is something that is inherent to the activity. Football being one of the most practiced sports and attracting huge crowds all over the world, the interest in the prediction of results and the introduction of statistical models for that end was an inevitability, either by motives of curiosity or monetary, in the case of the betting market. For this purpose, we propose Bayesian Hierarchical models to model the strength of the participating teams in the league to gain an understanding of how the compare between each other. We consider two types of approaches: the first where the goals scored by both teams are modeled, for any given game, and the second models the results of the game themselves in the form of a (win, tie, defeat) vector. Besides prediction, we also want to compare these approaches and use the predicted data to verify if any of these approaches is superior and inside each, find what the models are the better fit. In total, we apply four Bayesian models to data from season 2014-2015 of Primeira Liga from the Portuguese Football League.
In any type of sport the competitivity between players, teams and spectators is something that is inherent to the activity. Football being one of the most practiced sports and attracting huge crowds all over the world, the interest in the prediction of results and the introduction of statistical models for that end was an inevitability, either by motives of curiosity or monetary, in the case of the betting market. For this purpose, we propose Bayesian Hierarchical models to model the strength of the participating teams in the league to gain an understanding of how the compare between each other. We consider two types of approaches: the first where the goals scored by both teams are modeled, for any given game, and the second models the results of the game themselves in the form of a (win, tie, defeat) vector. Besides prediction, we also want to compare these approaches and use the predicted data to verify if any of these approaches is superior and inside each, find what the models are the better fit. In total, we apply four Bayesian models to data from season 2014-2015 of Primeira Liga from the Portuguese Football League.
Descrição
Tese de mestrado, Estatística e Investigação Operacional (Estatística), 2023, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Inferência Bayesiana Dados de futebol MCMC Previsão de resultados Ajuste de modelos Teses de mestrado - 2023