A Critical Comparison of Machine Learning Classifiers to Predict Match Outcomes in the NFL

Abstract

In this paper, we critically evaluate the performance of nine machine learning classification techniques when applied to the match outcome prediction problem presented by American Football. Specifically, we implement and test nine techniques using real-world datasets of 1280 games over 5 seasons from the National Football League (NFL). We test the nine different classifier techniques using a total of 42 features for each team and we find that the best performing algorithms are able to improve one previous published works. The algoriothms achieve an accuracy of between 44.64% for a Guassian Process classifier to 67.53% with a Naïve Bayes classifer. We also test each classifier on a year by year basis and compare our results to those of the bookmakers and other leading academic papers.

Predizendo os Vencedores dos Playoffs: Um Estudo de Caso com Aprendizado de Máquina em Partidas de Futebol Americano

“…- [Beal et al 2020] Comparar diferentes modelos de AM na classificação do resultado de partidas da NFL.…”

“…Um exemplo é a National Football League (NFL), o principal campeonato de futebol americano, que movimenta bilhões de dólares anualmente [Durand et al 2021]. Como resultado, diversos pesquisadores e apostadores têm se empenhado em enfrentar um grande desafio: a previsão dos resultados das partidas [Beal et al 2020]. Conforme [Beal et al 2020], tal desafio é fundamental para muitas partes interessadas no esporte, como as equipes para selecionar suas táticas, bem como as casas de apostas que definem probabilidades.…”

“…Como resultado, diversos pesquisadores e apostadores têm se empenhado em enfrentar um grande desafio: a previsão dos resultados das partidas [Beal et al 2020]. Conforme [Beal et al 2020], tal desafio é fundamental para muitas partes interessadas no esporte, como as equipes para selecionar suas táticas, bem como as casas de apostas que definem probabilidades. No entanto, prever resultados de jogos esportivos é um desafio complexo que requer a consideração de diversos fatores, tais como estatísticas da equipe, o ânimo/moral dos jogadores e a localização do jogo, entre outros.…”

A Learn-to-Rank Approach for Predicting Road Cycling Race Outcomes

“…Besides a variety of sports, there is also a variety of ML methodologies used to predict the outcomes. Beal et al (2020) did an extensive comparison of ML techniques for predicting outcomes of the National Football League (NFL). The authors found that the best performing techniques were Nave Bayes, AdaBoost, and Random Forest with accuracies of 68, 66, and 64%, respectively.…”

Features selection in NBA outcome prediction through Deep Learning

“…Considering the large interest and the increasing volume in sport betting, it is easy to understand the reason why the number of attempts in predicting games' results is continuously increasing, see for instance Bunker and Thabtha, 2019;Hubáček et al, 2019. Machine learning techniques for outcome prediction have been widely applied (Haghighat et al, 2013), covering all professional sports, from horse races (Davoodi and Khanteymoori, 2010) to hockey (Gu et al, 2016) and from American football (Beal et al, 2020;David et al, 2011;Kahn, 2003;Purucker, 1996) to football (Carpita et al, 2019;Min et al, 2008;Tax and Joustra, 2015), just to give some examples among others. Also basketball, of course, has been investigated under this perspective.…”