Several studies have attempted to predict soccer games using various machine learning algorithms. Few of them have succeeded in predicting soccer games with predictive accuracy (PA) as high as 54.6%. This paper aims to predict English Premier League (EPL) soccer games with PA higher than 54.6%. To reach this goal, we build several ordered probit models, and Artificial Neural Network (ANN) using EPL data from 8 seasons (2008-09 season to 2016-17 seasons). The results show that a simple statistical model comes closest to reaching the target set by a complex ANN model: an ordered probit model with 4 predictors has an average PA of 53.5%. However, the model is heavily reliant on betting odds data. Likewise, an ANN model with 4 predictors has an average PA of 50.8%. These results suggest that if we want to build a model with higher PA, which does not rely on betting odds data, then building more complex ANN model may be the key: specifically, building ANN models with more hidden layers and nodes.


Long, Colby


Mathematics; Computer Science


Applied Statistics | Artificial Intelligence and Robotics | Categorical Data Analysis | Data Science | Multivariate Analysis | Probability | Statistical Models


soccer prediction, Artificial Neural Networks, sports prediction, ordered probit, machine learning

Publication Date


Degree Granted

Bachelor of Arts

Document Type

Senior Independent Study Thesis



© Copyright 2021 Shivam K C