Football Predictions Made Easy by Anthony Constantinou

Scholar and machine expert Anthony Constantinou has come up with a new way of determining football scores using the Bayesian network model. The network works using probability and represents the conditional dependencies among uncertain variables. The variables can be objective or subjective. When using the Bayesian network model to predict Association football matches, the subjective variables represent the factors that are vital for prediction but have not yet been captured by historical data.

Anthony Constantinou has used the pi football model to generate predictions about the English Premier League matches in the 2010/11 season. Due to its success, the program can now be used in other league matches. To prove the legibility of the program, the predictions were displayed before the games begun. He further showed that the football forecasting models only needed to show three probability values, which are the away win, home win and draw. The biggest challenge is that the various scoring rules used for validation in the previous studies do not recognize that football outcomes are showcased in a ranked scale. Anthony Constantinou and his team are now looking into Rank Probability score, which has been missed by previous researchers but proves to perfect asses football forecasting models.

Anthony draws his expertise from his experience as an assistant professor in machine learning and data mining in the Queen Mary University of London. He is the current head of the Bayesian Artificial Intelligence Research lab. Besides, he is recognized as one of the prime Rating systems and Bayesian networks consultants. One if his significant success was in May this year when his model ‘Dolores’ was ranked 2nd in the Machine Learning for international soccer competition. It was also published in the Machine Learning Journal. Anthony was also awarded a fellowship by the prestigious Engineering and Physical Sciences Research Council.

More on Anthony Constatinou: