Forecasting of rare events

OData support
Dr. Pach Péter Pál
Department of Computer Science and Information Theory

Nowadays forecasting models are used in various area of life from meteorology to financials. There are several opportunities for applying these methods into sport also. Many studies deal with usability of them for association football.

This work is looking for answer for similar question: is it possible to create continuously profitable forecasting method for the result of football matches.

This thesis has two major goals, first is to create a forecasting system for association football and to find a highly profitable betting strategies using the prediction, second is to investigate the operation of betting market. The motivation of this thesis is to get a part of betting market’s money by beating the bookmakers.

In the first part of the thesis there is a short summary about gambling and betting industry. Furthermore, about how works as well as about their theoretical background and efficiency. The next chapter introduces the assumptions necessary for creating predictor models. Then I describe the used forecasting systems: ELO, Poisson, Bivariate Poisson and their mixed experts. Then there is short introduction about betting strategies. The next chapter shows the result of simulations. Finally, the last section contains a summary and my further development ideas.

Based on my current tests, betting market are not efficient and it can be achievable to find a good prediction model with a profitable betting strategies.


Please sign in to download the files of this thesis.