Methods for the identification and optimal exploitation of profitable betting scenarios

This thesis tackles the issue of how gamblers can profit from betting on the outcome of sporting events. In particular, it addresses issues which have arisen in recent years concerning both the inception of betting exchanges, and the technique of building complex statistical models to accurately predict the sporting outcomes. This thesis shows that bias in predictive models can be quantified from a collection of model outputs. It is shown that a Bayesian method can be constructed to derive accurate bias estimates, even when the model outputs are merely a collection of independent Bernoulli trials. In addition, the method is expanded, to allow the quantification of a time-varying bias, as long as it changes in a known, deterministic setting. The utility of this method is demonstrated via the correction of a simple football prediction model. The movements seen in betting markets before the event in question occurs are investigated. It is conjectured that the rate of increase of the amount of capital invested in the betting market is central to understanding other market movements. With this in mind, two approaches are derived, which both use a collection of historic market movements for past events for their predictions. It is shown that in many cases, some mix of the two approaches achieves the most accurate forecasts. A new gambling strategy, dubbed consolidated wagering is introduced. It is demonstrated that consolidated wagering outperforms all other candidate methods when considering string bets (multiple bets on the same event, at different odds). The application of these methods to investing in restricted markets in betting exchanges is demonstrated. Finally, the problem of string wagers under uncertainty is explored.

Name dc:type.qualificationname

Ph.D.

Level dc:type.qualificationlevel

doctoral

Grantor dc:publisher.institution

Lancaster University

Year dc:date.issued

2016

Authors dc:creator

• Flowerdew, Tom
• Kirkbride, Christopher
• Tawn, Jonathan
• Glazebrook, Kevin