This thesis considers the application of statistical exploratory methods and modelling techniques to sports data. Key to this investigation is the analysis of home advantage and the factors which drive it. The review of literature has shown that much conjecture has been made about the cause of home advantage, but little statistical investigation has been pursued into this area. Building on the model for association football goal counts discussed in Dixon and Coles (1997), reparameterisation to reflect time and team dependent home advantage was explored, alongside the effect of cards on home advantage. Covariate analysis was performed using parametric and semi-parametric models in an attempt to better interpret home advantage by analysing regularly hypothesised causal relationships. Over and under dispersion in goal counts may be the result of variation in team skill or the lack thereof. Censoring and threshold mixture models were explored to try and capture any over or under dispersion, with the aim of creating a more flexible model. As an aside, weighted likelihood based changepoint methods were also explored as a method of considering the reduction in information about the threshold position carried by observations far from the threshold position. Finally, a brief but insightful analysis of changes of performance in golf was carried out. The research contained within can be used to inform statistical models for sports results and impact betting strategies based upon such models.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:702590 |
| Date | January 2016 |
| Creators | Foulds, George |
| Contributors | Tawn, Jonathan ; Brooks, Roger ; Wright, Mike |
| Publisher | Lancaster University |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://eprints.lancs.ac.uk/84465/ |
Page generated in 0.0056 seconds