The Academic World of Sports Modelling

Dr. Alun Owen, a lecturer in statistics at Loughborough University, discusses the academic world of sports modelling and its publications, papers and conferences.

Categories: All Sports, Professional, Statistical models, Tools

I’m a lecturer in statistics at Loughborough University with research interests in the modelling and analysis of sports data, although I have recently made the step into my own prop trading with a statistical modelling based trading service at www.systematicsoccer.com.

My focus here though is to share with you what I am aware of in terms of some of the more recently published academic research and other work, that relates to my interests in modelling football and horse racing, and to highlight some relevant learned organisations and upcoming events that might be of interest to you. What is mentioned below is not meant to be exhaustive, rather a snippet of the vast amount of information that is potentially available.

In terms of some of the recently published work in football (soccer) modelling, Koopman and Lit (2013) outline an approach to developing a predictive goals based model, but which also allows teams’ attacking and defensive abilities to vary over time (i.e. to be modelled stochastically). In fact this presents an approach to modelling time-varying abilities that is an alternative to an approach I developed and published in Owen (2011). However Koopman and Lit’s paper represents a significant and interesting development since they apply their approach to modelling the time-varying aspect of teams’ abilities to a model which also includes aspects of the bivariate Poisson goals models of both Dixon and Coles (1997) and Karlis and Ntzoufras (2006). Published work on predictive models in football is quite rare to find these days, I guess due to the potentially commercially valuable nature of the IP contained within that type of material. However, I am keen to look again at my approach to modelling the time-varying aspects of teams’ abilities and plan to do some work this summer, perhaps with a view to presenting this at the Statistics in Sports stream at the next Royal Statistical Society International Conference. In a different, but still very relevant area of football modelling, McHale and Szczepański (2014) published very recently on their work on modelling the goal scoring abilities of players, with some very interesting findings indeed and so well worth a read.

If your trading interests are in horse racing rather than football, then the “Efficiency of Race Track Betting Markets” edited by Haush, Lo and Ziemba is a must. This includes the classic works on developing a statistical model for horse racing by Benter (1994) and Chapman (1994). The latter being a follow up to Bolton and Chapman (1986). There is also CX Wong’s book “Precision: Statistical and Mathematical Methods in Horse Racing”, which has much to offer, although the translation from Chinese to English is not always perfect! I can also recommend you have a look at the Smartersig website at www.smartersig.com and the web site of Smartsigger, which to some extent has taken over Smartersig’s mantle at www.smartsigger.com. Both sites offer numerous articles, utilities and a forum of direct relevance to betting on horse racing. I mention these sites as I previously wrote articles regularly for Smartersig (the magazine stopped in October 2013) and now write regularly for Smartsigger. At the moment I am writing a series of articles for Smartsigger which has just introduced a practical series on how to implement a predictive horse racing model using R based on the multinomial logistic regression model.

Looking now at other academic sources of information for the sports trader, there are a number of academic organisations and events that are held regularly, which might be potentially relevant to anyone interested in sports trading. For example the Royal Statistical Society (RSS) is the UK’s learned society for statistics and a professional body for statisticians, and runs numerous meetings and events each year. In particular, the RSS has a “Statistics in Sport Section”, for which I am currently Section Secretary. This aims to bring together a community of researchers interested in statistics in sport or those working in this field, to share ideas relating to research and applications of statistics in sport, and to support the communication of accurate and informed information on sports statistics to the wider community. The Statistics in Sport Section of the RSS usually runs twice-yearly meetings that consist of one or two presentations of relevance to the use of statistics in sport, followed by an opportunity to network with other people sharing an interest in this area. Since the section’s inception two years ago, there have been talks on prediction models for the Euro 2012 tournament by Dr Stuart Coles of Smartodds, in-play models for football by Professor John Goddard of Bangor University, an overview of sports analytics by representatives from Opta and IBM, rating systems by Professor Mark Glickman from Boston University School of Public Health, USA and more recently in December last year we had two talks on detecting corruption in football by Dr Giambattista Rossi from UCL and Dr James Reade from Reading University. This last set of talks couldn’t have been timed better if we had tried, given the publicity surrounding the issue of corruption in football that exploded onto the news channels at that time. The date for the next meeting of the Statistics in Sport Section of the RSS has not been set but this is likely to be held sometime in June this year.

Another significant event worth highlighting is the bi-annual mathematics in sport conference entitled MathSport International. Interest in this conference comes from a very broad audience and includes not just academics (statisticians, mathematicians, and computer scientists etc.), but also bookmakers and some professional sports modelling and betting/investment organisations, all looking for that edge with their models. The (next) 5th Mathsport International conference is actually being organized by myself along with colleagues at my own institution, Loughborough University in the UK, and is due to be held there in June 2015. Further details are available on our website and we currently interested in hearing from potential sponsors for this three day international event so do get in touch with me (at the email address given at the head of this article) if sponsorship of this even is of potential interest to you.

Previous MathSport International conferences have been held in Belgium in 2013, UK in 2011, Netherlands in 2009 and UK in 2007. To give you a flavor of the talks, last year’s conference (Belgium 2013) covered a wide range of sports including football, tennis, table tennis, NFL, basketball, AFL, baseball, volleyball, golf, athletics, rugby, cycling, cricket, orienteering, hockey and skiing. In addition, there were talks looking at betting strategies, staking and corruption. The MathSport International conference is held bi-annually, since it alternates with the Australasian Conference on Mathematics and Computers in Sport with their next (12th) due to be held in Darwin, Australia in June this year.

Other potentially relevant conferences include that organized by the International Association of Computer Science in Sport (IACSS). Their last conference was held in Instanbul, Turkey in 2013. Topics included for example, a talk by a colleague of mine from Loughborough University, Dr Chris Dawson, on the prediction of football match outcomes using artificial neural networks. Other notable talks include that by Kiyoshi Osawa from the Japan Institute of Sport Sciences, on the effect of fielding results on winning percentage in baseball, and Stuart Morgan from the Australian Institute of Sport, who had a paper on the prediction of shots and point outcomes in Tennis. The IACSS Conference in 2014 is actually scheduled to take place jointly with the 2014 Australasian Conference on Mathematics and Computers in Sport in June this year. The next IACSS Conference in 2015 is then actually scheduled to be held also at my institution Loughborough University in 2015, albeit later in the year in September.

Finally on the topic of conferences I should mention the MIT Sloan Sports Analytics Conference that is usually held in March each year. This also features presentations specific to predictive sports betting. There is lots of interesting information, including some videos of some of the talks on their website at http://www.sloansportsconference.com.

Of course there is much more I could mention than I have space to cover her, but it would be remiss of me not mention the book “The Numbers Game: Why Everything You Wanted to Know About Football is Wrong“ written by Chris Anderson and David Sally. This was published last year and exerts from the text have been published in the media and has a number of snippets if information useful to the football trader. The only disappointing thing from my perspective was that they included one or two snippets that I knew about but I hadn’t seen these published anywhere and so thought not many people were aware of them – the joys of keeping ahead! Chris Anderson by the way keeps a very interesting blog which I can recommend at http://www.soccerbythenumbers.com.

I hope you found some of this useful? Do please get in touch if you wish to discuss anything with me, especially if it relates to modelling football or horse racing, or trading.