Predicting bookmaker odds and efficiency for UK football

Abstract

The efficiency of gambling markets has frequently been questioned. In order to investigate the rationality of bookmaker odds, we use an ordered probit model to generate predictions for English football matches and compare these predictions with the odds of UK bookmaker William Hill. Further, we develop a model that predicts bookmaker odds. Combining a predictive model based on results and a bookmaker model based on previous quoted odds allows us to compare directly William Hill opinion of various teams with the team ratings generated by the predictive model. We also compare the objective value of individual home advantage and distance travelled with the value attributed to these factors by bookmakers. We show that there are systematic biases in bookmaker odds, and that these biases cannot be explained by William Hill odds omitting valuable, or excluding extraneous, information.

pi-football: A Bayesian network model for forecasting Association Football match outcomes

“…Overall, pi-football won approximately 35% of the bets simulated under all of the three scenarios, with the mean odds of winning bets at approximately 3.00. This suggests that the model was able to generate profit via longshot bets; what makes this especially interesting is that longshots are proven to be biased against the bettors (Cain et al, 2000, Forrest & Simmons, 2001Forrest et al, 2005;Graham & Stott, 2008;Constantinou & Fenton, 2012b). This implies that the model would have generated even higher profits if the betting market was to provide unbiased odds.…”

“…In (Dixon & Coles, 1997) authors claimed that for a football forecast model to generate profit against bookmakers' odds without eliminating the in-built profit margin it requires a determination of probabilities that is sufficiently more accurate from those obtained by published odds, and (Graham & Stott, 2008) suggested that if such a work was particularly successful, it would not have been published. Ours is the first study to demonstrate profitability against all of the (available) published odds.…”

“…Figure 11 illustrates the Risk of Ruin for up to a bankroll 100 times the value of a single bet. A bankroll of ~£55 (or 55 times the value of a single bet) and ~£45 is required to ensure that the probability to lose the specified bankroll **** See also the following studies on the football gambling market: (Pope & Peel, 1989;Dixon & Coles, 1997;Kuypers, 2000;Rue & Salvesen, 2000;Forrest & Simmons, 2001;Dixon & Pope, 2004;Goddard & Asimakopoulos, 2004;Forrest & Simmons, 2008;Graham & Stott, 2008). † † † † The bookmaker's odds are also provided by (Football-Data).…”

Systematic Positive Expected Returns in the UK Fixed Odds Betting Market: An Analysis of the Fink Tank Predictions

“…There is also a home advantage rating, which allows for the fact that clubs score more goals when playing at home. An early version of the Fink Tank model appeared as Graham and Stott [18].…”

Ordinal versus nominal regression models and the problem of correctly predicting draws in soccer

“…Kuypers (2000) argued that ordinal regression makes sense, since match outcomes are naturally ordered, and therefore did not test any nonordered regression models. Ordered regression models have later been used for prediction of match results by Koning (2000), Forrest and Simmons (2000), Dobson and Goddard (2001, 2003, Audas, Dobson, and Goddard (2002), Goddard and Asimakopoulos (2004), Goddard (2005), Forrest, Goddard, and Simmons (2005), Graham and Stott (2008), Hvattum and Arntzen (2010), and Hvattum (2015). Some research has relied on the use of ordered regression models as a means to normalize bookmakers' odds, such as by Štrumbelj (2014, 2016).…”