A Statistical Model of the 2013 Rugby Sevens World Cup

The purpose of this study was to establish a predictive model capable of simulating and predicting the outcome of the 6th Rugby Sevens World Cup, held in Moscow on the 28 th to the 30 th June 2013. Following a review of predictive modelling approaches in sport, a multivariate regression model was attempted, using International Rugby Board (IRB) ranking points and various travel effect descriptors (relative time zone, altitude, average temperature, estimated flight time and circadian rhythm offsets) to predict pointsdifference of international Rugby Sevens matches. A whole database approach was used to generate the model, based on the outcomes of all IRB Series matches dating from 2004 to the fifth round of 2013 (n = 3240). A step wise approach to establishing a model was attempted, though none of the travel effects tested correlated, with only IRB ranking difference between teams improving the model. Accordingly a bivariate model was used in simulating the tournament structure. The model correctly predicted New Zealand as tournament winners, as well as six of the eight cup quarterfinalists, which is in line with the tournament seeding. Half of the tournament final standings were within +/1sd of the prediction. The current research establishes that international Rugby Sevens match outcomes are dominated by difference in relative strength demonstrated over previous ranking events, and are largely insensitive to the travel effects estimated in this study. Meanwhile analysis of the input dataset revealed a 24.7% upset frequency in the period 2004-2013, which is higher than that presented in other texts for fifteen a side Rugby Union.