Predicting performance at the group-phase and knockout-phase of the 2015 Rugby World Cup

The primary aim of this paper was to produce a model that predicts outcome in the group-phase of the 2015 Rugby World Cup and to determine the relevance and importance of performance indicators (PIs) that are significant in predicting outcome. A secondary aim investigated whether this model accurately predicted match outcome in the knockout-phase of the competition. Data was the PIs from the 40 group-phase games of the 2015 RWC. Given the binary outcome (win/lose), a random forest classification model was built using the data sets. The outcome of the knockout-phase was predicted using this model and accuracy of prediction of the model from the group-phase. The model indicated that thirteen PIs were significant in predicting match outcome in the group-phase and provided accurate prediction of match outcome in the knockout-phase. These PIs were tackle-ratio, clean breaks, average carry, lineouts won, penalties conceded, missed tackles, lineouts won in the opposition 22, defenders beaten, metres carried, kicks from hand, lineout success, penalties in opposition 22 m and scrums won. For the group-phase matches tackle ratio, clean breaks and average carry were accurate standalone predictors of match outcome and respectively predicted 75%, 70% and 73% of match outcomes. The model based on the group-phase predicted correctly 7 from 8 (87.5%) knockout-phase matches. In the knockout-phase clean breaks predicted 7 from 8 outcomes, whilst tackle ratio and average carry predicted 6 from 8 outcomes.