Purpose: There are only a few reports that have explored the relationships between player actions in matches and match outcomes in Australian Football (AF). This study extends upon this work, making use of a larger set of performance indicators (Pis), it develops models that provides an indication of the relative importance of Pls to match outcome. This work offers novel insights into AF that can assist strategic and tactical decision making. Methods: Ninety-one (91) team performance indicators (Pis) from the 2001 to 2016 Australian Football League (AFL) seasons were used as Independent variables. The categorical Win-Loss and continuous Score Margin match outcome measures were used as dependent variables. The 16-year database was partitioned into two 8-year eras. Within each era, a feature selection process was applied to the independent variables. Decision tree and Generalised Linear Models (GLMs) were created to describe the relationships between the values of the Pls and match outcome. Results: The decision tree models achieved prediction accuracies between 83.5-88.9% using Win-Loss and 64.4-70.3% when using Score Margin. The GLMs predicted Score Margin achieving root mean squared error of 6.8-7.4 points and a Win-Loss prediction accuracy of 93.1-95.1%. Decision tree and GLMs identified the most important Pis in relation to match outcome to be Metres Gained, Time in Possession, Inside 50s, Inside 50s Per Shot, Turnovers Forced Score and Rebound 50s all in their relative form. Conclusions: In conclusion, the methods used in this work can be Implemented by other sport analysts and the findings can be used to support the decision-making process of coaches, who determine performance strategies. Further, the models created provide novel insights into AF, the most important Pis and the PI characteristics of successful match outcomes.
© Copyright 2018 World Congress of Performance Analysis of Sport XII. Veröffentlicht von Faculty of Kinesiology, University of Zagreb, Croatia. Alle Rechte vorbehalten.