Key performance indicators in Australian sub-elite rugby union

Supervised sequential pattern mining of event sequences in sport to identify important patterns of play: an application to rugby union

“…The importance of frequency-based performance indicators (M. D. Hughes & Bartlett, 2002) (PIs) has been studied in rugby through the use of statistical techniques such as the Wilcoxon signed rank test and discriminant analysis (Bishop & Barnes, 2013;Bremner, Robinson, & Williams, 2013;A. Hughes, Barnes, Churchill, & Stone, 2017;Ortega, Villarejo, & Palao, 2009;Vaz, Vasilica, Kraak, & Arrones, 2015;Watson, Durbach, Hendricks, & Stewart, 2017), and more recently, with random forests from machine learning (Mosey & Mitchell, 2019). Rugby has also been analysed at the more granular sequence level by analysing the duration of sequences.…”

“…The necessity for more advanced methods in rugby union was also highlighted by Watson et al (2017) and Colomer, Pyne, Mooney, McKune, and Serpell (2020), who suggested that methods from dynamical systems, machine learning, social network analysis, interpersonal distance and group behaviour, which have been applied in basketball and soccer, remain largely unused in rugby. Machine learning models have been used for the prediction of results in rugby (Mosey & Mitchell, 2019;O'Donoghue & Williams, 2004;O'Donoghue, Ball, Eustace, McFarlan, & Nisotaki, 2016;Reed & O'Donoghue, 2005), while Croft, Lamb, and Middlemas (2015) and Lamb and Croft (2016) used Self-Organising Maps (Kohonen, 1997) to identify important PIs and effective playing styles in New Zealand provincial rugby. Sasaki, Yamamoto, Miyao, Katsuta, and Kono (2017) applied network centrality to identify tactical and leadership structures and to improve the description of complex passages of play at the 2015 RWC, while (Coughlan, Mountifield, Sharpe, & Mara, 2019) applied K-modes cluster analysis to identify particular patterns of play that led to tries in the 2018 Super Rugby season.…”

A Random Forest approach to identify metrics that best predict match outcome and player ranking in the esport Rocket League

“…In traditional sports, identifying the PIs most important for successful task performance helps players and coaches to better direct focus to those key components to accelerate learning and, ultimately, improve performance. Thus, in traditional sport research, many have employed a notational approach to identify PIs in Australian Rules Football 3 , basketball 4,5 , ice hockey 6 , rugby league 7,8 , and rugby union [9][10][11][12][13][14] .…”

“…Moreover, the fact that Random Forests are an amalgamation of many CARTs using a bootstrapped data samples and a random selection of predictor variables for node splitting per tree, they inherently provide much greater predictive ability and reduce propensity for over tting when compared to the CART method alone 25,28 , making them suitable for large datasets. Given the above advantages over existing methods and that Random Forest have been used previously to identify PIs within traditional sports [7][8][9][10]13 they are arguably the most optimal method to identify PIs in Rocket League and esports more broadly.…”

A Random Forest approach to identify metrics that best predict match outcome and player ranking in the esport Rocket League

“…In traditional sports, identifying the PIs most important for successful task performance helps players and coaches to better direct focus to those key components to accelerate learning and, ultimately, improve performance. Thus, in traditional sport research, many have employed a notational approach to identify PIs in Australian Rules Football 3 , basketball 4,5 , ice hockey 6 , rugby league 7,8 , and rugby union [9][10][11][12][13][14] .…”