The use of technical-tactical and physical performance indicators to classify between levels of match-play in elite rugby league.

Abstract

This study aimed to identify which physical and technical-tactical performance indicators (PI) can classify between levels of rugby league match-play. Data were collected from 46 European Super League (ESL) and 36 under-19 Academy (Academy) level matches over two seasons. Thirty-one ESL players and 41 Academy players participated. Microtechnology units were used to analyse the physical PI and matches were videoed and coded for individual technical-tactical PI, resulting in 157 predictor variables. Data were split into training and testing datasets. Random forests (RF) were built to reduce the dimensionality of the data, identify variables of importance and build classification models. To aid practical interpretation, conditional inference (CI) trees were built. Nine variables were identified as most important for backs, classifying between levels with 83% (RF) and 78% (CI tree) accuracy. The combination of variables with the highest classification rate was PlayerLoad2D, PlayerLoadSLOW per Kg body mass and high-speed running distance. Four variables were identified as most important for forwards, classifying with 68% (RF) and 64% (CI tree) accuracy. Defensive play-the-ball losses alone had the highest classification rate for forwards. The identified PI and their unique combinations can be developed during training to aid in progression through the rugby league playing pathway.