The Validity of Automated Tackle Detection in Women's Rugby League.

ABSTRACT: Cummins, C, Charlton, G, Naughton, M, Jones, B, Minahan, C, and Murphy, A. The validity of automated tackle detection in women's rugby league. J Strength Cond Res 36(7): 1951-1955, 2022-This study assessed the validity of microtechnology devices to automatically detect and differentiate tackles in elite women's rugby league match-play. Elite female players (n = 17) wore a microtechnology device (OptimEye S5 device; Catapult Group International) during a representative match, which involved a total of 512 tackles of which 365 were defensive and 147 were attacking. Tackles automatically detected by Catapult's tackle detection algorithm and video-coded tackles were time synchronized. True positive, false negative and false positive events were utilized to calculate sensitivity (i.e., when a tackle occurred, did the algorithm correctly detect this event) and precision (i.e., when the algorithm reported a tackle, was this a true event based on video-coding). Of the 512 video-derived attacking and defensive tackle events, the algorithm was able to detect 389 tackles. The algorithm also produced 81 false positives and 123 false negatives. As such when a tackle occurred, the algorithm correctly identified 76.0% of these events. When the algorithm reported that a tackle occurred, this was an actual event in 82.8% of circumstances. Across all players, the algorithm was more sensitive to the detection of an attacking event (sensitivity: 78.2%) as opposed to a defensive event (sensitivity: 75.1%). The sensitivity and precision of the algorithm was higher for forwards (sensitivity: 81.8%; precision: 92.1%) when compared with backs (sensitivity: 64.5%; precision: 66.1%). Given that understanding the tackle demands of rugby league is imperative from both an injury-prevention and physical-conditioning perspective there is an opportunity to develop a specific algorithm for the detection of tackles within women's rugby league.

Need a break? The locomotor and tackle pacing profile and loads of women's rugby league match-play following various between-match turnaround durations.

OBJECTIVES: The study investigated the locomotor and tackle pacing profile and loads of female rugby league players following various between-match turnaround durations. Specifically, the study examined the (1) pacing of locomotor and tackle loads across the time-course of a match and; (2) whole-match and peak locomotor and tackle loads of match-play. METHODS: Microtechnology data were collected from elite female rugby league players (n = 172) representing all National Rugby League Women's teams (n = 6 teams) across two seasons. Players were categorised into backs, adjustables, forwards or interchange players. Data was calculated for the whole-match (m), per minute (m.min-1) and peak (running: m.min-1; acceleration: m.s-2) locomotor and tackle loads (number and efficiency (%)) of match-play. The pacing as well as the locomotor and tackle loads of match-play were examined following short (≤6 days), normal (7 days) or long (≥8 days) turnarounds. RESULTS: The pacing profile of playing positions varied across short, normal and long match turnarounds. Trivial to moderate differences existed in the whole-match, per minute and peak locomotor loads across match turnaround durations (effect size ≤ 1.2). CONCLUSIONS: Following various between-match turnaround durations (i.e., short, normal and long match turnarounds), there were variations in the locomotor and tackle pacing profile and loads whereby, the pacing profile of positional groups was more affected than the load profile. The findings can be used in applied settings to guide the recovery strategies and training plans of female rugby league players to optimise performance and wellbeing across various match turnaround durations.

How fast is fast? Defining velocity zones in women's rugby league.

OBJECTIVES: The study aimed to: 1) apply a data-mining approach to identify velocity zone thresholds for female rugby league players and 2) apply these velocity zones to examine the locomotor demands of match-play. METHODS: Microtechnology data were collected from elite female rugby league players representing all National Rugby League Women's teams (n = 85 players; n = 224 files) over one season. Spectral clustering with a beta smoothing cut-off of 0.1 was applied to each player's instantaneous match-play velocity data for the identification of four zones. To account for outliers within repeated data-points, the velocity zones for each player were calculated as the median. The overarching velocity zones were determined through an incremental search to minimise the root mean square error. RESULTS: Through a data-mining approach, four velocity zones were determined. Rounded to the nearest 0.5 km.h-1 the velocity values across each zone were classified as low (<11.49 km.h-1), moderate (11.50 to 17.49 km.h-1), high (17.50 to 20.99 km.h-1) and very-high (>21.00 km.h-1). Practical application of the zones demonstrated positional group differences in the absolute (effect size (ES): 0.03 to 1.77) and relative (ES: 0.04 to 1.60) locomotor demands of match-play. The back positional group covered greater absolute and relative distances at a very-high velocity than all other positions. CONCLUSIONS: This work informs the velocity zones that could be applied consistently to women's rugby league data within practical (i.e., in the training and monitoring of players) and academic (i.e., as a model for future research to analyse locomotor demands) settings.

Changing gears: data-driven velocity zones to support monitoring and research in men's rugby league.

OBJECTIVES: The study aimed to (1) apply a data-mining approach to league-wide microtechnology data to identify absolute velocity zone thresholds and (2) apply the respective velocity zones to microtechnology data to examine the locomotor demands of elite match-play. METHODS: League-wide microtechnology data were collected from elite male rugby league players representing all National Rugby League (NRL) teams (n = 16 teams, one excluded due to a different microtechnology device; n = 4836 files) over one season. To identify four velocity zones, spectral clustering with a beta smoothing cut-off of 0.1 was applied to each players' instantaneous match-play velocity data. Velocity zones for each player were calculated as the median while the overarching velocity zones were determined through an incremental search to minimise root mean square error. RESULTS: The velocity zones identified through spectral clustering were 0-13.99 km · h-1 (i.e., low velocity), 14.00-20.99 km · h-1 (i.e., moderate velocity), 21.00-24.49 km · h-1 (i.e., high velocity) and >24.50 km · h-1 (i.e., very-high velocity). CONCLUSIONS: The application of spectral clustering (i.e., a data-mining method) to league-wide rugby league microtechnology data yielded insights into the distribution of velocity data, thereby informing the cut-off values which best place similar data points into the same velocity zones. As the identified zones are representative of the intensities of locomotion achieved by elite male rugby league players, it is suggested that when absolute zones are used, the consistent application of the identified zones would facilitate standardisation, longitudinal athlete monitoring as well as comparisons between teams, leagues and published literature.

Women's Rugby League: Positional Groups and Peak Locomotor Demands.

The aims of this study were to (a) use a data-based approach to identify positional groups within National Rugby League Women's (NRLW) match-play and (b) quantify the peak locomotor demands of NRLW match-play by positional groups. Microtechnology (Global Navigational Satellite System [GNSS] and integrated inertial sensors; n = 142 files; n = 76 players) and match statistics (n = 238 files; n = 80 players) were collected from all NRLW teams across the 2019 season. Data-based clustering of match statistics was utilized to identify positional clusters through classifying individual playing positions into distinct positional groups. Moving averages (0.5, 1, 2, 3, 5, and 10 min) of peak running and average acceleration/deceleration demands were calculated via microtechnology data for each player per match. All analysis was undertaken in R (R Foundation for Statistical Computing) with positional differences determined via a linear mixed model and effect sizes (ES). Data-based clustering suggested that, when informed by match statistics, individual playing positions can be clustered into one of three positional groups. Based on the clustering of the individual positions, these groups could be broadly defined as backs (fullback, wing, and center), adjustables (halfback, five-eighth, and hooker), and forwards (prop, second-row, and lock). Backs and adjustables demonstrated greater running (backs: ES 0.51-1.00; p < 0.05; adjustables: ES 0.51-0.74, p < 0.05) and average acceleration/deceleration (backs: ES 0.48-0.87; p < 0.05; adjustables: ES 0.60-0.85, p < 0.05) demands than forwards across all durations. Smaller differences (small to trivial) were noted between backs and adjustables across peak running and average acceleration/deceleration demands. Such findings suggest an emerging need to delineate training programs in situations in which individual playing positions train in positional group based settings. Collectively, this work informs the positional groupings that could be applied when examining NRLW data and supports the development of a framework for specifically training female rugby league players for the demands of the NRLW competition.