Given the varying minutes of playing time among players, new analytical approaches are required to quantify full-match (rather than partial match) and ball-in-play (BiP) demands in rugby union. A two-year prospective, longitudinal study was conducted on a single Australian Super Rugby Women`s rugby union team (n = 41 players, 10 matches). GPS-derived metrics (total distance, acceleration load, high-speed running, high-metabolic load distance, acceleration count) and video-coded contact events (tackles, ball-carries, attackers/latches, collisions), total breakdown involvements, scrums, and mauls were modelled using Bayesian multilevel regression modelling. Bayesian models explained 90-98% of data variance (Bayesian R2 : 0.904-0.988) for continuous metrics and 73-92% for count-based metrics (Bayesian R2 : 0.732-0.923). The models demonstrated excellent predictive performance for continuous metrics (Pearson r2 : 0.923-0.990) and good-to-excellent performance for count-based metrics (Pearson r2 : 0.766-0.943). Resulting position-specific full-match reference values for running and contact demands were similar to, or higher than, previous domestic reports. Overall, Bayesian multilevel regression modelling provided robust estimates of full match-play demands and a practical tool to derive position-specific reference values, address data incompleteness, impute missing GPS or contact data, and manage small sample sizes.
© Copyright 2026 Journal of Sports Sciences. Taylor & Francis. Alle Rechte vorbehalten.