GPS-Derived Spatial–Temporal Metrics for Characterizing Collective Tactical Behavior in Elite Rugby Union

Spatial–temporal metrics offer a promising framework for quantifying collective tactical behavior in rugby union, yet their application across multiple matches and try-scoring contexts remains limited. Building on a single-match case study introducing these metrics to rugby union, this study examined how collective spatial–temporal organization relates to try-scoring in elite professional rugby union. GPS-derived positional data from 118 players across six Super Rugby matches were analyzed for the 30 s preceding 43 try-scoring events. Centroid distance, convex hull, stretch index, and approximate entropy (ApEn) were calculated for both attacking and defending teams. Linear mixed-effects models assessed the influence of try source, field location, and match quarter on spatial–temporal organization. Linebreaks (23%) were the most frequent try-scoring precursor. No significant between-phase differences were observed in stretch index or convex hull, in contrast to our earlier single-match case study, suggesting that this finding may not generalize across matches and warrants further multi-match verification. Tries from linebreaks, mauls, and offloads occurred at significantly greater inter-team centroid distances compared to tackle breaks (e.g., linebreaks: 3.6 m greater than tackle breaks, p = 0.023), while offloads and wide passes were associated with higher centroid distance ApEn, indicating greater spatial unpredictability. It appears that inter-team spacing and its unpredictability, rather than unilateral spatial expansion, are key spatial–temporal features of successful attacking play. Centroid distance metrics offer coaches and analysts a practical means of benchmarking attacking spacing by try source and field zone, while ApEn provides a complementary indicator of movement predictability that could inform defensive alignment strategies against attacking sets associated with greater tactical unpredictability. As only successful try-scoring sequences were analyzed, and several try source comparisons (e.g., mauls, grubber kicks) were based on very few observations, these findings should be interpreted as preliminary characteristics associated with successful tries rather than as factors discriminating scoring success from unsuccessful attacking outcomes. Multi-season replication and inclusion of unsuccessful attacking sequences are needed to confirm their generalizability and practical application.