Description of collective team behaviours and team performance analysis of elite rugby competition via cooperative network analysis

This study investigated cooperative passing interactions in elite rugby match play. Associations between team network metrics and match outcomes were also investigated. A cross-sectional approach was adopted, using data from four Australian Super Rugby teams, across five seasons. 44,178 passing actions were included across 321 team-fixture observations. Network metrics were calculated for each positional group within each match, and two statistical models were developed; First: a mixed-effects multinomial regression to identify differences between positional groups; and second: a mixed-effects binomial logistic regression to determine the association between team-level network metrics and match outcomes. Differences were identified between positional groups e.g. Halves had the highest out-degree centrality and betweenness, while Centres had higher eigenvector centrality than all other positions. Within the Forwards pack, the Back Row had greater in-degree, out-degree, and betweenness than the Tight Five. Regarding match outcomes, the model explained only 6.9% of variance, although greater in-degree centralisation (OR = 1.847 [1.241–2.749], p = 0.002) and lower eigenvector centralisation (OR = 0.655 [0.440–0.975]; p = 0.037) were associated with successful outcomes. Cooperative passing networks in rugby union may provide useful information to describe how various positions interact, and some behaviours may contribute towards successful team performance.