Author(s): CHIARELLO, F., Institution: UCAM, Country: ITALY, Abstract-ID: 473

INTRODUCTION:
Performance analysis is widely used in team sports [1],[2],[3],[4].
With the end goal being winning, understanding which KPI’s are more useful to comprehend the differences between winning and losing efforts should be the main concerns for practitioners [5],[6],[7],[8],[9]. The KPI’s selection usually is limited to a standardize group of few variables, for example, in basketball stats as rebounds or field goal made [9], or in rugby union indicators as carries or lineout won are commonly used [5],[6]. The KPI’s selection [7] should be prioritized in these types of works. The goal is to establish a workflow that could help with data selection while analyzing opposition.
METHODS:
1073 professional rugby games from the French first division (TOP14), the English Premiership and The United Rugby Championship (URC) from the 2023-2024, 2024-2025, and the 2025-2026 seasons (up to 3/1/2026) were selected. From this 1073 games, 1053 were considered for the analysis, the games that ended in a draw were excluded. 42 Professional rugby teams took part in these games. For the analysis were included 110 KPI’s, 22 of which were the performance indicators commonly used in the scientific literature [5], [10]. Two databases were created; both contained data from the 1053 games. The first one included all game data in absolute form (DBGa) and the second one included game data represented as percentiles of the of the maximum occurrence for that variable (DBGp). A Student’s T-Test for independent samples was carried out for each database, with the KPI’s grouped for result (W-L), with the assumption that there were significant differences for the result. Then for the Percentile Database a MANCOVA was carried out, to evaluate if variables between wins and losses varies significantly among teams. 108 KPI’s were used as variables, team and result as factors and Points Scored and Point Against as Covariate.
RESULTS:
The two databases showed respectively 55,5%(DBGa) and 54,5%(DBGp) of the 110 KPI’s with significant differences between winning and losing outcomes, p-value <.001. Out of the 22 standard KPI’s alone 17(77%) showed p-value < .001. The MANCOVA test showed that were not significant differences in the variables between win and losses among teams, Lambda-Wilks p = 0.458.
CONCLUSION:
Incrementing the number ok KPI’s considered can help with a better understanding of winning and losing efforts. The lack of significant differences in the variables between wins and losses among teams lead to the assumption that the KPI’s that showed significant relevance for game outcome can be used for every team. Being percentile a reliable way to represent data the percentile of the difference between percentile for winning and losing effort can be used to underline for a particular team the importance of a variable in determining winning and losing. [1] Hughes 2002/ [2] Colomer 2020/ [3] Hughes 2004/ [4] McLellan 2013/ [5] Ungureanu 2019/ [6] Scott 2023/ [7] Martìn-Castellanos 2024/ [8] Ortega 2009/ [9] Chen 2025.