Deconstructing agility in elite handball: PLSR-based modeling of speed, strength, and body composition on change-of-direction and reactive agility

Abstract

This study aimed to identify multidimensional physical correlates of change-of-direction speed (CODS) and reactive agility (RA) in elite handball players and to examine potential sex-related association patterns using partial least squares regression (PLSR). Forty-eight National Level I (or above) handball athletes (18 males, 30 females) were tested during the preseason. Athletes completed assessments of sprint performance (10–30 m), isometric mid-thigh pull (IMTP; peak force and 0-100ms rate of force development), vertical jump performance (CMJ, SJ, DJ, and reactive strength index), and body composition. Testing was completed across two standardized sessions separated by 48 h; CODS (T-test) and RA (Y-agility) were assessed in the second session. Sex-stratified PLSR models were fitted, predictor importance was quantified using variable importance in projection (VIP), and model performance was summarized using R 2 Y and Q 2 Y. The one-component sex-stratified PLSR models showed acceptable explanatory and predictive performance (female: R 2 Y = 0.604, Q 2 Y = 0.559; male: R 2 Y = 0.631, Q 2 Y = 0.531). Sprint ability, particularly 30 m performance, and RSI showed the highest overall importance (VIP ≈ 1.2–1.3) in both models, whereas higher body fat percentage (BFP) tended to relate to longer CODS and RA completion times (i.e., poorer performance). Beyond these shared predictors, height exceeded the VIP threshold in males (VIP = 1.26), whereas BFP exceeded the threshold in females (VIP = 1.18), reflecting modest between-model nuances in secondary predictors, while the top predictors were largely consistent across sexes. Agility outcomes in elite handball players appear to be associated with both neuromuscular and morphological factors in this preseason, cross-sectional dataset. Across sexes, sprint performance and reactive strength emerged as prominent multivariate correlates of CODS and RA, while higher BFP tended to relate to longer completion times. These findings support a multifactorial view of agility and provide preliminary, model-based evidence for future hypothesis-driven research. Sex-stratified patterns should be interpreted as exploratory given the smaller male subgroup and the constrained RA task.