Abstract
OBJECTIVES: To quantify upper- and lower-limb power responsiveness to maximal strength (MSTG) versus plyometric training (PTG) versus control (CG), and to identify stable responder signatures from the 2D change vector (ΔUpper, ΔLower). METHODS: Twenty-seven university swimmers were randomized to MSTG, PTG, or CG (n = 9 each) for 6 weeks with testing at Pre, Mid, and Post. Upper- and lower-limb power constructs were derived via baseline-fitted PCA from bench press power plus medicine-ball throw and from CMJ, SJ, DJ, plus SLJ, respectively; responsiveness was Post-Pre. Group contrasts used permutation tests with Holm adjustment and bootstrap confidence intervals. Responder signatures were identified by Ward clustering with cluster-number selection and bootstrap stability. RESULTS: ΔUpper was 0.962 ± 0.129 (MSTG), 0.762 ± 0.218 (PTG), and 0.332 ± 0.058 (CG); MSTG-PTG mean difference was 0.200 (95% CI [0.047, 0.356], p = 0.030, g = 1.065), and both exceeded CG (p < 0.001). ΔLower was 0.822 ± 0.125 (MSTG), 0.758 ± 0.150 (PTG), and 0.388 ± 0.059 (CG); MSTG-PTG was 0.065 (p = 0.331), while both exceeded CG (p < 0.001). Clustering selected k = 2 (silhouette 0.608) with high stability (ARI 0.840 [0.591, 1.000]) and strong group association (χ(2) = 18.900, p < 0.001). CONCLUSION: In this sample, MSTG elicited larger upper-limb responsiveness than PTG, while both approaches improved upper- and lower-limb constructs versus CG, and responder signatures were stable and strongly aligned with training modality. In exploratory models within our sample, short-duration intervention, training allocation (stimulus) dominated modeled responsiveness and baseline sprint performance showed an inverse association with high-responder membership, patterns consistent with short-block trainability/ceiling effects, therefore these predictive findings should be interpreted as context-specific and not generalized beyond similar swimmer levels and intervention doses.