Concurrent Validity and Between-Session Reproducibility of Agreement of GPath for Mean Propulsive Velocity Assessment During the Bench Press Exercise

Growing use of inertial measurement units (IMUs) in velocity-based training has outpaced the evidence supporting their accuracy, making device-specific validation an essential aspect before adoption in research or practice. This study evaluated the concurrent validity of the GPath 6-axis inertial sensor against the Chronojump linear position transducer (LPT) for mean propulsive velocity (MPV) measurement during the bench press exercise. Twelve physically active males performed repetitions at 20%, 40%, 60%, and 80% of one-repetition maximum (1RM) across two separate testing sessions. Concurrent validity was evaluated using Pearson’s correlation, concordance correlation coefficient (CCC), Bland–Altman analysis, typical error, and coefficient of variation (CV). The GPath systematically overestimated MPV relative to the LPT at all loading conditions (p < 0.001), with bias decreasing from 0.253 m/s (20% 1RM) to 0.091 m/s (80% 1RM). CV remained low and consistent across conditions (5.12–5.98%). CCC was trivial > 1 m/s (20–40% 1RM: 0.079–0.056; combined: 0.331) and moderate < 1 m/s (60–80% 1RM: 0.430–0.433; combined: 0.833), with an overall CCC of 0.877. Between-session reproducibility of agreement was high (Session 1: CCC = 0.878; Session 2: CCC = 0.877), with between-session bias differences not exceeding 0.026 m/s. Despite its systematic overestimation, the GPath showed a reproducible bias structure between sessions, which may indicate utility for longitudinal within-athlete monitoring of relative MPV changes. Although, it cannot be recommended for absolute velocity measurement or 1RM estimation without prior correction.