Validity and within-session reliability of the Exerfly system: A comparison of motorised and non-motorised flywheel squats

Abstract

This study examined the concurrent validity and within-session reliability of the Exerfly flywheel application. Twelve resistance-trained men completed repeated flywheel back-squat sessions across six inertial loads (0.025–0.15 kg·m 2 ) using motorised and non-motorised modes. Mean and peak concentric and eccentric force and velocity were recorded simultaneously using the Exerfly rotary encoder system and a MuscleLab unit. Concurrent validity was assessed using Bland-Altman limits of agreement (LoA) and concordance correlation coefficients (CCC) while accounting for repeated measures. Proportional bias was examined where present. Within-session inter-set reliability was evaluated using ICC(2,1), CV, and SEM. Mean concentric velocity demonstrated the highest agreement between systems, with small, non-significant bias and very large concordance in both non-motorised (0.04 m/s [95% CI: −0.01, 0.08], CCC = 0.80) and motorised (0.01 m/s [95% CI: −0.02, 0.04], CCC = 0.86) conditions. Motorised mean eccentric force also showed trivial bias and nearly perfect concordance (−5 N [95% CI: −23, 13], CCC = 0.94), although with wide LoA. In contrast, mean eccentric velocity and all peak variables exhibited systematic bias and lower concordance, with proportional bias evident particularly at higher forces and velocities. Despite this, both systems demonstrated good-to-excellent inter-set reliability across variables (ICC range: 0.64–0.99; CV%: 1.13–10.16). Agreement between systems was limited for interchangeable use across most mechanical outputs. Mean concentric velocity and motorised mean eccentric force showed the most acceptable agreement, whereas the remaining force and velocity measures demonstrated substantial bias. Nonetheless, both systems exhibited strong reliability, supporting their independent use for monitoring flywheel training performance. Protocol registration: The original protocol was prospectively registered at the Open Science Framework ( https://osf.io/acpvs ).