Abstract

The study aim was to compare different predictive models in one repetition maximum(1RM) estimation from load-velocity profile (LVP) data. Fourteen strength-trained men under-went initial 1RMs in the free-weight back squat, followed by two LVPs, over three sessions. Pro-files were constructed via a combined method (jump squat (0 load, 30–60% 1RM) + back squat(70–100% 1RM)) or back squat only (0 load, 30–100% 1RM) in 10% increments. Quadratic and linear regression modelling was applied to the data to estimate 80% 1RM (kg) using 80% 1RM mean velocity identified in LVP one as the reference point, with load (kg), then extrapolated to predict 1RM. The1RM prediction was based on LVP two data and analyzed via analysis of variance, effect size (g/η2p),Pearson correlation coefficients (r), paired t-tests, standard error of the estimate (SEE), and limits of agreement (LOA).p< 0.05. All models reported systematic bias < 10 kg, r> 0.97, and SEE < 5 kg, however, all linear models were significantly different from measured 1RM (p= 0.015 <0.001). Significant differences were observed between quadratic and linear models for combined (p< 0.001;η2p= 0.90) and back squat (p= 0.004,η2p= 0.35) methods. Significant differences were observed between exercises when applying linear modelling (p< 0.001,η2p= 0.67–0.80), but not quadratic(p= 0.632–0.929,η2p= 0.001–0.18). Quadratic modelling employing the combined method rendered the greatest predictive validity. Practitioners should therefore utilize this method when looking to predict daily 1RMs as a means of load autoregulation.