Reproducibility of 20-min Time-trial Performance on a Virtual Cycling Platform

This study aimed to analyse the reproducibility of mean power output during 20-min cycling time-trials, in a remote home-based setting, using the virtual-reality cycling software, Zwift. Forty-four cyclists (11 women, 33 men; 37±8 years old, 180±8 cm, 80.1±13.2 kg) performed 3×20-min time-trials on Zwift, using their own setup. Intra-class correlation coefficient (ICC), coefficient of variation (CV) and typical error (TE) were calculated for the overall sample, split into 4 performance groups based on mean relative power output (25% quartiles) and sex. Mean ICC, TE and CV of mean power output between time-trials were 0.97 [0.95-0.98], 9.4 W [8.0-11.3 W], and 3.7% [3.2-4.5], respectively. Women and men had similar outcomes (ICC: 0.96 [0.89-0.99] vs. 0.96 [0.92-0.98]; TE: 8.3 W [6.3-13.1] vs. 9.7 W [8.2-12.2]; CV: 3.8% [2.9-6.1] vs. 3.7% [3.1-4.7], respectively), although cyclists from the first quartile showed a lower CV in comparison to the overall sample (Q1: 2.6% [1.9-4.1] vs. overall: 3.7% [3.2-4.5]). Our results indicate that power output during 20-min cycling time-trials on Zwift are reproducible and provide sports scientists, coaches and athletes, benchmark values for future interventions in a virtual-reality environment.