This study investigated the reliability and accuracy of self-regulated recovery time and performance during repeated sprinting. On 4 occasions, 14 men (24.5 ± 5.0 years) completed 10 × 6 seconds cycle sprints against 7.5% body mass, self-regulating (SR) recovery time to maintain performance. Subjects then repeated the test, but with a reduced recovery (RR) of 10% less recovery time. Across the first 4 trials, there were no between-trial differences in peak power output (PPO) or mean power output (MPO), recovery time, or fatigue index (p > 0.05). Random variation in recovery time was reduced across trials 3–4 (coefficient of variation [CV] = 7.5%, 95% confidence limits [CL] = 5.4–12.4%) compared with trials 1–2 (CV = 16.0, 95% CL = 11.4–27.0%) and 2–3 (CV = 10.1%, 95% CL = 7.2–16.7%) but was consistent across trials for PPO and MPO (between-trials CV, ≤3.3%). There were no trial effects for any performance, physiological, or perceptual measures when comparing SR with RR (p > 0.05), although heart rate and perceptual measures increased with subsequent sprint efforts (p ≤ 0.05). After 2 familiarization trials, subjects can reliably self-regulate recovery time to maintain performance during repeated sprints. However, subjects overestimate the amount of recovery time required, as reducing this time by 10% had no effect on performance, perceptual, or physiological parameters. Self-regulated sprinting is potentially a reliable training tool, particularly for sprint training where maintenance of work is desired. However, overestimation of required recovery time means that performance improvements may not be achieved if the goal of training is improvement of repeated sprint performance with incomplete recovery.