Purpose: Repeated-sprint training in hypoxia (RSH) leads to great improvements in anaerobic performance. However, there is no consensus about the optimal level of hypoxia that should be used during training to maximize subsequent performances. This study aimed to establish whether such an optimal altitude can be determined and whether pulse oxygen saturation during RSH is correlated with training-induced improvement in performance.Methods: Peak and mean power outputs of healthy young males [age (mean ± SD) 21.7 ± 1.4 years] were measured during a Wingate (30 s) and a repeated-sprint ability (RSA; 10 x 6-s sprint with 24-s recovery) test before and after RSH. Participants performed six cycling sessions comprising three sets of 8 x 6-s sprint with 24-s recovery in normobaric hypoxia at a simulated altitude of either 1,500 m, 2,100 m, or 3,200 m (n = 7 per group). Heart rate variability was assessed at rest and during recovery from Wingate test before and after RSH.Results: The subjective rating of perceived exertion and the relative exercise intensity during training sessions did not differ between the three groups, contrary to pulse oxygen saturation (p < 0.001 between each group). Mean and peak power outputs were significantly increased in all groups after training, except for the mean power in the RSA test for the 3200 m group. Change in mean power on RSA test (+8.1 ± 6.6%) was the only performance parameter significantly correlated with pulse oxygen saturation during hypoxic training (p < 0.05, r = 0.44). The increase in LnRMSSD during recovery from the Wingate test was enhanced after training in the 1,500 m (+22%) but not in the two other groups (≈– 6%). Moreover, the increase in resting heart rate with standing after training was negatively correlated with SpO2 (p < 0.01, r =–0.63) suggesting that hypoxemia level during training differentially altered autonomic nervous system activity.Conclusion: These data indicate that RSH performed as early as 1,500 m of altitude is effective in improving anaerobic performance in moderately trained subjects without strong association with pulse oxygen saturation monitoring during training.