ABSTRACT Resisted sprint training (RST) is effective for improving sprint acceleration. However, the choice of RST loads−ranging from light (≤ 10% of unresisted sprint maximum velocity decrement v dec , 10%–30% v dec ) and very heavy RST were more effective than UST, whereas light and heavy loads (> 50%–80% BM or > 30%–50% v dec ) were not. Two‐ and multiple‐comparison models indicated that very heavy RST loads produced significantly greater performance improvements than light loads. The effect of RST on performance compared to UST became more pronounced with an increasing load ( β = −0.005, p = 0.018), with a significant threshold at 20.7% BM, beyond which RST led to significantly greater performance improvements than UST ( p < 0.05). Very heavy RST loads were particularly effective in improving F 0 , P max , and RF max compared to UST, with adaptations in F 0 increasing significantly in tandem with RST ( β = 0.010, p = 0.023). These effects were more pronounced in trained and highly trained individuals, with negligible improvements in recreationally active participants. There was an inverted U‐shaped relationship between training volume (total distance) and training effect: heavy loads at small volumes yielded similar benefits to light loads at large volumes. Finally, participant characteristics (age, sex) and training parameters (duration, frequency, number of training sessions, sprint phase, session volume) also influenced sprint performance improvements. Very heavy loads were generally more effective at increasing sprint performance than UST and light RST loads, with light RST showing comparable effects to UST. These findings were likely attributable to improvements in the early acceleration phase (especially 0–5 m). Heavier loads were optimal for trained and highly trained groups, but no clear benefit of higher loads for recreational groups, suggesting light RST or UST are probably sufficient. Coaches and practitioners can refine protocols based on these findings to boost sprint performance outcomes. Trial Registration The original protocol was prospectively registered (osf.io/fu82d) with the Open Science Framework
Xu et al. (2025) studied this question.