Author: Jenny Murphy et al (2026)
Journal: Sports Medicine

AI generated summary

This study investigated how replicable findings are within sports and exercise science, a field increasingly concerned with research reliability. Using a large-scale statistical approach, the authors estimated the proportion of published findings that would successfully replicate if the same studies were repeated under similar conditions.

Rather than conducting direct replication experiments, the study applied modern statistical modeling techniques to reported effect sizes and p-values from a broad sample of sports and exercise science papers. This approach allowed the authors to estimate replicability across the literature while accounting for publication bias and selective reporting.

The results suggest that replicability in sports and exercise science is moderate but lower than ideal. A substantial proportion of reported effects are likely overestimated, and a meaningful number may fail to replicate altogether. Smaller sample sizes, flexible analytical choices, and an overreliance on statistical significance were identified as key contributors to reduced replicability. Studies reporting larger effect sizes and stronger statistical evidence were more likely to be replicable.

The authors conclude that while many findings in sports and exercise science are likely genuine, the field would benefit from improved research practices. Recommended solutions include larger sample sizes, preregistration of study protocols, transparent reporting, data sharing, and greater emphasis on estimation and uncertainty rather than sole reliance on p-values. Strengthening these practices is essential to improving confidence in research used to inform training, performance, and health-related decision-making.