Where this summary comes from
Çabuk S, Ulupınar S, İnce İ, Özbay S. Can OpenCap deliver valid and reliable kinematic data for motion analysis? A systematic review and three-level meta-analysis. Biology of Sport, 2026. PMID 41783455.
The paper itself — doi.org/10.5114/biolsport.2026.154942
Study design: Systematic review and three-level meta-analysis · 12 studies · 184 participants · 1,087 error values
What the paper asked
How close does markerless motion capture from phone cameras come to laboratory systems?
What was found
- Correlation with the criterion device was good to excellent: r = 0.845. When a joint angle rises, this method sees the same rise.
- The absolute error is not small, though: RMSE 5.88°, falling to 5.20° in sensitivity analysis and 4.94° after correcting for publication bias.
- The systematic difference from the criterion device was statistically significant but practically trivial (ES = −0.140).
- Test–retest consistency ran moderate to very good, but not evenly across joints and tasks.
What does not follow from it
- In high-velocity movements and complex joint actions the variability grows markedly. The more a task resembles real sport, the less the number can be leaned on.
- 184 participants in total, and the authors themselves write that evaluation across more diverse populations and a wider range of tasks is still needed.
- An error of about five degrees means the method is built to show large differences and trends, not to adjudicate a two-degree difference.
What it means for someone here
Gait analysis here uses this family of methods, and this number states its honest limit: large differences and change over time can be trusted, while two values a few degrees apart are, in practice, the same value. That is why this clinic's report leans on trend and range rather than on the last digit.
Related findings
This page summarises one study. It is not treatment advice, and no programme is written for a person from a single paper.