3× fewer
ACL injuries.
NC State Statistics. 989,405 player-exposures. U15–U19 girls. The SoccerPulse cohort recorded one-third the ACL injury rate of the national HS Girls' Soccer benchmark.
Three rates. One conclusion.
3.94
ACL injuries per 100,000 exposures — SoccerPulse cohort. 8 injuries across 203,112 player-exposures.
12.2
ACL injuries per 100,000 exposures — national HS Girls' Soccer benchmark. 96 injuries across 786,293 player-exposures.
3×
Posterior median reduction factor in favour of the SoccerPulse cohort. 95% credible interval: 1.5× to 6.75×.
What the study measured
Dr. Sujit K. Ghosh, Professor in the Department of Statistics at NC State University, was asked to assess whether ACL injury rates among U15–U19 girls using SoccerPulse differed from the published national benchmark.
Cohort one: 203,112 player-exposures across three coaches at one club, all using the SoccerPulse daily wellness, RPE, and injury-reporting workflow. 8 ACL injuries. Cohort two: 786,293 exposures of Girls' HS Soccer from Hootman et al. (NIH PMC3867093). 96 ACL injuries.
Because both rates are very small in absolute terms, classical asymptotic tests would understate the evidence in the data. Dr. Ghosh used Bayesian methods to recover the full posterior distributions — letting the data speak for itself.
How the analysis was done
Bayesian
Two-sample Beta-Binomial comparison with a Jeffrey's prior (a = b = 0.5). 100,000 posterior samples per cohort recover the full distributions of the rate difference, log ratio, and log odds ratio.
Stochastically smaller
The posterior CDF of the SoccerPulse rate lies entirely to the left of the benchmark CDF — every quantile of p₁ is below the corresponding quantile of p₂.
Conclusive
Posterior probability that the SoccerPulse rate equals or exceeds the benchmark: ≈ 0.018% (≈ 1 in 5,500). Far stronger than any conventional p < 0.05 threshold.
Where the reduction lives
3.03×
Posterior median reduction. log(p₁/p₂) = -1.11. The most likely single answer.
6.75×
Upper bound of the 95% credible interval. log ratio = -1.91. The optimistic tail.
1.5×
Lower bound of the 95% credible interval. log ratio = -0.46. Even the conservative tail is a real reduction.
Dr. Sujit K. Ghosh
Professor, Department of Statistics, NC State University. Analysis performed independently in R using a standard two-sample Beta-Binomial Bayesian comparison. The R script, posterior diagnostics, and exact quantile output are reproduced in the appendix of the downloadable PDF.
Control-group data: Hootman et al., published at https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3867093/.