I just wrapped a six-week pilot with a Championship club where a CatBoost injury-risk model (AUC 0.78) fed a live minutes-cap visualization in Power BI, and staff used it to hold two players under 65 minutes without pushback. Is anyone else seeing executive sign-off on model-driven availability this season, and how are you quantifying and communicating uncertainty bands to coaches on matchday?
We got sign-off at our MLS club once we stopped showing a single risk score and framed it as “cap 55–65’ (80% PI) + expected days lost avoided vs baseline,” with a simple traffic-light band so coaches didn’t feel it was a black box. Caveat: we recalibrate to actual exposures every Monday and keep a documented override for tactics — treat it like a speed limit, not a brick wall.
Added SHAP-driven ‘why now’ card + ±8’ band in Power BI; kept 10’ discretionary buffer…