прогнозирование готовности к тренировке по данным носимых устройств с учётом heart rate recovery

Predicting an athlete's subjective training readiness from wearable-device data is a fast-growing field. Heart rate recovery (HRR) is the rate of heart-rate decline after exercise and a recovery indicator recognized in sports physiology that is not displayed to the user in the Fitbit app. The open PMData dataset holds 1,757 daily records from 16 recreational participants collected between November 2019 and March 2020. From the heart-rate data in this dataset, HRR60 was computed – to the best of our knowledge, for the first time for this dataset – after each of 772 training sessions (495 with a valid result), and it was tested whether it adds value to readiness prediction beyond metrics already visible to the user, using Ridge, Random Forest, and Gradient Boosting regression with Leave-One-Participant-Out validation. Neither the ablation test, nor permutation feature importance, nor direct correlation with same-day or next-day readiness showed a meaningful gain in accuracy. This result tempers overstated expectations of "hidden" wearable-device metrics and points to the need for ablation testing of such hypotheses.