À retenir
In 7 male competitive cyclists tracked for sixteen weeks (590 athlete-days), adding total work in kilojoules and session RPE to past HRV gave only modest reductions in prediction error. With so few athletes and big differences between them, the models are not a basis for forecasting your own HRV.
Study design
Article de recherche
Reading basis
Published abstract
Reading focus
Design & applicability
Cette explication éditoriale existe en anglais, chinois simplifié et chinois traditionnel. Cette interface affiche la lecture en anglais.
Question de recherche
Many athletes judge recovery from day-to-day heart rate variability (HRV), but it is unclear how well tomorrow's HRV can be forecast from training data people already collect. The authors tested whether machine learning models can predict next-day HRV in competitive cyclists using external load (total mechanical work in kilojoules, kJ) and internal load (session rating of perceived exertion, RPE).
Population et méthode
Seven male competitive endurance cyclists were monitored daily for sixteen weeks, giving 590 athlete-days (seven independent time series). Support vector regression (SVR) and extreme gradient boosting (XGBoost) were compared with a conventional autoregressive model with exogenous inputs (ARX). Each model was trained per athlete under two scenarios (past HRV only, or past HRV plus kJ and RPE) and across lag orders of 1, 4, 7, 10 and 14 days. Accuracy was expressed as root mean squared error (RMSE). The abstract does not state which HRV metric was forecast or give RMSE values.
Résultats
The authors’ reported results. Read each estimate with its population and uncertainty.
- 1Across all athletes, adding kJ and RPE to past HRV produced only modest reductions in RMSE compared with HRV-only models.
- 2XGBoost had the lowest one-step-ahead RMSE at short lag, while all models converged at longer lag orders.
- 3Prediction accuracy differed markedly between athletes, in line with the individual nature of autonomic responses.
Limites et incertitude
- 1Only 7 male cyclists were studied, so results may not apply to women, other sports or recreational athletes.
- 2The abstract gives no RMSE values, so the size of the improvement from adding load cannot be judged, and models were fitted separately per athlete.
- 3Only kJ and RPE were tested; the authors say broader contextual variables are probably needed, and this reading is based on the abstract only.
Mise en perspective
Total work and perceived effort tell you little extra about tomorrow's HRV beyond your recent HRV history. Sleep, stress, illness and other context probably matter too. Treat HRV as one input to how you feel, not something a training log can predict for you.
Vos données
If you log training load (kJ from a power meter, session RPE) and morning HRV, compare your HRV against your own recent range under the same conditions rather than expecting load numbers to explain it. Expect large person-to-person differences, so another athlete's pattern is a poor guide for yours.
Comprendre ce type de preuve
The index does not establish a precise design. Read the methods before making causal claims; check population, comparator, outcomes and uncertainty in the source.
Résumé structuré des auteurs English
Le résumé original en anglais est reproduit avec attribution sous la licence ouverte indiquée. Il s’agit d’une source, pas d’une interprétation MoveVitals. L’analyse préparée figure ci-dessus ; vérifiez les conclusions importantes dans la source.
Introduction
Day-to-day fluctuations in heart rate variability (HRV) are widely used to infer autonomic recovery in endurance athletes. However, the extent to which HRV can be forecast one day ahead from readily available external and internal training-load metrics remains unclear. In this study, we evaluated whether machine learning models can predict next-day HRV in competitive cyclists using the two load descriptors most commonly collected in practice: external load quantified as total mechanical work in kilojoules (kJ) and internal load quantified as session rating of perceived exertion (RPE).
Methods
Seven male competitive endurance cyclists were monitored daily for sixteen weeks, yielding 590 athlete-days of longitudinal data (seven independent time series). Two machine learning approaches-support vector regression (SVR) and extreme gradient boosting (XGBoost)-were compared with a conventional autoregressive model with exogenous inputs (ARX) as a traditional time-series benchmark. Each model was trained individually per athlete under two predictor scenarios (using past HRV-only or past HRV plus kJ and RPE) and across multiple lag orders (1, 4, 7, 10 and 14 days), with forecasting accuracy expressed as root mean squared error (RMSE).
Results
Across all athletes, adding kJ and RPE to the past HRV produced only modest reductions in RMSE relative to HRV-only models. XGBoost achieved the lowest one-step-ahead RMSE at short lag, while all models converged at longer lag orders. Predictive accuracy differed markedly between athletes, reflecting the well-known individual nature of autonomic responses.
Conclusions
These findings suggest that the two routinely collected load descriptors examined here-total work (kJ) and RPE-add limited information beyond recent HRV history for forecasting next-day HRV, and that broader contextual variables are likely required to meaningfully improve athlete monitoring.
Sources et attribution
Predicting Next Day Heart Rate Variability Based on Training Load in Cyclists Using Machine Learning.
Barsumyan A, Saukkonen A, Soost C, Graw JA, Burchard R.
Sports (Basel, Switzerland) · 2026-06-30
DOI: 10.3390/sports14070271 ↗PubMed · 42506814 ↗Europe PMC · Licence indiquée: cc by ↗Abstract © the original authors, reproduced under the recorded open licence. Translations support reading; use the English record to check terminology and numerical results. Our methodological interpretation is separate from the authors’ conclusions.
Interprétation éditoriale MoveVitals · contenu éducatif · aucune évaluation clinique individuelle enregistrée
Sources vérifiées 2026-11-01
Guides pratiques
À lire ensemble
Short-term HRV was higher during slow breathing in 962 military men, injured or not
In 962 military men (median age 37.6 years), every short-term heart rate variability (HRV) measure was higher during slow breathing at six breaths a minute than during spontaneous breathing, injured or not; in fast resting breathers RMSSD rose from 15.57 to 21.89 ms. It was one session in a fixed order, so it shows an immediate effect, not lasting benefit.
European Journal of Applied Physiology
Cold-water immersion: no strength gain, briefly weaker jumps, soreness relief uncertain
Across 22 randomised trials (375 participants, mostly men), one cold-water immersion after exercise did not change maximal isometric strength (g 0.08, 95% CI −0.11 to 0.26) and was followed by weaker jumping straight afterwards (g −0.68), gone by 24–48 hours. Lower soreness and creatine kinase looked favourable but weakened in sensitivity checks, and certainty was low to very low.
PeerJ
Cold-water immersion after hard exercise: less soreness, better power, no strength gain
In a meta-analysis of 52 randomised trials, cold-water immersion after high-intensity exercise was followed 24 hours later by lower soreness (SMD −0.89), higher perceived recovery (0.66) and better muscular power (0.22), with no effect on strength. Participants could not be blinded, most were men, and many pooled results rest on few studies with high heterogeneity.
Sports Medicine
Acute:chronic workload ratio: no evidence it can reduce injury risk, authors argue
The authors argue there is no evidence for using the acute:chronic workload ratio (ACWR) in training-load management or to reduce injury risk: no study has properly estimated a causal effect, and the ratio is statistically inaccurate and ambiguous. This is a methodological critique, not a new data analysis, so it does not show that training load is unrelated to injury.
International Journal of Sports Physiology and Performance

