Recovery Score Signals Repeat at Rest, Not After Workouts
Repeat measures of HRV agreed at rest and fell apart after a 30-minute ride. No study found here tested a named recovery score day to day.
The signal behind the score repeats well at rest and poorly after exercise. In 18 healthy adults who did a 30-minute cycling session twice, resting HRV indices reproduced with an intraclass correlation of 0.81 to 0.86. Measured 30 to 60 minutes after exercise, the range fell to -0.06 to 0.60. The wearables' own HRV error can also approach the size of a change worth noticing. No study found here tested the day-to-day reliability of a named recovery score.
Does the HRV behind a recovery score repeat from day to day?
At rest, mostly yes. After exercise, no. Eighteen healthy adults (mean age 27, eight women) did a 30-minute cycling session at 60% of peak workload twice, plus a control session. Heart rate variability and blood pressure were measured at rest and 30 to 60 minutes afterward.
Resting HRV indices had moderate to excellent reproducibility (ICC 0.81 to 0.86). After exercise it ranged from -0.06 to 0.60, which the authors call poor. Blood pressure was excellent both times (ICC 0.93).
The exercise was the same in both sessions. The measure still shifted, so a post-exercise HRV reading carries more noise than a resting one.
1 study
- In 18 healthy adults (age 27, eight women) doing a 30-minute cycling session twice, HRV indices had ICC of 0.81 to 0.86 at rest and -0.06 to 0.60 30 to 60 minutes after exercise. Systolic blood pressure had ICC of 0.93 both before and after.
Is the wearable's own HRV number accurate enough?
For heart rate, yes. For HRV, the error is close to the size of the changes being tracked. A WHOOP validation over 15 opportunities found heart rate bias under 0.39% against ECG. For ln RMSSD, bias was 1.66% and limits of agreement were plus or minus 5.93%. That approached or exceeded the smallest worthwhile change (1.5% to 6.5%) and the natural day-to-day variation (3% to 13%).
The Oura Ring was tested against a chest ECG in 35 healthy people overnight at home. Mean biases were low. Heart rate, RMSSD, average beat interval and pNN50 correlated highly with ECG. SDNN and high-frequency power correlated moderately, and low-frequency power and the LF:HF ratio correlated poorly.
The WHOOP paper discloses a conflict of interest. A co-author's position is sponsored by WHOOP Inc. Sleep-tracking accuracy is covered in how accurate Fitbit, Garmin and Whoop sleep tracking is.
2 studies
- In a WHOOP versus ECG validation over 15 opportunities, heart rate bias was 0.39% or less. For ln RMSSD, bias was 1.66% (SD 1.80) with limits of agreement of plus or minus 5.93%, which approached or exceeded the smallest worthwhile change (1.5% to 6.5%) and coefficient of variation (3% to 13%).
- In 35 healthy people overnight at home, the Oura Ring had low mean bias against a chest ECG. Heart rate, RMSSD, AVNN and pNN50 correlated highly, SDNN and HF moderately, and LF and LF:HF ratio poorly.
Does a recovery score track anything real?
It tracks performance in one large dataset. A study of 389 professional golfers across 521 events covered 35,140 nights of wearable data. Its composite recovery score, integrating sleep and biometrics, averaged 59.1% (SD 9.9%).
Golfers with higher recovery had better scores. Between golfers, each 10-point higher recovery went with a golf score 0.476 lower. Within a golfer, a 10-point improvement went with 0.238 lower. Lower is better in golf.
It is an association in one sport. It does not show that raising the score improves play. The composite's formula is not described in the abstract.
1 study
- In 389 professional golfers over 35,140 nights, a 10-percentage-point higher composite recovery score was associated with a 0.476 lower golf score between golfers and 0.238 lower within golfers.
What can a single day's number tell you?
Less than a pattern. A study of 25 experienced athletes found many illnesses occurred when an individual exceeded a personal training threshold. The thresholds were mostly related to training strain, calculated over rolling six-week averages.
A survey of 41 professional football clubs found they monitor players with questionnaires (68%) and submaximal exercise tests (41%) alongside objective data. Clubs reported a gap of 23% between expected and actual effectiveness for injury prevention. Concerns over reliability of assessment tools were one cause.
See recovery score versus HRV and why recovery scores drop. For the scores' inputs, see recovery score versus readiness score.
What does this not settle?
No source here measures the day-to-day reliability of a named consumer recovery score. The reproducibility study covers standardized exercise in 18 healthy adults, not a wearable's scoring window.
The golf study links a composite score to performance. It does not test whether day-to-day movements in the score are noise.
The reproducibility finding comes from 18 healthy adults aged about 27 doing one standardized 30-minute cycling session, with HRV measured by research methods. It does not test any wearable recovery score, other exercise types, or people who are older or less fit.
Common questions
How reliable is HRV from day to day?
At rest, HRV indices were reproducible in 18 healthy adults (ICC 0.81 to 0.86). Measured 30 to 60 minutes after exercise, reproducibility was poor (ICC -0.06 to 0.60).
Why does my recovery score jump around?
One reason is that HRV taken after exercise reproduces poorly. Wearable HRV error is also close to the size of a meaningful change, at plus or minus 5.93% for WHOOP ln RMSSD against a smallest worthwhile change of 1.5% to 6.5%.
Is wearable heart rate more reliable than wearable HRV?
Yes in the validations reviewed. WHOOP heart rate bias was under 0.39% against ECG, while RMSSD limits of agreement were plus or minus 5.93%.
Does a recovery score predict performance?
In 389 golfers over 35,140 nights, a 10-point higher composite recovery score went with a 0.476 lower golf score between golfers and 0.238 lower within a golfer. It is an association.
Has anyone tested day-to-day reliability of a recovery score?
Not in the sources reviewed here. They test the HRV signal, its measurement accuracy, and links to performance.
KnowYourPrime Evidence Registry: Finding KYP-ER-0005: Readiness and Recovery Score Validity · Finding KYP-ER-0024: Oura Ring HRV Accuracy Against ECG
Sources
- Reproducibility of autonomic cardiovascular function and hemodynamics at rest and during recovery from exercise
- Wrist-Based Photoplethysmography Assessment of Heart Rate and Heart Rate Variability: Validation of WHOOP. [conflict statement: an author's post is sponsored by WHOOP, from after data collection]
- Accuracy Assessment of Oura Ring Nocturnal Heart Rate and Heart Rate Variability in Comparison With Electrocardiography in Time and Frequency Domains: Comprehensive Analysis. [acknowledgements: Oura gave the authors access to the data]
- Wearable-Derived Sleep and Physiological Metrics Are Associated With Performance in Professional Golfers. [author affiliations include Whoop]
- Monitoring training in athletes with reference to overtraining syndrome.
- Training Load and Player Monitoring in High-Level Football: Current Practice and Perceptions.