On 9 September 2026, Apple and Zwift both moved to put a recovery score between you and your next workout. Apple announced Readiness for the Apple Watch Series 12, a 0–10 score with a one-line recommendation. The same day, Zwift began adjusting its workout recommendations whenever a connected WHOOP shows a red recovery day. In the same month, Rouvy’s 4.7 update added “Today’s suggestion”, a daily workout picked from its own scores.
If you plan with a Performance Management Chart, you already have a number for this job: Training Stress Balance (TSB). So what do you do when the watch says Recover and TSB says you’re fine? This article sets out what each score is built from, according to the companies that make them, what the research says about training by heart rate variability (HRV), and how to decide when the numbers disagree.
What TSB Measures, and What It Can’t See
TSB comes from training load alone. The TrainingPeaks help centre defines it as yesterday’s Fitness (CTL) minus yesterday’s Fatigue (ATL). CTL is an exponentially weighted average of daily Training Stress Score (TSS) with a 42-day time constant; ATL uses 7 days. TSS in turn combines the duration and intensity of each session relative to your threshold. TrainingPeaks presents TSB not as a predictor of performance but as “a measure of how adapted an athlete is to their training load.”
Coach Joe Friel publishes TSB ranges that he calls guidelines, and TrainingPeaks’ coaching guide gives the same main ones: −10 to −30 during hard, focused training, which Friel calls a productive range for most athletes, with risk rising sharply below −30; and roughly +15 to +25 on the day of an A race, though Friel notes that some athletes race best at +5 to +10. Our CTL, ATL and TSB guide walks through the maths.
Because TSS is the only input, TSB knows what you did, not how your body absorbed it. A short night, a cold or a hard week at work leave TSS, and therefore TSB, unchanged.
What the Readiness Scores Are Built From
Apple Watch Readiness
Apple says Readiness analyses recent activity, training load, vitals and sleep score, and turns them into a 0–10 score. Its support pages give the bands: Recover (0–1), Pace Yourself (2–4), Ready (5–7) and Go For It (8–10). The inputs fall into three groups:
- Activity: active calories, workout effort ratings and Apple’s Training Load, which compares the intensity and duration of your workouts over the last 7 days with the previous 28 days.
- Vitals: overnight heart rate, respiratory rate, wrist temperature, HRV and blood oxygen, plus daytime heart rate, each compared with your recent baseline. Building that baseline takes at least 7 nights of sleeping with the watch in the past 49.
- Sleep: your recent sleep score.
The score updates during the day as you burn calories, nap, or your daytime heart rate changes, and the app shows which factors are driving it. Apple says the algorithm was developed with data from the Apple Heart and Movement Study, but the pages we read give no formula. Readiness runs on the Series 12 and Ultra 4. Apple calls it a general wellness feature that should not be the only basis for deciding whether it is safe to exercise.
Note the overlap with TSB. Apple says higher recent activity can lower the score, so after a big week a low reading can reflect the same load your PMC already shows. Check the factors before you change the plan.
WHOOP Recovery, as Used by Zwift
Zwift’s fall 2026 announcement, as published by Zwift Insider, says that once WHOOP is connected, Zwift automatically adjusts your Personal Recommendations on red days so you still get a useful session without overreaching. Zwift has not said how the session changes, and the announcement mentions only red days.
WHOOP’s developer documentation describes Recovery as a 0–100% score calculated each morning from measurements taken the previous day and during your sleep: resting heart rate, HRV, respiratory rate, and sleep duration and quality, plus skin temperature and blood oxygen for WHOOP 4.0 members. Green (67–100%) means ready for a strenuous day, yellow (34–66%) moderate strain, and red (0–33%) that rest is likely what the body needs; WHOOP lists overtraining, sickness, stress, lack of sleep and lifestyle factors as possible reasons. Training load is not among the listed inputs.
Rouvy’s Today’s Suggestion
Today’s suggestion, in the Rouvy Companion app, is one workout a day built from two scores: harder when you have trained consistently and recovered well, easier when you haven’t. Rouvy’s support pages define both. The Training Score is your 6-week average Rouvy Training Load (the intensity and duration of each synced activity) on a 0–100 scale. The Recovery Score, also 0–100, is fed by your rides and other synced activities. Rouvy can display HRV, resting heart rate and sleep from WHOOP, Oura and Apple Health (in beta), but says these do not feed the Recovery Score yet, and it does not disclose its formulas. So Rouvy’s suggestion is a load model, closer to TSB than to a watch score.
| Input | Apple Watch Readiness | WHOOP Recovery (used by Zwift) | Rouvy Today’s suggestion | TSB (PMC) |
|---|---|---|---|---|
| Workouts and training load | Yes: active calories, effort ratings, Training Load | Not listed | Yes: rides and synced activities | Yes: daily TSS only |
| Heart rate variability | Yes (overnight) | Yes | Shown, not used | No |
| Resting / overnight heart rate | Yes, plus daytime heart rate | Yes | Shown, not used | No |
| Sleep | Yes (sleep score) | Yes (duration and quality) | Shown, not used | No |
| Respiratory rate | Yes | Yes | No | No |
| Skin / wrist temperature | Yes | Yes (WHOOP 4.0 members) | No | No |
| Blood oxygen | Yes (not on some US models) | Yes (WHOOP 4.0 members) | No | No |
| Formula on the maker’s pages | No | No | No (Rouvy says it is not disclosed) | Yes (TrainingPeaks) |
| Output | 0–10: Recover, Pace Yourself, Ready, Go For It | 0–100%: red, yellow, green | One suggested workout | CTL minus ATL; can be negative |
| Look-back | Load: last 7 days vs previous 28; vitals baseline needs 7 of the past 49 nights | Previous day and last night’s sleep | Training Score: 6-week average | Time constants of 7 days (ATL) and 42 days (CTL) |
Load-Based vs Physiology-Based
The two families answer different questions. TSB tells you how your recent load compares with what you are adapted to. A vitals-based score tells you how your body looked overnight. Each has its own weak spots:
- HRV is non-specific. A 2026 review by Schaffarczyk and Sperlich lists illness, alcohol, menstrual cycle phase and sleep among the things that move it. In a three-month study of 39 WHOOP 4.0 wearers (Ungaro et al., 2026, funded by the Gatorade Sports Science Institute), nightly HRV averaged 54.5 ms on days participants reported drinking alcohol against 66.2 ms on days they didn’t, and it did not track self-reported stress or nervousness.
- Wrist sensors are not an ECG. The same review says pulse rate variability from wrist or ring optical sensors should not be treated as fully interchangeable with ECG-derived HRV.
- Load is only as good as the TSS behind it. TSS scores each session as a factor of your threshold, and TrainingPeaks’ coaching guide warns that incorrect zones undercut any planning built on the data.
What the Research Says About HRV-Guided Training
Readiness scores are new. HRV-guided training is not, and it has been tested against fixed plans. In those trials, each athlete’s HRV was compared with a personal normal range. When it fell below that range, the planned session was reduced, postponed, or replaced by easy exercise or recovery; when HRV was stable or rising, hard training went ahead (as summarised by Schaffarczyk and Sperlich).
One of the cycling trials in this literature is Javaloyes et al. (2019). Seventeen well-trained cyclists did four weeks of standardised training to establish a resting HRV baseline, then spent eight weeks on either HRV-guided or traditional periodization. The HRV group improved peak power output by 5.1%, power at the second ventilatory threshold by 13.9% and a 40-minute time trial by 7.3%; the traditional group did not improve significantly. The authors found no statistical difference between the groups, although a magnitude-based analysis rated the HRV group’s time trial and peak power gains likely and possibly beneficial.
Pooled across studies, the advantage is small:
- Manresa-Rocamora et al. (2021), 8 studies and 199 participants: HRV guidance improved vagal HRV, but the gains in VO2max, maximal power or speed, power at VT2 and endurance performance were small and not statistically significant. If HRV guidance beats fixed plans for fitness at group level, the authors conclude, it is only by a small margin.
- Düking et al. (2021), 8 studies and 198 participants: a significant effect on submaximal markers, which the authors call medium-sized, and small, non-significant effects on performance and VO2peak. HRV-guided training produced fewer performance non-responders, and most HRV-guided programmes included fewer moderate- or high-intensity sessions than the fixed plans.
- Granero-Gallegos et al. (2020), 6 studies and 195 participants: VO2max rose in both groups, with a larger effect in the HRV groups (0.402 against 0.215 for controls) and better results in amateur and female subgroups. Heterogeneity was high (I² = 94%).
Manresa-Rocamora et al. (2021): 8 studies, 199 participants
Düking et al. (2021): 8 studies, 198 participants
Effect size: right of zero favours HRV-guided training
Schaffarczyk and Sperlich’s 2026 review reads the evidence the same way: no clear gain in VO2max or performance, and the main benefit may be more consistent training adaptation. Most of the trials lasted eight weeks or less and had small samples. They also guided training from HRV itself. Composite scores blend HRV with sleep, other vitals and, in Apple’s case, load. The same review notes that their algorithms and weighting strategies “remain largely undisclosed,” with limited independent validation.
When They Agree, and When They Don’t
Put the two signals on a grid and four cases appear. The TSB split at −30 is Joe Friel’s guideline, which TrainingPeaks’ coaching guide also gives. Rouvy’s suggestion belongs on the TSB side, because its scores come from your activities.
Train as planned
In the HRV trials, stable or rising HRV cleared high-intensity work, and Friel calls −10 to −30 a productive range for hard training.
Soften, move or swap the session
TSB does not flag the load, so check which factors drive the score. If it is not load, do what the HRV trials did (reduce, postpone or replace the session with easy riding) and look at sleep, illness, alcohol and stress.
Do the plan, add nothing
Overnight signals look normal, but load is past the range Friel calls productive. Keep the recovery days that follow: TrainingPeaks’ coaching guide says values beyond −30 need consecutive days of rest.
Easy day or rest
Both signals point the same way: WHOOP says rest is likely what the body needs on a red day, and Friel warns that going below −30 greatly increases risk.
A Practical Routine
- Plan the week by load. Set targets with CTL and keep TSB in the range you intend; our guide to building a training plan with CTL covers the steps.
- Check readiness on hard-day mornings, and look at which factor is driving it. A low Apple score driven by training load is the PMC’s message again, not new information.
- One low reading with normal TSB: soften, move or swap the session, as the trials did. Every pooled fitness and performance estimate above favoured the HRV-guided groups, mostly by small margins, even though most HRV-guided programmes in Düking et al.’s review had fewer moderate- or high-intensity sessions.
- Several low readings in a row with normal TSB: look outside training: sleep (see why sleep is your best legal PED), illness, alcohol and stress. Apple advises stopping a workout or seeking medical attention if you have concerning symptoms, and not using Readiness in place of professional medical judgement.
- Deep negative TSB with a good score: don’t let a green morning talk you into extra work. The recovery days in the plan are there for the load the PMC can see.
So which one should decide? Neither on its own. TSB tells you whether the plan’s load is sensible; a readiness score tells you whether today is the day to carry it out. The research supports letting a low physiological reading soften a session, but it does not show that doing so makes you much fitter than following a predefined plan.
See your PMC chart in real-time
Paincave calculates CTL, ATL, and TSB automatically from your Strava data. Watch your fitness evolve with every ride.
Start freeSources
- Apple (read October 2026): Apple Newsroom, “Introducing Apple Watch Series 12, with the all-new Health Sensing System” (9 September 2026); Apple Support, “Understand your readiness score on Apple Watch” (14 October 2026); Apple Watch User Guide, “Use the Readiness app on Apple Watch” and “Track your training load on Apple Watch”.
- WHOOP and Zwift: WHOOP Developer Platform, “WHOOP 101” and “Recovery” (read October 2026); Zwift’s “This Season On Zwift” fall 2026 press release as published by Zwift Insider (Eric Schlange, 9 September 2026).
- Rouvy (read October 2026): “What’s new” (Rouvy 4.7, September 2026); Rouvy Support, “Training Insights in the Companion App” and “Multi-sport training and health metrics in Companion App”.
- Training load: TrainingPeaks Help Center, “Form (TSB)” (24 October 2022), “Fitness (CTL)” (8 October 2025) and “Fatigue (ATL)” (24 August 2026); Andrew Simmons, “A Coach’s Guide to ATL, CTL & TSB” (TrainingPeaks, updated 9 June 2026); Joe Friel, “Part 3: Training Stress Balance…So What?” (joefrieltraining.com, 12 July 2015).
- Research: Javaloyes A, Sarabia JM, Lamberts RP, Moya-Ramon M (2019), “Training Prescription Guided by Heart-Rate Variability in Cycling”, International Journal of Sports Physiology and Performance 14(1):23–32; Manresa-Rocamora A et al. (2021), “Heart Rate Variability-Guided Training for Enhancing Cardiac-Vagal Modulation, Aerobic Fitness, and Endurance Performance: A Methodological Systematic Review with Meta-Analysis”, IJERPH 18(19):10299; Düking P et al. (2021), “Monitoring and adapting endurance training on the basis of heart rate variability monitored by wearable technologies: A systematic review with meta-analysis”, Journal of Science and Medicine in Sport 24(11):1180–1192; Granero-Gallegos A et al. (2020), “HRV-Based Training for Improving VO2max in Endurance Athletes. A Systematic Review with Meta-Analysis”, IJERPH 17(21):7999; Schaffarczyk M, Sperlich B (2026), “Heart rate variability-guided endurance training: evaluating strengths, weaknesses, opportunities, and threats for load prescription and adjustment”, Frontiers in Sports and Active Living 8:1858271 (27 August 2026); Ungaro CT et al. (2026), “Disconnection Between Self-Reported Wellbeing and Heart Rate Variability from Wearables”, Sensors 26(4):1325.
This article describes how each score is built and what published research found; it is not a product comparison. No affiliate links were used.