Your watch gives you a number each morning and calls it recovery, or readiness, or something similarly confident. It is usually a percentage, which implies a precision the underlying measurement does not have.
The ingredients are real. The confidence of the presentation is the part worth interrogating.
Key takeaways
Recovery scores are mostly a repackaging of heart rate variability, resting heart rate and sleep, weighted by an algorithm the vendor does not publish.
HRV between healthy people varies almost as much as its own average: mean RMSSD 57 ms, standard deviation 49 ms across 840 adults (Adamu et al. 2024).
That is why every score is computed against your own baseline, and why the first weeks on a new device are meaningless.
A low score is a workload report, not a grade.
What is actually inside a recovery score?
Three signals, mostly, and no vendor publishes exactly how they are weighted.
Heart rate variability, which usually carries the most weight. The gaps between your heartbeats vary by fractions of a second, and that flexibility reflects the balance between the calming and alerting halves of your autonomic nervous system. When you have capacity to spare, the calm side keeps timing loose and variable. Under load, timing turns rigid and HRV falls. Variability is a proxy for spare capacity.
Resting heart rate, which is remarkably stable and therefore a good tripwire. The median person's moves only about 3 bpm week to week.
Sleep, usually duration and sometimes stage estimates, though stage estimates are the least reliable part of what a wrist device produces.
Some devices add skin temperature or overnight breathing rate. The output is a weighted blend, presented as one number, with the weighting undisclosed.
That is not a scandal. It does mean two things: you cannot audit why the score moved, and you cannot meaningfully compare your score to anyone else's, because it is not the same calculation.
Why is the score always relative to me?
Because absolute HRV values are almost uninterpretable across people.
In 840 healthy adults, mean RMSSD was 57 ms with a standard deviation of 49 ms (Adamu et al. 2024). The spread between healthy individuals is nearly as wide as the average. Men averaged higher than women (63.6 versus 50.9 ms), and there was no significant difference across ages from 15 to 40.
No sensible algorithm could turn a raw HRV number into a universal score, so none of them try. Every recovery score is a comparison between today and your own recent history.
Two consequences follow, and both are practical:
A new device needs a baseline period. Its first readings have nothing to compare against, so early scores are close to meaningless. Give it a few weeks before believing anything.
Switching devices resets everything. Different sampling windows, sensors and maths produce different numbers from the same body on the same night. A change in your score after switching is a change in method, not in you.
Is a low score a reason to skip training?
It is a reason to ask a question, not a reason to obey.
A low reading means your body is committing resources to something. The list is ordinary: short sleep, a drink last night, a hard session yesterday, a stressful week, a hot room, or an infection your immune system noticed before you did. We go through the causes in why your HRV tanked last night.
None of that is a grade. A low morning is your body reporting honestly, and the useful response is usually proportionate rather than dramatic: an easier session, an earlier night, more water.
Where these scores earn their keep is the pattern rather than the day. Several low days in a row with no obvious explanation is worth acting on. One low Tuesday after a late Monday is just arithmetic.
Does meditation or mindfulness move the score?
Gently, sometimes, and it is worth being honest about how gently.
A systematic review of 47 trials with 3,515 participants found mindfulness meditation programmes produced moderate evidence of improved anxiety (effect size 0.38 at 8 weeks) and depression (0.30), with low evidence of improvement in stress and distress (Goyal et al. 2014, JAMA Internal Medicine).
Notably, that same review found low evidence of no effect, or insufficient evidence of any effect, on sleep. So the common promise that a mindfulness habit will fix your sleep, and therefore your recovery score, is not well supported.
Breathwork has better short-term evidence than meditation for physiological arousal specifically. Five minutes a day of exhale-focused cyclic sighing outperformed mindfulness meditation on mood and arousal over a month (Balban et al. 2023). If your goal is to move the signals a recovery score is built from, that is the more direct route.
The honest framing: recovery signals answer to sleep, alcohol and training load first. A mindfulness habit is a supporting act, and a rough week will mask it entirely without meaning the practice failed.
How should you actually use it?
Treat it as one input with a wide error bar, not a verdict.
Read the direction, not the digit. A score of 62 against 67 is noise. Four consecutive low days is information.
Check it against your own explanation. If you know why it is low, the number has told you nothing you did not already know, which is fine.
Do not let it dictate. Feeling good on a low-score day is allowed and common. The score is an estimate assembled from proxies.
Ignore other people's scores entirely. Different algorithm, different baseline, different body.
Give a new device weeks. Its early numbers have no history to compare against.
The most useful thing about a recovery score is that it makes you notice a pattern you would otherwise have to remember. That is genuinely valuable, and it is a lower bar than the percentage implies.
Frequently asked questions
What does a recovery score actually measure?
Mostly heart rate variability, resting heart rate and sleep, blended by an undisclosed algorithm and compared against your own recent baseline. Some devices add skin temperature or breathing rate. Because the weighting is proprietary and differs by vendor, scores are not comparable between devices or between people.
Why is my recovery score low when I feel fine?
Because it is an estimate built from proxies for autonomic balance, and those proxies respond to things you may not consciously notice: a drink, a warm room, a hard session two days ago, or an immune response before symptoms. Feeling well on a low-score day is common and not a contradiction.
Is HRV the best measure of recovery?
It is the most informative single signal available from a wrist device, but its between-person spread is enormous, with a standard deviation nearly as large as the mean across 840 healthy adults (Adamu et al. 2024). It only means something against your own history, and it is one input rather than a verdict.
Can meditation improve my recovery score?
Modestly at best. Across 47 trials, mindfulness programmes showed moderate evidence for anxiety and depression but low or insufficient evidence of an effect on sleep (Goyal et al. 2014). Exhale-focused breathwork has better short-term evidence for reducing physiological arousal (Balban et al. 2023).
How long does a new device need before its scores mean anything?
A few weeks. Recovery scores are computed against your own baseline, so a device with no history has nothing to compare today against. The same applies after switching devices, since different sensors and algorithms produce different numbers from the same body.
Knit shows patterns, not diagnoses. Talk to a clinician about readings that concern you.
