How Is a Readiness Score Calculated? Oura Uses Nine Inputs, Apple Uses Three, and None of Them Publish the Formula

A user wakes up to three numbers that claim to answer the same question.

WHOOP publishes a member-average Recovery of about 58%, which sits in yellow (34–66%) [3][5]. An Oura 72 is Good (70–84) [1]. Apple has not published which integers on its 0–10 map onto Recover, Pace Yourself, Ready or Go For It — only the four labels [2]. Garmin Training Readiness 62, constructed here as an example, is Moderate (50–74) on a five-band scale, and it is not a synonym for Oura’s 72; Body Battery on the same wrist is a different object entirely [4].

None of those figures is a measurement of readiness in the way resting heart rate is a measurement of heart rate. Each is a composite: several signals, a proprietary weighting, a branded scale, and a sentence the marketing team liked. A 2025 review of fourteen such scores across ten manufacturers found that not one disclosed the formula, and that few had tested the composite against any external reference [1].

That is not an argument against showing a daily capacity number. It is the reason you cannot treat 72, 58% and an unpublished 0–10 as translations of each other. This post is the vendor map — what each score is actually built from, where the scales disagree, one reconstructed morning where the inputs are shared and the composites are not, and what a product should do instead of picking a favourite and hoping.

What are these scores actually made of?

Short answer: mostly HRV, resting heart rate, sleep and yesterday’s activity, combined in ways nobody will show you. Doherty, Baldwin, Lambe, Burke and Altini catalogued fourteen composite health scores in Translational Exercise Biomedicine in April 2025. HRV appeared in 86%, resting heart rate in 79%, physical activity in 71%, sleep duration in 71%. Sleep architecture showed up in 57%, temperature in 29%, respiratory rate in 14% [1].

So the shopping list is shared. The recipe is not.

Two rows in that table are the ones products mix up. Garmin ships both Body Battery and Training Readiness. Apple’s new score is 0–10, not 0–100, and it does not run on last year’s Watch [2]. Polar never used 0–100 at all. The English word is readiness. The objects are different.

Why the same morning produces different numbers

The window is different

Even when two scores both “use HRV,” they do not average the same minutes. The 2025 review set the windows out explicitly [1]:

• Fitbit Daily Readiness: HRV from the longest sleep period over three hours
• Garmin Training Readiness: five-minute HRV windows averaged across the night
• Oura: mean of all five-minute samples across the whole sleep period
• Polar Nightly Recharge: a four-hour window starting 30 minutes after sleep onset
• WHOOP: HRV weighted toward the last slow-wave-sleep period

We have already shown, in what a good HRV even is, that high-frequency HRV is more than twice as high in slow-wave sleep as in REM. A score that leans on the last deep-sleep period will read higher than a score that averages the whole night, from the same heart, before anyone applies a weight. That is one documented reason an Oura number and a WHOOP number can diverge. It is not a measured share of any given morning’s gap, and it is not a bug in either product.

The scale is different

Rescaling Apple’s 7 to 70 because “both are 70% of max” is a category error. Apple’s 0–10 is a four-bucket recommendation system with a numeric handle, and Apple has not said which integers sit in which bucket [2]. Oura’s 70 sits at the bottom of Good, one point above Fair [1]. WHOOP’s 70% is just inside green (67–100%), a statement about strain capacity, not a percentile of sleep quality [5]. Polar’s +3 is not 65 on a 0–100 line. There is no honest linear map.

The question is different

Oura is asking how recovered you are relative to your baselines, including two months of activity balance [1]. WHOOP is asking how much strain you can absorb today [5]. Apple is asking how to take the day, and it will change the answer after an afternoon workout [2]. Garmin Body Battery is asking how much charge is left in the tank. Those are neighbouring questions. They are not one question with four UIs.

One morning, five scores

The formulas are unpublished, so nobody outside the vendor can emit the composite. What we can do is hold the physiology still and say which listed inputs move.

Take a constructed morning, labeled as such: 7 hours asleep after a hard evening session, sleep onset two hours later than the prior fortnight, overnight HRV below the person’s usual range, resting heart rate a few beats up. Body Battery has been draining all evening.

The useful output of that table is not a fake 61 / 7 / 54%. It is that the same night is a different subset of the same body. A product that averages the composites is averaging four different questions. A product that shows the factors that moved is showing the night.

Nobody will show you the weights

The review’s blunt finding is the one that should sit on every product spec: no manufacturer discloses the algorithm or the relative weights [1]. Polar will say the autonomic-nervous-system piece is weighted above sleep. Fitbit will say HRV and sleep matter more than resting heart rate. That is the ceiling of public detail. Everything else is a contributor list, which is not a formula.

Citing a study of the heart-rate sensor, or of sleep-stage detection, is evidence for a component, not validation of the composite. Apple’s September announcement said the readiness algorithm was developed with data from the Apple Heart and Movement Study [2]. That is a serious construction claim. It is still not an independent test of whether a 6 predicts anything a 9 does not.

Apple just made this a platform default

Until this month, readiness was a specialist product: Oura, WHOOP, Garmin, Polar, Fitbit, Samsung’s Energy Score. On 9 September 2026 Apple shipped the category onto the newest Watch — 0–10, four verbs, Series 12 and Ultra 4, watchOS 27 on 14 September — and kept it off HealthKit [2][6]. We covered the strategic half in the Series 12 event piece: the score is real, gated, unpaid, and not yours to query.

The product half is this. Users will now expect a daily capacity number with a plain-language verb on hardware that did not have one before Series 12. Matching Apple’s 0–10 on Apple-only, for the people who just bought a Series 12, is not a strategy. Showing Oura’s 0–100 next to Apple’s 0–10 as if 70 ≈ 7 is worse. The category is now default. The construct is still vendor-private.

What a product should do with this

Do not compare composites across brands. If a user pairs an Apple Watch and an Oura Ring, you will get two readiness-like numbers for one body. Averaging them is how you manufacture a third fiction. Pick a primary source per user, or compute one score from the underlying samples, the same way you already have to deduplicate overlapping streams.

Show the factors, not just the grade. A 58% that is sleep-debt and a 58% that is a suppressed HRV after a late interval session are different product moments. The vendors that name contributors (Oura, Apple’s “what’s driving the score”) are doing the useful part. The grade is a headline. The headline is not the feature.

Do not wait for a readiness API. Apple has not announced one [2]. HealthKit still measures. If you need the same daily capacity signal on iOS, Android, and the phone-only majority, you compute it. How Sahha does that — 0–100, eight factors, four of them from the phone alone — is the Readiness Score guide, not this page.

A readiness score is a useful object. It is also a branded compression of several other objects, on a scale the vendor chose, with weights the vendor will not share. Treat it that way and it helps. Treat 72, 58% and an unpublished 0–10 as the same fact and you have built a feature on a translation that does not exist.