Recovery score on Apple Watch: what Apple offers and how to read it

Apple's pages do not describe a ‘recovery score’. What exists is the Readiness app, only on Apple Watch Series 12 and Ultra 4 and for ages 13 and older (Apple Watch User Guide, Readiness). The rest with recovery in the name measure something else.

What Apple offers and what it does not

Sources: (Apple Support, Vitals), (Apple Watch User Guide, Training load), (Apple, 2024) and (Apple Support, Heart rate). Apple sorts the Readiness score into four bands (Apple Support, Readiness):

How to get the score

With one of those two models, Apple asks for this before it gives a score:

1. Enter your height, weight and age in the Fitness app on iPhone.
2. Set up Activity.
3. Set up Sleep and wear the watch to bed for at least 7 nights.
4. Set up overnight vitals.

Then open the Readiness app on the watch and tap Activity, Vitals or Sleep for detail. On iPhone it is in the Fitness app, sometimes at the bottom of the screen. If it does not appear one day, check you slept with the watch on. The steps may change with your watchOS and iOS versions.

The score updates during the day with calories burned, naps and daytime heart rate, and it can go down after a workout. On any other model, Apple does not offer this score: you will have to read vitals, load and sleep separately.

Cardio recovery is something else

When a workout you started ends, the watch keeps measuring your heart rate for three minutes to estimate it. It is a prediction of your 60-second heart rate recovery after submaximal effort at 85% of predicted maximum, capped at 55 bpm. Apple says higher values can indicate better cardiovascular health (Apple, 2024). It is a figure tied to a workout, not a score for how you are today. The paper does not say where to view it.

What these scores are usually built from

They usually combine sleep, HRV, resting heart rate and recent load. Each signal has its weak point.

Why they are estimates

Halson, 2014, reviews markers of load and fatigue in athletes: very few have solid scientific evidence, and no single definitive marker exists. It is about athletes, not watches, but it describes the problem: a score combines imperfect signals using weights someone chose. Apple says it developed its algorithm with data from the Apple Heart and Movement Study (Apple Newsroom, 2026). It does not give the weights on the pages we read, and as of 2 November 2026 we have found no independent validation on PubMed.

The limits of load ratios

The acute:chronic workload ratio divides recent load by the load of the preceding weeks. Impellizzeri et al., 2020, conclude there is no evidence supporting its use in load management aimed at reducing injury risk, that it is an inaccurate metric and that it adds noise. Their critique concerns injury risk, not whether it is worth looking at how much you train.

The ratio rises because the denominator fell, not because you did more. Apple's Training Load also compares 7 days with 28, but its page does not give the formula, so we do not know whether this critique applies to it.

What makes a score trustworthy

• It compares you with yourself. Halson, 2014, presses the case for individualising, and Apple analyses vitals against your recent baseline.
• It shows what is missing: with no sleep from last night, it says so and does not quietly fill the gap.
• It gives a range or a confidence level. A bare 74 looks more exact than it is.
• It says what is moving it, so you can check.
• It is clear about when its data are from: some scores change during the day and some are fixed in the morning.

These criteria are our own, not from a study.

How GoForIt builds its number

GoForIt is not out yet, so this describes how its readiness is designed:

• It is an estimate from 0 to 100 that combines sleep, heart rate variability, resting heart rate and recent load.
• It is always compared with your own range, never a generic average.
• When a signal is missing, it says so. Anything it estimates from your usual range is marked ‘Estimated’, with a confidence level and a likely range.
• What you train today counts for tomorrow: it does not change today's number.
• It is not a medical diagnosis: it is an estimate that gets more precise with more days.

With sample data it would look like this: 74, ‘Estimated · Medium confidence · 70-78’. The detail is in how it works. To read the signals one at a time, see the HRV guide and HRV on Apple Watch.