Garmin vs Oura vs Whoop: which recovery score should you believe?

If you wear more than one device you have already had the morning where the ring says 84 and the strap says 41. Both are measuring you correctly. The disagreement is not a malfunction, and understanding where it comes from tells you which number to trust for what.

What is each recovery score actually built from?

All three start from overnight autonomic signals — heart rate, heart-rate variability, sleep — and then weight them differently according to what the company thinks recovery means.

Garmin Body Battery is a running energy budget across the whole day, draining with stress and activity and charging with rest, derived largely from HRV-based stress. It answers "how much have you got left right now".

Oura Readiness leans hardest on sleep and on overnight physiology, including body temperature deviation, and compares them to your own baselines. It answers "how well did your night go".

Whoop Recovery weights HRV and resting heart rate during slow-wave sleep against your baseline, explicitly in the context of accumulated strain. It answers "how ready are you to take on load".

Those are three different questions. A morning where you slept well after a very hard week can genuinely produce a high Oura score and a low Whoop score, because both are right about the thing they are measuring. The general anatomy of these scores is in training readiness score explained.

Why do the scores disagree even when the inputs agree?

Three reasons, in order of size.

Different weightings and different baselines. Each score is normalised against your history on that device. Swap devices and your own score resets; the number is not portable and was never meant to be.

Different measurement windows. A score built from the last three hours of sleep and one built from the whole night will disagree on a night with a rough start and a calm finish.

Sensor conditions. Optical heart-rate accuracy is not uniform. Bent and colleagues, testing six consumer and medical wearables, found significant differences between devices and between activity types, with absolute error during activity averaging about 30% higher than at rest — and notably no significant difference across skin tones. A finger ring and a wrist strap are not in the same measurement conditions overnight.

PeakU does this from your own data. A coach that reads everything you already track. Join the free beta →

How accurate is the sleep data behind the scores?

Sleep staging is shakier than heart rate. Chinoy's laboratory comparison of seven consumer devices against polysomnography found most were decent at detecting sleep versus wake and considerably weaker at staging it. Treat duration and consistency as solid; treat a deep-sleep percentage as an estimate.

This matters for the scores because they weight sleep differently. A score that leans on total sleep time is standing on firm ground. One that leans on how much of it was deep is standing on an estimate, and two devices will estimate it differently for the same night.

Which recovery score is the most accurate?

None of them, individually, as an instruction. All of them, collectively, as evidence. The question is slightly wrong: they are not measuring the same construct, so there is no single truth for one of them to be closest to. Judge them on the inputs you can compare across devices, resting heart rate and HRV against your own baseline, rather than on the score, which you cannot. What a comparable HRV baseline looks like is in what is a good HRV for my age.

Which one should you believe this morning?

The useful rules: trust trends over single days, because every one of these scores is noisy day to day and informative week to week. Trust agreement — when two independent devices both say you are down, that is a real signal. Trust the inputs over the score: if resting heart rate is up five beats and HRV is well under baseline, that is a fact, whereas the score built on top of it is an opinion about what the fact means.

And when they disagree sharply, ask which question you actually have. Deciding whether to do intervals is a Whoop-shaped question. Deciding whether you are getting ill is an Oura-shaped one. Deciding whether to take the stairs at four in the afternoon is a Garmin-shaped one.

A practical habit: pick one device as the tie-breaker for training decisions, so that a disagreement never becomes a reason to skip. Read the other as a second opinion on illness and sleep. If the tie-breaker says train and your resting heart rate is five beats up on both devices, the raw inputs win, and you train easy. The rule costs nothing and removes the morning argument with yourself.

Should you wear two recovery devices?

If you already own two, yes, and use the disagreement. If you are choosing one, pick the one that answers the question you ask most mornings and wear it the same way every night. Two devices measuring the same night independently is more information than one; two devices both writing the same data into your health store is a duplication problem, not a measurement one, and the fix for that is in how to combine Garmin, Samsung Health, Strava and Hevy without double-counting.

1. Let each device's own app write its sleep and HRV; stop one of them writing steps and workouts.
2. Compare the trends, not the morning numbers.
3. When they agree, act. When they disagree, read resting heart rate and HRV directly.

What none of the scores can see

Your training outside the device's view. A wrist device records a lifting session as an hour of elevated heart rate and no idea of the load. None of the three knows what you squatted, and a ring will often miss that you trained at all.

They also cannot see the fitness markers that move over months rather than nights; a recovery score tells you nothing about whether your aerobic capacity is improving, which is what VO₂max for your age measures. Recovery is a daily question, fitness a quarterly one, and the scores only answer the first.

What PeakU does with this

PeakU does not add a fourth opinion, and it does not ask you to pick a winner. It reads whichever of these you already have through Apple Health or Health Connect, plus the training your wearable never sees — lifts from Hevy, runs from Strava — and computes one Peak Score over the combination.

Two things follow from that. It can see the disagreement and say so, rather than silently picking one. And it shows the contributions: which signal pushed the number up, which pulled it down, and by how much, so you can check its reasoning against what you know about your own week. Your watch's score is an input to the answer, not a rival to it.

Hen Eytan · Founder, UPeak

Builds the UPeak apps and writes up the research behind them.