Why your sleep score and how you feel disagree — and which one to trust

Short answer

If your sleep tracker says you slept well but you feel tired, the most likely explanation is that both are right about different things. A wearable estimates your night from movement and heart rate — how long you lay still, how your pulse behaved, roughly when you drifted off. You judge the night by the day that follows, and your day is shaped by pain, mood, stress, illness, medication and hormones as much as by the night itself. Research on how people define sleep quality finds exactly this: good sleepers and people with insomnia alike rate a night mainly by how tired they feel on waking and through the day (1).

So when the number and the feeling point in opposite directions, you are not imagining it, and nothing has malfunctioned. The device answered "how long, and how still?" You answered "how do I function today?" Different questions, different answers — and the useful move is not to pick a winner but to learn what each is for.

Note: this article is for education and is not medical advice. Persistent unrefreshing sleep, heavy daytime sleepiness, loud snoring or pauses in breathing, or exhaustion that does not improve with rest deserve assessment by a clinician, whatever your tracker says.

Why does my sleep tracker say I slept well when I feel tired?

Because a consumer sleep tracker is good at one part of the question and weak at the rest — and weakest at the part you care about most.

What the hardware does. A wrist or finger device has two signals: an accelerometer that senses movement, and an optical sensor that reads your pulse through the skin. From those an algorithm infers when you fell asleep, when you woke, and — more speculatively — which stage of sleep you were in. There is no electrode on your scalp, so no direct measurement of brain activity, which is what sleep staging is defined by.

How well that works against the laboratory standard. In a sleep laboratory, 34 adults were recorded for three consecutive nights — one deliberately disrupted — with full polysomnography alongside seven consumer devices. Epoch by epoch, the devices were very good at spotting sleep (sensitivity at least 0.93 across all of them) and much weaker at spotting wakefulness (specificity between 0.18 and 0.54). Stage results were mixed, and every device did worse on the disrupted night (4). A second study put six widely used devices on 53 adults against polysomnography: agreement on "asleep or awake" was 86–89%, but agreement on which specific stage you were in fell to 50–65% (5).

So a tracker knows roughly when you were asleep. It is unreliable about the minutes you spent lying awake, and it is not a staging instrument at all — even though the stage chart is usually the prettiest screen in the app. The American Academy of Sleep Medicine's position is that, lacking validation and regulatory clearance, consumer sleep technologies cannot be used to diagnose or treat sleep disorders (6).

And you are not rating the night at all. You are rating the morning and the day — which is why a seven-hour night can arrive with a heavy, fogged morning attached, and both accounts can be accurate.

Why doesn't feeling unrefreshed show up in the numbers?

Because "unrefreshing sleep" is its own thing, and sleep science has spent decades failing to find it reliably in measurements.

The clinical term is non-restorative sleep: persistently waking unrefreshed despite a normal sleep duration, with no sleep disorder to explain it. A review of the literature found this experience is reported by a substantial share of the population and tracks with daytime impairment, pain and fatigue — and that it has no agreed objective definition, which is precisely why it is so hard to see in a dataset (8). If what you are experiencing has never been pinned to a measurable feature of the night, no sleep score can be expected to reflect it.

In fibromyalgia, symptoms track what people report, not what the device records. Seventy-five people with fibromyalgia wore an actigraph continuously and kept an electronic morning diary for a week. The two methods disagreed by an average of 73 minutes per night, and on roughly one night in five by more than two hours. Crucially, symptom severity was related to people's own report of poor sleep — and not to the objective sleep data (9).

In ME/CFS, objective sleep looks close to normal while the complaint is severe. A review of the polysomnography literature in chronic fatigue syndrome/ME concluded that objective measures have found few differences from healthy controls, in sharp contrast to the near-universal complaint of disturbed, unrefreshing sleep (10). NICE defines the symptom plainly: sleep that is non-restorative, where even after a full night people do not feel refreshed (18).

And the day does not run only on the night. Pain, upright posture, the aftermath of exertion, mood and infection all shape how a morning feels, and none of them is a sleep variable. A wrist sensor sampling movement and pulse has no access to any of them.

So when a tracker shows a decent night and you feel wrecked, the reading is neither "the device is broken" nor "you are exaggerating". Whatever is making you feel this way is not something a wrist sensor measures — and in several conditions, not something a sleep laboratory reliably measures either.

Which one is right when they disagree?

Usually both, about different questions. The table below splits the night and the day into the parts a device can answer and the parts only you can.

Read down the last column and the rule writes itself. For how long you slept across a run of nights, believe the device. For how you are doing, believe yourself. For noticing slow change, use a consistent device against your own baseline. For deciding what to do today, your body holds information the device does not.

Physiology and feeling can also diverge for weeks, in the direction people least expect. In a two-year prospective cohort of 4,795 people tracked through infections with smartwatches, those with moderate-to-severe COVID-19 reported symptoms resolved after about 12 days — but smartwatch-measured heart rate and heart-rate-variability-based stress took an additional 60 days on average to return to baseline (17). Feeling recovered and being physiologically back to baseline are not the same moment.

Can seeing a number change how you feel?

Yes, measurably — which is the strongest practical reason to be deliberate about when you look.

In a controlled experiment, 63 adults meeting diagnostic criteria for insomnia disorder wore an actigraphy-and-diary watch and were randomly assigned to positive or negative sham feedback about their sleep efficiency at their usual rise time, delivered the way a wearable would deliver it. The nights were not different. The days were: by evening, the group told they had slept badly showed lower alert cognition (effect size d = 0.79) and more sleepiness and fatigue (d = 0.55) (13). The same effect appears in people without insomnia: across two studies with 164 participants, people told — on no real basis — that they had spent an above- or below-average share of the night in REM sleep scored accordingly on attention and verbal-fluency tests, while their own reported sleep quality predicted nothing (14).

That is not a reason to distrust all data. It is a reason to notice what a number does to you before you build a day on it.

The discrepancy can also work the other way. In a trial with 48 people with insomnia, half were shown the gap between their sleep diary and their actigraphy through a hands-on behavioural experiment, and half were simply told about it. The behavioural-experiment group had far larger reductions in self-reported sleep impairment, insomnia symptoms and sleep-related anxiety and distress (effect sizes 0.79 to 1.25) than the group given verbal feedback (−0.06 to 0.31) (15).

Discovering you slept more than you thought can make you feel better — if you discover it rather than being told. The risk sits at the other end of the same mechanism: clinicians describe patients arriving distressed about self-diagnosed sleep problems based on tracker output, who find the tracker's version of the night more convincing than polysomnography. The term is orthosomnia (7), covered in our article on orthosomnia and chronic illness.

Why do recovery scores disagree with you even more than sleep scores?

Because of what they are built from. A sleep score mostly summarises the night: duration, timing, continuity. A recovery or readiness score leans on overnight heart rate and heart rate variability, which respond to a far wider world than the night alone.

HRV's link to how people feel is real but small. Across five longitudinal studies using consumer wearables — two smartwatches, two chest straps and one smart ring — resting HRV showed small-to-moderate associations with health measures including depressive symptoms (r = −0.22) and sleep difficulty (r = −0.11) (16). A correlation that size means the two move together on average across a population, not that they will agree for you on a Tuesday.

HRV is also naturally noisy day to day. In athletes monitored for 16 weeks with a wrist wearable, the week-to-week coefficient of variation in log-transformed HRV was about 5.4%, and in resting heart rate about 7.6% — figures the authors note sit within or below the roughly 3–13% day-to-day variability seen with other HRV measurement protocols (12). That variability is not device error; it is how the measure behaves. It is why a single morning reading carries far less information than a seven-day trend, and why a recovery score can swing while nothing about your night changed.

And it responds to yesterday, not only to last night. Alcohol, a late meal, a hard session, the start of an infection, a warm room, a stressful evening, how still you were lying when the reading was taken — all move overnight heart rate and HRV. The infection cohort above is the clearest published example of physiology and felt experience running on different clocks for weeks (17).

None of that is a flaw. A measure that responds to your autonomic nervous system is useful precisely because it picks up things you do not consciously notice. It was simply never answering the question "do I feel rested". To take a morning reading consistently enough for the trend to mean something, see our guide to morning HRV and pacing.

What does "unrefreshing sleep" mean if the numbers look fine?

For many people with a long-term condition, this is the version of the mismatch that hurts most: sleep that looks adequate on every measure and still leaves you feeling as though you did not sleep. It is a recognised feature of ME/CFS, fibromyalgia, long COVID and several other conditions.

What a wearable sees on those nights is often unremarkable — a reasonable duration, a normal-looking heart rate curve, a score in your usual range. What it cannot see is whatever stops the sleep restoring you. In ME/CFS that is not a failure of consumer hardware: laboratory measures have also struggled to find consistent differences from healthy controls (10). A normal sleep score is not evidence against your experience. It is evidence that what you are experiencing is not what a sleep score counts.

But a normal-looking tracker is not a reason to stop looking. In 343 consecutive patients attending a fatigue clinic with a chronic fatigue syndrome diagnosis, a single night of polysomnography found a primary sleep disorder sufficient to explain the presentation in 104 of them — 30.3%. Of the remaining 239, 89.1% met criteria for at least one objective sleep problem (11). That is close to a third of a fatigue-clinic population with a treatable sleep disorder no wrist device would have identified. Sleep apnoea, restless legs, pain that fragments the night, medication effects and hormonal change can all produce unrefreshing sleep, and several need a sleep assessment to find.

If you also track resting heart rate and find it higher than the charts say it should be, our article on resting heart rate in chronic illness explains why that is usually the wrong comparison.

How should you use metrics that disagree with you half the time?

Rate how you feel before you look. Write down your sense of how rested you are, 1 to 5, and only then open the app. This is the practical answer to the sham-feedback research: a number that arrives second cannot set your mood (13).

Use trends, not single mornings. A seven-day average of sleep duration or resting heart rate carries far more signal than one reading, given how much these measures move on their own (12).

Treat the data as a second opinion. When the number says you should feel fine and you do not, it is answering a different question. Use your body for today and the number for the pattern over months.

Ignore the stage chart. Of everything on screen, sleep-stage percentages agree least with laboratory measurement (5).

Watch for sustained disagreement. One odd morning means nothing. Weeks of waking unrefreshed despite fine-looking data is a clinical signal (11).

Do not let a good number overrule a bad day. In fibromyalgia, symptom severity tracked people's own reports of poor sleep and not their objective sleep data (9). A normal number does not mean your symptoms are not real.

A two-week experiment to learn how your data and your body relate

Population averages tell you that disagreement is common. They cannot tell you how your numbers relate to your experience. Two weeks of simple records usually can.

Every morning, before you look at anything: rate how rested you feel from 1 to 5. Add one word if something stands out — pain, headache, wired, groggy.

Then check your data and write down two numbers: total sleep time, and whichever recovery or HRV measure you use. Two, not ten.

Every evening: rate how well you functioned that day from 1 to 5. That is the outcome you care about, and the one no device records.

At the end of two weeks, look for four things:

1. Does your morning rating move with your sleep duration at all? For some people it does; for many it barely does, and knowing that is worth more than any score.

2. Does your recovery measure line up better with your evening rating than with your morning feeling? Sometimes the body signal is ahead of conscious awareness.

3. On the days the two disagreed most, what else was going on — pain, stress, alcohol, a late meal, your cycle, an infection starting, a big day before?

4. Which measure, if any, is worth keeping? If a number has told you nothing in fourteen days, you are allowed to stop looking at it.

For a ready structure for the daily entries, see our guide to an activity and symptom diary.

What does the evidence not show?

That wearables are useless. They are reasonably good at what they actually measure — sleep duration and heart rate trends against your own baseline (4).

That feelings are unreliable. How you function is the outcome that matters, and no device measures it. The non-restorative sleep literature treats the subjective report as the thing to be explained, not as noise (8).

Prediction of anything. Nothing here supports using a sleep or recovery score to forecast tomorrow, or to anticipate a crash, a flare or an illness. These studies measure agreement with polysomnography on the night in question.

Generalisability to everyone. Most device validation has been done in small samples of healthy adults, in laboratories, over one to three nights (4, 5). Performance in older people, in fragmented sleep and in specific conditions is much less established — and the published signal is that it degrades on exactly the disrupted nights patients have most of.

Living with data that disagrees with you: the questions people ask next

Should I keep wearing my tracker if it keeps saying I slept fine when I feel awful?

That depends on what it is doing to you. If the score has become something you argue with every morning, the research on sleep feedback suggests it can worsen the day before it has started (13). A practical approach is a two-week trial: keep wearing it, rate how you feel before you look, then decide whether any number earned its place. Some people keep only sleep duration and resting heart rate and hide the rest; some stop for a month and find their mornings improve.

How do I explain to my manager that I slept eight hours and still cannot work?

Keep the tracker out of it. A sleep score invites the reply "but it says you slept". What travels better at work is function: what you can do, for how long, and what happens afterwards. "I can do focused work for about two hours in the morning and need a break before the afternoon" is a workable sentence; "my recovery score was 40" is not. For a formal conversation, bring two weeks of your own 1-to-5 function ratings rather than screenshots, and name the adjustments you need.

I feel terrible after exercise but my sleep score looks fine. Should I push through?

No — and the right next step depends on which condition you have, because the guidance differs. If you live with ME/CFS or long COVID with post-exertional malaise, NICE's guideline is explicit that physical activity may make symptoms worse, and that activity should stay within your energy limits and be reduced if symptoms worsen; fixed incremental increases are not the approach (18). For POTS, fibromyalgia, MS, lupus and rheumatoid arthritis, structured exercise is part of standard care. Ask your doctor which applies to you, and see our guide to pacing with chronic illness.

Does alcohol explain why my recovery score drops even when I slept my usual hours?

It is one of the common explanations, and a good illustration of why recovery scores are not sleep scores. Overnight heart rate and HRV respond to many things unrelated to how long you lay in bed — alcohol, a late or heavy meal, a hard session, a warm room, the start of an infection, how still you were when the reading was taken. These measures also move by several percent from day to day when nothing notable happened (12). To test it, log your evenings and look across several weeks, not one night.

My sleep data looks worse on holiday but I feel better. What is going on?

This is the whole thesis in holiday form. A later bedtime, a strange bed, a warm room and a drink with dinner can push duration, continuity and overnight heart rate in the "worse" direction, while the thing that actually governs how you feel — no alarm, no commute, no deadline — has improved enormously. People rate a night largely by the day that follows it (1), and your day improved. Trust the holiday feeling; it is the measure with clinical meaning.

How do I explain to my family that my data looks fine but I feel terrible?

One sentence usually does more than a chart: "the watch counts how long I lay still, not whether the sleep did anything for me." You can add that this is a described, named thing — non-restorative sleep, waking unrefreshed despite normal duration, with no agreed objective marker (8). If someone wants a number to hold on to, give them your own morning and evening 1-to-5 ratings instead of the app's score. Those describe what they can actually see in you.

Everyone keeps telling me my numbers look great. How do I stop doubting myself?

By separating two claims. "Your sleep score is normal" is a statement about movement and pulse overnight. "You feel unrestored" is a statement about your day. The evidence says the second is not reducible to the first: non-restorative sleep has no agreed objective marker (8), and in fibromyalgia symptom severity tracked people's own reports rather than their actigraphy (9). A normal score is not a counter-argument to your experience. It measures something else.

Is this forever? Will my data and my feelings ever line up?

For some people they come closer. If a treatable sleep disorder is found and treated, if pain is better controlled, if an infection resolves, the number and the feeling can start moving together. For others — particularly where unrefreshing sleep is a core feature of a long-term condition — the gap stays, and the goal changes from closing it to using each side for what it is good at. The research does not promise convergence. It does show the gap is a described phenomenon rather than a personal failure.

How to bring this up with your doctor

See a clinician if you have felt unrefreshed by sleep for three months or more, if you are excessively sleepy in the day, if someone has noticed loud snoring or pauses in your breathing, or if exhaustion is not improving with rest — whatever your tracker says.

Say the mismatch out loud, in one sentence. "My tracker says I sleep seven to eight hours, but I wake unrefreshed almost every day." That describes a recognised clinical picture — non-restorative sleep with normal duration (8) — and opens the right conversation without turning the appointment into a debate about gadgets.

Bring averages and your own ratings, not screenshots. One page: two weeks of total sleep time and time in bed, your daily 1-to-5 morning rating of how rested you felt, and your evening rating of how you functioned. Label the device numbers as personal trends from a consumer wearable, not measurements.

Ask these specifically. Could this be a sleep disorder such as sleep apnoea that a wearable would not detect? Should I have a sleep study? Could my condition, my pain or my medication explain unrefreshing sleep? Would cognitive behavioural therapy for insomnia be appropriate for me?

Expect the data to be treated cautiously, and that is correct. The American Academy of Sleep Medicine's position is that consumer sleep technology cannot be used to diagnose or treat sleep disorders, but can support the conversation inside a clinical evaluation (6).

If you are dismissed because the numbers look fine. "I understand the numbers look normal. What I am reporting is how I feel every morning, and it has not improved. Can we record that and consider a sleep assessment?"

And do not put these down to bad sleep: chest pain, fainting, breathlessness at rest, a new severe headache, sudden weakness on one side, or witnessed pauses in your breathing need urgent care — call 911 in an emergency.

How Welltory helps — and what it cannot do

The limits first. Welltory is a general wellness product, not a medical device. It does not diagnose, predict, monitor, prevent or treat insomnia, sleep apnoea, ME/CFS, fibromyalgia or any other condition, and it cannot tell you whether a night restored you — no device can, because non-restorative sleep has no agreed objective marker. What it can do is keep a consistent physiological record beside your own account of the day.

Log what happened, in your own words. If you wear an Apple Watch or Oura, the Today screen (iOS) flags stress stretches and asks "What happened?". Tap a suggested tag, type a few words, or just talk: "woke unrefreshed", "slept 8h still wrecked", "pain woke me", "wired at 2am", "alcohol", "late meal", "period day 1", "big day yesterday". You can also add a note any time with the plus icon ("Share your thoughts…"); it goes into your Journal. Rate how you feel before you open the app — a number cannot set your mood if it arrives second.

Look at the one to three days before, not just this morning. Put sleep analysis, stress minutes, Battery, resting heart rate and HRV beside your own notes for the days leading up to a bad morning. Stress minutes and sleep analysis need a supported wearable; with the morning phone-camera (PPG) reading alone you have your spot readings, and Heartbeat Report is built for taking those under the same conditions each day. Read recovery here too — as a trend against your own baseline, never a score to beat.

Check My Patterns after two to three weeks of tagging. My Patterns collects the tags you add to stress and rest stretches on the Today screen, so it needs iOS with an Apple Watch or Oura; a phone-camera reading alone does not create those stretches. Patterns begin to appear at around 7 tagged events, and insights typically need roughly 7 occurrences of a tag in the current month plus some history from the month before. It shows which tagged situations tend to come with stressful stretches, rarer tags your body reacts to strongly, trends by day of week, and every time a tag occurred. Some sections need a paid plan.

Build a log for your appointment. The Journal (Premium) holds HRV measurements, tags, mood, how you feel physically, notes and workouts. For a longer record, export a CSV from the web app (Dashboard → choose a chart → Export) and bring it alongside your own ratings. The free version keeps 30 days, so export what you want to keep.

Anything you find this way is an association to discuss with your doctor — not proof of a cause, and not a warning system.

Where to find other people in the same situation. If you live with an energy-limiting condition, Welltory runs a paid, moderated community called Energy Lab for women aged 18 to 65 living with ME/CFS, long COVID, fibromyalgia, POTS, MCAS and similar. It runs alongside the app: you keep collecting your own data, and the Lab is where people learn to read it together, with a medical board answering the science. It is education and peer support, not medical care, it does not replace your clinician, and there is a 14-day money-back guarantee on a first purchase.

How we made it

Made with AI tools, then edited and fact-checked by the Welltory team. See our Editorial & AI policy.

Data analysis by Jane Smorodnikova, co-founder of Welltory and the person who built the methodology behind how we read physiological data.

Written by Tatsiana Yashyna.

Discounts for blog readers: up to 36% off

See what affects your energy, stress, sleep, and daily state with Welltory

Explore your data