The green lights on the back of your watch
Turn your watch over. There are small lights on the underside, usually green, sometimes red, sometimes a dull infrared you can only half see. They are on all night.
That is how a watch knows your heart rate, your heart rate variability, your breathing and most of what it claims about your sleep. It is worth understanding, because everything on the rest of this site rests on whether those numbers are any good.
What the lights are doing
The technique is called photoplethysmography, or PPG. The word is unhelpful and the idea is simple.
Blood absorbs light. Every time your heart beats, a pulse of blood pushes through the vessels under your skin, and for an instant there is more blood there and it absorbs more light. A moment later the vessels empty slightly and absorb less.
So the watch shines a light into your wrist and measures how much comes back. What returns is a wave: a little less light at each beat, a little more between them. Count the peaks and you have a heart rate. Measure the gap between each peak and the next, to the millisecond, and you have heart rate variability.
That is the whole trick. A light, a detector, and the shadow your own pulse casts.
Why green
Different colours of light reach different depths, and that turns out to matter more than anything else.
Green, around 520 to 530 nanometres. Haemoglobin absorbs green strongly, so the signal is crisp. Green does not penetrate far, which sounds like a weakness and is actually the point: shallow light only sees the small vessels just under the skin, so it picks up less movement from the deeper tissue sloshing about when you walk. Green gives a better signal-to-noise ratio and better heart rate detection than infrared, and it holds up better in motion.
Red, around 660 nm, and infrared, around 940 nm. These go deeper. Depth is what you want for blood oxygen — the difference between how red and infrared light are absorbed is how a watch estimates SpO2 — but depth also means the reading picks up far more movement artefact. The deeper the light goes, the more your arm swinging affects it.
So most wrist wearables use green for heart rate and HRV, and red and infrared for blood oxygen. Some use the infrared channel as a movement reference: read the noise on the deep channel, subtract it from the green one. Done well, that takes heart rate error in motion from around 4 beats a minute down to about one.
Your phone camera can do the same thing with a fingertip and the flash, which is what HRV4Training and similar apps use.
Is it as good as an ECG?
An ECG measures the electrical signal that tells the heart to beat. PPG measures the pulse of blood that arrives afterwards. They are not the same event, and the gap between them wanders slightly, which is why researchers call the PPG version pulse rate variability rather than heart rate variability.
The honest answer has two halves, and the second half is the important one.
At rest, it is very close indeed
A 2026 systematic review and meta-analysis of 43 studies found small pooled errors against ECG — about 0.19 of a standard deviation for RMSSD and 0.13 for SDNN. Small enough to work with.
A 2025 study looked specifically at posture, and found:
| RMSSD | SDNN | |
|---|---|---|
| Lying down | 0.955 | 0.980 |
| Seated | 0.834 | 0.921 |
Those are intraclass correlations against ECG, where 1.00 would be perfect agreement. Lying down, a wrist sensor and a medical ECG are telling you almost exactly the same thing.
Overnight it is the same story. A validation of WHOOP against ECG during slow-wave sleep found correlations of 0.98 to 0.99 for RMSSD.
In motion, it is not
Everything above falls apart when you move. Sitting up costs you accuracy. Walking costs more. Running costs a great deal. Skin temperature, how tight the strap is, and where on the wrist it sits all matter as well.
Which is exactly why the readings are taken the way they are
This is not a detail. It is the reason the protocol looks the way it does.
The overnight numbers come from the quietest hours you have. Asleep, still, warm, the strap in the same place all night. That is the single best condition PPG has, which is why the overnight average is the steadiest number a watch produces.
The morning reading is taken lying down, before you get up. Not because it is convenient — because supine is where agreement with ECG is near perfect, and sitting up measurably degrades it. Same posture every day, before coffee, before your phone.
A wrist sensor used this way is doing something close to laboratory work. The same sensor read during a workout is doing something much rougher, and nothing here treats those two numbers as the same.
What to be careful about
RMSSD tends to read slightly high on PPG. It is the measure most prone to systematic overestimation. That does not matter for tracking yourself over time, because the bias is in the same direction every day. It does matter if you compare your number to somebody else’s, or to a figure from a paper that used ECG.
Skin tone. A systematic review of ten studies found the evidence mixed — four found significantly reduced accuracy in darker skin tones, four found no difference, two were mixed. Not settled, and not nothing. Green light is absorbed by melanin as well as by blood, and anyone whose readings look implausible should suspect the sensor before they suspect themselves.
Tattoos, cold hands and a loose strap all degrade the signal for the same reason: less light getting back cleanly.
Every brand’s algorithm is its own. Two watches on the same wrist will not agree, and neither is lying. Pick one and stay with it, because the comparison that matters is you against yourself.
The short version
The lights on your watch measure the shadow your pulse casts. Green for heart rate because it stays shallow and shrugs off movement; red and infrared for oxygen because they go deep.
At rest, and lying still, what comes back is close enough to an ECG to work with — and that is precisely the condition these readings are taken in. In motion it is a rougher instrument, and it is treated as one.
Twenty years ago this was a laboratory. Now it is on your wrist, and it runs every night whether you think about it or not.
Every figure above is from a published source, and they are listed with their DOIs on the references page.