The instrument

The Even score

One number for how you are doing, and everything behind it.

Zero to a hundred, and higher is lighter. Fifty is the middle of your own range, not the middle of everybody's. The number moves because something measurable moved, and every part of it opens into what it was read from.

Five words, five marks

The score lands in one of five bands. They are symmetric about Steady, which is a true neutral: two bands above it and two below, so the middle of the scale does not already lean one way. The same five run along the check-in dial, so what Even reads and what you say use one vocabulary.

The Spent mark, a soft rose form with a downturned mouth — Spent0 – 19

The Stretched mark, a yellow form pulled thin at the waist — Stretched20 – 39

The Steady mark, a blue cushion at rest with a level mouth — Steady40 – 59

The Light mark, a teal bloom, open and smiling — Light60 – 79

The Bright mark, a green bloom, further open and smiling wider — Bright80 – 100

Below the middle the form deforms, at the middle it rests, above it the form opens.

Read against you, not against everyone

A resting heart rate of 61 is meaningless on its own. Against your own last ninety days it can be four beats above where you usually sit, which is a fact about your week. Almost everything Even says is that second kind of statement, because it is the kind that can be acted on.

Population comparisons are used in exactly one place, sleep regularity, because that is the one measure in the app with a published normative distribution Even can honestly place you in. Everywhere else the app speaks about you relative to you, and where a comparison would need a reference that does not exist, it stays quiet rather than inventing one. Heart rate variability is the clearest example: Apple Watch samples it opportunistically rather than over a fixed overnight window, no published distribution describes that measurement, so Even will never tell you which percentile your HRV is in.

The five parts

Each is a 0 to 100 reading in its own right, weighted into the whole. Tap any of them in the app and it opens into the nights, samples and citations it was read from.

Sleep

Length against the sleep you told Even you aim for, how broken the nights were, how late they ran, and how steady one night is against the next. Read from sleep and, where an Apple Watch records it, sleeping wrist temperature. The steadiness half is the Sleep Regularity Index, and it is the flagship measure in the app.

Body

What your body did while it was not doing anything: heart rate variability, resting heart rate and breathing rate, each against your own baseline rather than a chart. This is the part that notices a hard fortnight before you do, because resting physiology drifts before anybody feels it drift.

Demand

How hard the day itself was, from four quick sliders at check-in. They are adapted from the NASA Task Load Index, the workload instrument published by Hart and Staveland in 1988 and used for whole-day rating in more recent work. This is the part a wrist cannot see: two identical physiological days can be a quiet Sunday and the worst Tuesday of your year.

Mood

How it landed, said in ten seconds. Even logs on Apple's State of Mind model, so a mood recorded in Even appears in Health, and a mood recorded anywhere else is read back into the score.

Activity

Steps, active energy, exercise minutes and time in daylight. Movement and light are both inputs to how a day goes and outputs of how it went, which is why they sit in the score rather than in a separate fitness view.

What Even does when it does not know

Thin data gets a best estimate and a quiet note that it will sharpen, never a fabricated figure and never a wall of what is missing. A reading that cannot be made honestly is shown as a countdown rather than hidden, so you can see what is still coming.

This is enforced in the code rather than left to good intentions. Nothing in the app can turn a plain string into a claim on screen: a claim is computed from a measured value that carries its own provenance and confidence, and a comparison against a population cannot be constructed at all without a real citation attached. Sentences that would be true of anybody are treated as bugs, and the test suite fails the build if one sentence turns up in more than half of two hundred generated profiles.

It tunes itself to you

The weights are not fixed forever. As you mark how sessions landed and how days actually felt, Even tilts the parts of the score toward the ones that predict relief for you. Someone whose bad days are almost always short nights and someone whose bad days are almost always heavy ones should not be read by the same instrument, and after a few weeks they are not.

For anyone who tracks their cycle, the second-half rise in resting heart rate and temperature is measured from your own data and corrected for, rather than being read as strain. That correction comes from your history, not from a textbook curve.

What the score is not

It is not a readiness score for training. It is about the weight on your mind and what your body is doing under it.
It is not a diagnosis, a screening tool, or a measure of any medical condition.
It is not a grade. A low morning is information about a week, not a verdict about a person, and Even is built so that it can never make you feel bad for not opening it.

The Today screen: an Even score of 26 of 100, stretched and below the person's usual, with sleep, body, mood and activity chips underneath and one plain-language line — Twenty-six of a hundred, stretched, with the four parts it was made of and one line about what to do with it.

Even describes patterns in your own data. It is not a medical device, it does not diagnose or treat anything, and nothing in it is medical advice.

NextSleep regularity, and why it beats sleep duration AlsoThe nine studies Even is designed on