The Sleep Screen: Duration, Efficiency and Consistency

The formula

efficiency  = asleep / inBed                                      -> 40%
duration    = clamp(asleep / 8h, 0, 1)                            -> 40%
consistency = 1 - clamp(|bedtime - avg14dBedtime| / 90min, 0, 1)  -> 20%
score = round(100 * (0.4*efficiency + 0.4*duration + 0.2*consistency))

Efficiency is time asleep over time in bed, read from the platform's sleep samples. It captures both how long it took you to fall asleep and how much of the night you spent awake.

Duration is measured against an eight-hour ideal and clamped at 1, so nine hours does not score above eight. Sleeping longer than you need is not a training advantage, and letting it inflate the score would reward the wrong thing.

Consistency is how close your bedtime was to your own fourteen-day average, with a ninety-minute tolerance before it starts costing anything. It is your own average, not a prescribed bedtime — the app has no opinion about whether you should go to bed at ten or at one, only about whether it was the same as usual.

Why eight hours

The evidence for sleep as a training variable is stronger than for almost any supplement.

Sleep restriction measurably suppresses muscle protein synthesis. It raises perceived exertion at a fixed load, which means the same weight genuinely feels heavier rather than just feeling worse. And it degrades the endocrine environment both strength and endurance adaptation depend on.

Eight is a working target rather than a universal requirement, and clamping at eight rather than penalising nine is the acknowledgement of that.

Why awake time is not rated

There is no defensible target band for it.

Some awake time in bed is normal and unproblematic. A lot of it usually indicates something — but "a lot" depends on total time in bed, on age, and on what the person was doing. Rating it directly would mean picking a threshold that is wrong for a substantial fraction of people.

It is judged indirectly, through efficiency, which is where it actually shows up: time awake in bed sits in the denominator and not the numerator. That is the honest place for it.

The hypnogram, and what to distrust about it

The screen draws the night's stages — awake, REM, core, deep — as a chart. It is genuinely useful for spotting the shape of a night: fragmentation, a very late first deep block, an unusual amount of awake time in the small hours.

It is also less accurate than the confident four-colour chart implies, and that is worth saying on the page rather than leaving to be discovered.

Consumer wearables infer sleep stages from heart rate, heart rate variability and movement. Stage classification against polysomnography — the clinical standard, with electrodes on your scalp — is imperfect, and it is particularly weak at distinguishing light sleep from deep sleep. Total sleep time and efficiency are far more reliable than the stage breakdown.

That is exactly why the score is built on duration and efficiency and not on deep-sleep minutes. Deep sleep minutes are the number people fixate on and the number the device is least confident about.

Reading it as a hybrid athlete

Two things about sleep are specific to training both halves.

Two-a-days compress the sleep window from both ends. An early session and a late one leave a shorter night in between, and the common failure is subtracting an hour of sleep to fit a doubled day in. That is a bad trade: the session gains a stimulus and loses the conditions under which stimulus becomes adaptation.

Sleep degrades before performance does. In practice sleep score is often the first of the four fatigue signals to move, and it moves a week or so before the strength trend does. It is a leading indicator, and treating it as one is the useful way to use this screen.

What the screen will not do

It will not show a score when there is no data. "No sleep data for last night", not a zero. Every missing input in the app behaves this way, and the reason is that one fabricated zero teaches you every other number might be decorative.

It will not diagnose a sleep disorder. Persistent low efficiency, heavy fragmentation or unusual respiratory rate patterns are worth taking to a doctor. This is a training log, not a medical device.

It will not give you a percentile. Why refusing a number is also showing the math.

What to do next

If your score is consistently low, look at which of the three components is dragging it. Low duration is a schedule problem. Low efficiency is usually an environment or timing problem — caffeine, alcohol, room temperature, screen exposure. Low consistency is the easiest to fix and the one people ignore.

The recovery screen is where this score becomes 20% of a decision about today's training.