Most fitness metrics answer the wrong question. They tell you how you compare to other people, when the only comparison that predicts your day is how you compare to yourself. This page explains the idea Helix is built on, and what the confidence label is really telling you.
Why population norms mislead
Take a resting heart rate of 58 bpm. Against population tables, that is a good number: comfortably below average, the kind of value a chart colors green. But for a trained runner whose normal is 47, waking at 58 is a red flag. Something, a hard block, a bad night, an oncoming illness, is loading the system. The same reading is reassuring for one person and a warning for another, and a population chart cannot tell you which one you are.
The same is true of HRV, respiratory rate and sleep need. These signals vary enormously between people and are remarkably stable within one person, which is exactly the shape of data where “normal for you” carries information and “normal for humans” mostly doesn’t.
Rolling baselines: you vs. you
So Helix keeps a rolling personal baseline for every overnight signal it reads: HRV, resting heart rate, respiratory rate, wrist temperature, blood oxygen, sleep. Each baseline is built from your own recent history, typically about the last two months of you, and each morning’s readings are scored as deviations from it. The question is never “is 58 a good resting heart rate” but “is 58 normal for you, lately”.
Because the window rolls, the baseline is a picture of who you are now, not who you were when you installed the app. Get fitter over a season and your baseline quietly moves with you; the score keeps measuring you against your current self.
The outcome loop
Baselines make the inputs personal. The outcome loop makes the weighting personal. When you rate how a day felt, Helix compares that rating against the morning’s inputs and gradually re-weights them. If your rough days follow HRV dips but shrug off short sleep, HRV slowly earns more of your score. Two people with identical overnight data can wake to different scores, and both scores are right, because they are answering different questions: one for each body.
What the confidence label means
A baseline built on three nights is honest data, thinly sampled. So while your history is short, the score carries a visible confidence indicator. It is not an apology; it is a unit of trust. A score with low confidence still points the right way, it just should not settle arguments yet.
Confidence strengthens quickly across the first days as the window fills, is largely settled within the first weeks, and the personalization keeps refining well beyond that as the outcome loop accumulates rated days. You never have to manage any of this; you just watch the label fade in importance.
What weakens a baseline, and what doesn’t
- Long gaps in data weaken it. A stretch of nights without the watch leaves the window thinner, and confidence eases back until fresh nights refill it.
- A new watch is fine. Baselines belong to your body, not the device. Upgrade and your history carries straight over; a newer watch may simply add signals, like wrist temperature, that start building their own baselines.
- Illness stays visible. A rough week is part of your history, not an outlier to be scrubbed. Helix does not delete it; the rolling window simply carries you past it, and in the meantime those red mornings are the score doing its job.
Why the morning verdict is frozen
One more design choice belongs to this philosophy. Once your night is complete and scored, the morning verdict does not drift for the rest of the day. New afternoon data changes today’s strain, never this morning’s recovery. A readiness number that keeps moving after you have made your training decision is a number you learn to ignore, and a score you ignore measures nothing. Frozen is what lets you act on it.
To see all of this in motion from day one, read your first morning; for what the score does with these baselines, see the recovery score.