Learn

readiness = 62 + 11·HRV − 12·RHR + 15·sleep
· HRV: your HRV trend vs baseline, in units of your own night-to-night variability. A single wrist reading is noisy, so it's smoothed into a trend — and last night is trusted only as far as your resting heart rate agrees with it: genuine recovery shows higher HRV AND lower resting HR together, so a lone HRV blip barely counts, but a corroborated one moves the score
· RHR: last night's resting HR vs baseline, in the same personal-variability units (a steadier, harder-to-fool signal, so one night counts)
· sleep: a blend of last night and your past week, quality-aware — one short night only lightly dents it
Small deviations are treated as noise and ignored (the "smallest worthwhile change").

• Don't chase the number — improve its inputs: consistent sleep, and the habits that lift HRV and lower resting HR over weeks.
• Respect low scores occasionally: an easy day when readiness tanks often buys back two good ones.
• A single low night barely moves it by design — a sustained dip across several days is the real signal. Look for repeat offenders in your logs: late caffeine, alcohol, late meals, under-sleeping.

We take your wearable's estimated VO₂max (its cardio-fitness number), smooth it over a couple of weeks, and find the age at which that value is typical, using published age×sex norm tables. That age is your cardio age.
· Resting heart rate and HRV can nudge it by a year or two (a low resting HR reads younger); the nudge is capped so noise can't swing it.
· Sleep and steps are shown as context but never fed into the number — that would double-count.
· If we don't know your sex, we compute both norms and only call you "ahead" or "behind" when it holds either way, otherwise "on track". Set your birth year and sex in Settings for a precise number.

• Easy "zone 2" cardio (conversational pace) a few times a week is the biggest lever on VO₂max.
• Occasional higher-intensity intervals raise the ceiling faster.
• It moves slowly — think weeks and months, not day to day.
• Treat it as a fitness estimate, not a verdict: wrist-derived VO₂max is approximate, and this is not a biological-age or longevity measure.

Energy starts the day at your readiness score, then every 5 minutes:
· drains gently just from being awake, and faster when your heart rate climbs well above resting (your effort/stress proxy)
· charges a little when your heart rate is low and you're genuinely at rest (a nap recovers more)
· a logged coffee or mate adds a small, temporary "alertness" lift that fades over a few hours
Only real overnight sleep refills it to full — that's tomorrow's fresh start.

• Protect the start: Energy begins at your readiness, so the real lever is sleep and recovery, not the gauge itself.
• Spend it deliberately — a hard session is *supposed* to drain it; that's the point, not a failure.
• Genuine rest (a real sit-down, a short nap) charges it back; scrolling while your heart rate stays up doesn't.
• Caffeine's lift is borrowed alertness, not recovery — the gauge shows it as a small, fading bump on purpose.

• Consistent aerobic exercise (zone 2 especially) lowers it over months — the classic training adaptation.
• Alcohol is the most reliable way to spike it overnight; hydration and cool bedrooms help it fall.
• Track the trend after illness — it returning to baseline is a decent "fully recovered" indicator.

• Sleep is the #1 lever — both duration and regularity of bed/wake times.
• Alcohol is the #1 destroyer: even two drinks measurably suppress HRV for a night or two.
• Zone-2 cardio, breathing exercises (slow exhale-biased breathing), and managing late-evening caffeine all show measurable effects.
• Don't judge single days — react to 3–5 day trends vs baseline.

• Zone-2 cardio (conversational pace, 30–45 min) done consistently is the classic VO2max builder; add occasional harder intervals once it's a habit.
• It moves on a scale of weeks to months, not days — don't read single-day wiggles as change.
• Because there's no ceiling in the mortality data, going from 'fair' to 'good' is worth it even if you'll never be 'excellent'.

duration    d = full across 7–9h asleep, tapering outside it
restorative r = min(1, (deep+REM) ÷ asleep ÷ 0.40) — vs a 40% ideal
continuity  c = (efficiency − 60) ÷ 35, clamped 0–1

score = 95 × (0.5·d + 0.3·r + 0.2·c)²      (min 15, max 95)

Duration is credited on a curve, not a hard 8h target: the clinical guideline is "7+ hours", so anything 7–9h earns full duration credit, 6–7h costs a little, and it falls off faster below 5h. This stops a short-but-efficient night from being scored as harshly as a genuinely poor one.

worked example — a real short night:
  asleep 5h38 → d 0.81 (below the 7h band) · efficiency 97% → c 1.0
  deep 1h02 + REM 1h07 = 38% restorative → r 0.95
  base = 0.5×0.81 + 0.3×0.95 + 0.2×1.0 = 0.89 → 95 × 0.89² ≈ 76

• Duration is half the score — the biggest lever is protecting a 7h+ window; past 7h the gains flatten.
• The restorative share responds to the usual levers: regular schedule, no late alcohol, caffeine early only.
• REM concentrates in the final cycles, so cutting a night short cuts the restorative share too — the components compound.
• Compare against efficiency: long night + low score = architecture problem; high efficiency + low score = too short.

fell asleep in  = minutes from getting into bed to the first sleep stage
awakenings      = distinct awake periods AFTER onset, excluding the final wake-up
awake after onset = total minutes of those interruptions (WASO)
restorative     = deep + REM minutes combined
longest stretch = the single longest unbroken run of any sleep stage

• Consolidation responds to the basics: consistent bed/wake times, a cool dark room, no late alcohol (it fragments the second half of the night).
• A long 'fell asleep in' plus early awakenings often points at caffeine or screen time too close to bed — log them and watch your own pattern.
• Judge the weekly shape, not one night: everyone has the occasional broken night.

• Regularity beats duration: same bedtime ±30 min moves every downstream metric.
• Caffeine has a ~5–6 h half-life — a 6pm mate is still ~half-active at midnight. Log it and watch your own correlation.
• Alcohol near bedtime suppresses REM in the first half and fragments the second half — visible in your hypnogram the next morning.
• Cool, dark room; consistent wind-down; no doomscrolling in bed (light + arousal both delay deep sleep).

• Side-sleeping reduces positional breathing disturbances for many people.
• Alcohol before bed relaxes airway muscles — a common cause of worse overnight SpO2.
• Treat congestion; consider a professional sleep study if lows persist — this one is a signal to escalate, not self-optimize.

• Log every rescue-inhaler puff (💨 on the quick-log strip). International asthma guidance (GINA) treats needing a reliever more than twice a week as a sign of poorer control, and steadily rising use as a reason to see your doctor — frequency over time is the single best self-tracking signal, far more validated than any wearable vital.
• Cold-air mitigation has real trial evidence: pre-warming the air you breathe with a scarf or heat-exchange mask before going out in the cold cut airway narrowing about as well as a pre-emptive puff of albuterol in studies — directly relevant to warm-home-to-cold-street winter swings.
• Indoor air: the strong evidence is about avoiding damp and mould (clearly linked to asthma), not chasing humidity — keep it moderate (~40–50%), not aggressively humid.
• This is a watch-and-discuss metric: consumer overnight SpO2 in particular is unreliable on the wrist, so bring the trends to a doctor rather than acting on a single reading or adjusting medication yourself.

• There's no "improving" temperature — it's a messenger. Keep the bedroom cool and consistent so the signal stays clean.
• When it spikes with a high RHR: hydrate, skip intensity, and watch for symptoms.

baseline  = median HR of your awake 5-min windows
episode   = ≥2 consecutive windows above max(baseline+15, baseline×1.22)
            (single quiet windows in between are bridged)
walk      = ≥15 min at ≥60 steps/min, merged into overlapping episodes
suggestion rank = frequency prior × hour-of-day match × duration match
                  × HR match × step-density match  (from your history)

• Label honestly and early — the first ~5 confirmations do most of the teaching.
• Use "split walk + burst…" on mixed outings so each activity learns its true HR/steps signature.
• Dismiss what isn't an activity (stress spike, hot shower) — dismissals prevent bad suggestions later.
• Deleting an activity log puts its episode back up for labeling; the learned profile keeps history.

• One tap at the moment of consumption beats reconstructing the day later.
• Log the misses too (the late espresso, the third beer) — the "bad" data points are the informative ones.
• Use ＋ Custom to backfill anything you forgot — timestamps matter more than completeness.