Methodology

How your scores are calculated

Apex Trace's Sleep, Readiness, Cardio Load, Calm, and Body Age numbers are estimates, computed entirely on your device from your own data. Nothing is sent anywhere to produce them — there is no server, and no black box. This page shows exactly how each one is figured out.

Fitbit's Premium scores aren't available through the Google Health API, so Apex Trace builds its own from the raw signals your device records — sleep stages, heart rate, heart-rate variability (HRV), resting heart rate (RHR), breathing rate, and skin temperature. Every calculation below runs locally on your phone.

Two ideas show up throughout:

A baseline is simply your own recent normal — the average of a metric over your previous 30 days (we need at least 7 days of data; with less, any score that depends on it gracefully steps aside rather than guessing).
A z-score measures how far today sits from that baseline, in units of your own day-to-day variation. Zero means "a typical day for you"; positive means above your normal, negative means below.
When an input is missing (for example, no HRV was recorded), it simply drops out and the remaining inputs are re-weighted so the score still makes sense.

01 Sleep Score (0–100)

The Sleep Score blends three things about last night, then scales the result to 0–100:

Duration — 50%. How much you slept versus an 8-hour target, capped at 100%.
Deep + REM — 30%. The share of your sleep spent in the restorative deep and REM stages, measured against a healthy 45% of total sleep and capped at 100%.
Restfulness — 20%. Starts at full marks and is reduced for time spent awake during the night and for low sleep efficiency (time asleep versus time in bed).

duration (50%): min(time asleep / 8h, 1.0) deep+rem (30%): min((deep + REM) / (45% of sleep), 1.0) restful (20%): starts at 1.0, minus awake-time and low-efficiency penalties Sleep Score = round(100 × weighted sum of the three)

02 Readiness (0–100)

Readiness estimates how recovered you are this morning by comparing four signals to your own baseline:

HRV — 30%. Higher heart-rate variability than your normal nudges the score up; lower nudges it down.
Resting heart rate — 20%. A lower RHR than your normal is better, so the direction is flipped.
Last night's sleep — 35%. Your Sleep Score from section 1, on a 0–1 scale.
Recent load — 15%. Yesterday's Cardio Load relative to your target; training well past your target eases this component down.

HRV score (30%): centered at 0.5, ± your HRV z-score RHR score (20%): centered at 0.5, ∓ your RHR z-score (lower = better) sleep score (35%): last night's Sleep Score ÷ 100 load score (15%): eased down when yesterday's load exceeds target Readiness = round(100 × weighted sum) label = ≥70 "go for it" · ≥40 "steady" · else "take it easy"

The "biggest factor" note you sometimes see (e.g. "you slept 1h 47m less than usual") is just the single input that fell furthest below your baseline that day.

03 Cardio Load (today vs. target)

Cardio Load turns your day's effort into one number by weighting time spent in each heart-rate zone — harder zones count for more:

Fat Burn minutes count ×1, Cardio ×2, Peak ×3, summed across the day's workouts and activity.
Your target is the average daily load over your previous 28 days — so "37 / 35" means today edged just past your recent norm.

daily load = Σ zone minutes × (Fat Burn=1, Cardio=2, Peak=3) load target = average daily load over the previous 28 days

04 Calm

Calm is the recovery axis on the You-screen radar. It reads three overnight signals against your baseline and centers a typical night at the midpoint:

Breathing rate — a slower-than-normal rate reads as calmer (inverted).
HRV — higher than normal reads as calmer.
Skin-temperature deviation — closer to your normal reads as calmer (inverted).

If any of the three wasn't recorded, it drops out and the others are re-weighted, so a missing sensor never zeroes the score.

Calm = clamp 0–1 of: 0.5 ∓ breathing-rate z-score × 0.10 (slower = calmer) ± HRV z-score × 0.10 ∓ skin-temp-deviation z-score × 0.05 (missing inputs drop out, weights re-normalized)

05 Body Age (physiological-age estimate)

Body Age estimates how old your body is physiologically — specifically, how your cardio fitness compares to the US population median for your age group. The core question it answers: if a typical person had your fitness level, how old would they be? A fitness age lower than your real age means your cardio system is ahead of the curve; higher means there's room to improve.

Everything is computed on your device. If you choose to enter body measurements (date of birth, sex, height, weight, and optionally waist circumference), those measurements are stored only on your phone and never transmitted anywhere. No exercise test is needed.

VO₂max estimate

The calculation starts by estimating your VO₂max — a standard measure of how efficiently your body uses oxygen, and the most widely used single marker of cardio fitness. Two published non-exercise regression models are used depending on whether you've entered a waist measurement:

With waist circumference (more accurate) — Nes et al. 2011, sex-specific model from the HUNT Study:

Men: VO₂max = 100.27 − 0.296·age + 0.226·PA − 0.369·waist(cm) − 0.155·restingHR Women: VO₂max = 74.736 − 0.247·age + 0.198·PA − 0.259·waist(cm) − 0.114·restingHR

Without waist circumference (BMI fallback) — Jackson et al. 1990:

VO₂max = 56.363 + 1.921·PA − 0.381·age − 0.754·BMI + 10.987·sex (male = 1, female = 0) BMI = weight(kg) / height(m)²

Adding your waist measurement improves accuracy over the BMI fallback — the app will prompt you to add it if it's missing. In both models:

PA (physical-activity index, 1–5) is derived on-device from your weekly active-zone-minutes and steps from your Fitbit data.
Resting HR is your 30-day baseline mean (the same baseline used for Readiness, section 02).

Fitness age and delta

With an estimated VO₂max in hand, the app asks: at what chronological age does the typical person have this VO₂max? It answers that by inverting the US FRIEND/ACSM reference VO₂max-by-age curve — a piecewise-linear median curve built from a large US fitness database (Kaminsky et al., Mayo Clin Proc 2015). The US reference is intentional: it reflects the population the app primarily serves. The result is clamped to the range 20–80 years.

fitness age = age at which the FRIEND/ACSM median VO₂max equals your VO₂max (piecewise-linear inversion, clamped 20–80 years) delta = fitness age − your chronological age (negative = biologically younger; display-clamped to ±15 years)

Pace of Aging

Pace of Aging shows the direction your fitness age is moving, not just where it stands today. It compares how much your fitness age changed in the past 30 days against how much it changed in the 60 days before that, and expresses the ratio centered on 1.0:

Pace of Aging = (fitness-age change over last 30 days) ÷ (fitness-age change over the preceding 60 days) centered on 1.0: >1 → aging faster than your own baseline 1 → same rate as your baseline <1 → slowing down relative to your baseline

Because this ratio depends on trends over time, it is most meaningful after several months of data. When fewer than 14 days of VO₂max history are available, the gauge is hidden rather than showing a noisy early reading. It improves as your history grows.

Sources

Nes BM et al. "Estimating V̇O₂peak from a non-exercise prediction model: the HUNT Study." Med Sci Sports Exerc 2011;43(11):2024–2030.
Jackson AS et al. "Prediction of functional aerobic capacity without exercise testing." Med Sci Sports Exerc 1990;22(6):863–870.
Kaminsky LA et al. "The importance of cardiorespiratory fitness in the United States: a call to action from the American Heart Association." Mayo Clin Proc 2015;90(2):167–172. (FRIEND/ACSM US reference curve.)

06 What these numbers are — and aren't

These scores are estimates for general wellness, meant to help you spot your own trends. They are computed on your device from your own data and are not a medical measurement, diagnosis, or advice. They're also not the same as Fitbit's Premium scores, which use Fitbit's own private formulas. If a number ever looks off, the raw data behind it lives right alongside it in the app so you can check.

The above mirrors the scoring used in the app. For how your data is handled, see the Privacy Policy.

Get started

See these scores on your own data.

Private. On-device. No subscription.