Metric definitions
Measurements, estimates and missing data
Analysis preserves original activity records and frozen facts. Viewing reports does not rewrite distance, duration, source or raw events. Fitness, load, energy and charts are derived views; applicable results may change with calibration or profile updates.
Real zeros participate in totals. Null means unavailable and remains a gap or -- in the interface. Indoor stair climbing does not create a running distance. A merged activity and its constituent records are not counted twice. Coverage counts identify incomplete distance or elevation totals.
Heart-rate zones and load
Displayed zones follow the account heart-rate reference: heart-rate reserve (HRR) can be used with resting heart rate, otherwise maximum-heart-rate percentages apply. Z1–Z5 start at 50% with 10% boundaries. Valid samples below Z1 contribute to coverage but not to those five zones.
Edwards load always uses percentages of maximum heart rate. HRR zones must not be substituted. Minutes in the five bands receive weights 1 through 5; values below 50% contribute no band load. Daily charts integrate observed samples and do not invent heart rates for uncovered time.
Session RPE equals active minutes multiplied by a recorded effort rating from 1 to 10. Missing feedback does not create sRPE. The activity recovery model chooses sRPE, supported heart-rate evidence, or duration and terrain. Recovery duration and load in AU are distinct quantities.
Chronic load, acute load and balance
Daily heart-rate load requires at least 80% coverage for each activity, with no more than 5% of observed heart-rate time above the adopted maximum-heart-rate reference. If any activity lacks sufficient evidence, that day's load is unknown. Only days without activities receive zero.
CTL and ATL use exponential smoothing with 42-day and 7-day constants:
L_today = L_yesterday + (load_today - L_yesterday) × (1 - exp(-1 / days))
Balance is same-day CTL minus ATL. Up to 180 preceding days warm up the selected interval. Missing load breaks the series; the next available stretch starts a new baseline. The interface shows available history and flags fewer than 42 days. These describe the recorded workload, not injury probability, guaranteed performance or a medical diagnosis. Identically named metrics on other platforms may use different conventions.
Pace, elevation and energy
Average pace is saved active duration divided by distance. Grade-adjusted pace (GAP) and active energy use the same terrain and sport models as the phone. Energy details identify the weight source: recorded weight takes precedence, with current profile weight used when applicable.
Climbs use a smoothed elevation profile at 25-metre intervals within continuous route segments. A significant climb spans at least 300 metres, gains 30 metres and averages 3% grade. A cumulative 15-metre descent confirms the summit. A recording ending uphill may have an unconfirmed summit. Missing elevation or a route outside analysis limits does not produce a fabricated climb.
Efficiency and aerobic decoupling
Efficiency is speed in m/s divided by heart rate in bpm. Eligible two-minute windows need at least 90% GPS and heart-rate coverage, speed variation no greater than 10%, and absolute grade no greater than 1%.
Decoupling requires at least 40 continuous minutes of eligible windows. It compares their first and second halves as (early efficiency - late efficiency) / early efficiency × 100%. Positive values mean lower efficiency later. Heat, terrain, fatigue and sensor error can affect the result; insufficient evidence produces no value.
Running ability and cardio fitness
Running ability uses eligible running and race evidence within its observation window, with equivalent distance times. Estimates and observed times are shown separately. Cardio fitness distinguishes HealthKit, workout estimates and calibration; VO₂max is not a race-performance score.
Gait, in-workout HRV, training effects and recovery estimates retain their measurement or training-analysis basis. In-workout HRV cannot establish overnight readiness. Fitness uses the selected end date; current recovery uses the present time.
Continuous efforts and peaks
Distance efforts cover 400 metres through 100 kilometres within continuous GPS segments, without crossing pauses or gaps. Running, trail running, walking and hiking exclude a net descent exceeding 0.5% of the target distance; cycling is compared separately and retains descents. Continuous edges are limited to 10 m/s for foot sports and 35 m/s for cycling. Intervals over 60 seconds, excessive speed or horizontal accuracy worse than 50 metres break a window. Unknown elevation remains unknown and does not certify a level course.
Elevation records find the largest cumulative positive gain over each continuous distance on the same 25-metre smoothed profile, with interpolated endpoints. Missing elevations break windows. These differ from whole-activity ascent and are separated by sport.
Heart-rate and power peaks span 5 seconds through 2 hours. They use time-weighted integration over completely covered windows, not the largest isolated sample. Road running, trail running and other sports are compared separately. Annual curves contain only supported distances or durations within the selected period.
Updated