From the Coach guide
Your hardest runs show what you can do now, and how that is changing.
Coach finds the fastest 1 mi, 2 mi or 5 km inside each run that was above easy effort. It converts each one to a 5K time.
Coach compares your 2 best efforts from the last 4 weeks with the 4 weeks before. A change counts only when it is larger than your own usual spread and at least 2%.
A best effort shows what you did, not your limit. Training runs read slower than races. Hills, heat and GPS errors make them slower too.
If you had no hard run in the last 4 weeks, Fitness shows your last reading. A gap is not a decline.
Coach also compares your pace at a given heart rate on easy runs. It changes a lot from run to run, so it only supports the reading. It decides only when you have no hard runs.
A threshold test or a race is a direct measurement. When you have one, it replaces the best-effort reading.
Apple's Cardio Fitness is a VO2max estimate from your Apple Watch, shown here as a slow, 28-day-smoothed trend — it never sets a pace, zone, VDOT or target. Independent lab comparisons found it can underestimate true VO2max by around 6 mL/kg/min, with a typical error of about 7. The estimate also depends on the watch and its software. Heat, dehydration, caffeine and altitude all tend to push a single reading low, which is why only a change that holds for two consecutive 28-day windows is ever surfaced, and even then only as a suggestion to retest sooner.
Race-performance equations
Daniels and Gilbert built equations that rate a performance from its distance and time. They make a 1 mi, a 2 mi and a 5 km effort comparable.
The equations come from races. A training run that was not flat out reads lower than a race would. Coach uses the trend, not the number alone.
Test–retest reliability study
Trained runners repeated a track test of critical speed within 2.0%, and within 1.3% once they knew the test. Coach counts a change only when it is at least 2% and larger than the spread of your own best efforts.
The study tested 10 trained men on a track, not training runs. Your own spread is usually larger, and then Coach uses it instead.
Monitoring research; app-defined trend
Buchheit’s review discusses exercise heart rate as one way to monitor training responses, interpreted alongside training context and measurement variability.
Coach’s speed-to-heart-rate ratio, four-week comparison, and change threshold are an implementation choice. This review does not validate them as a fitness test or VO₂ max estimate.
Independent lab validation studies
Lambe and colleagues compared Apple Watch Cardio Fitness against lab-measured VO₂max in two separate cohorts and found a consistent underestimate, with a mean error of roughly 6 mL/kg/min and typical absolute error around 7.
These studies measure Apple's algorithm's accuracy against a lab standard; they say nothing about Coach's 28-day-median, two-window flagging rule, which is an app-chosen display threshold, not a validated one. Accuracy can differ between watch models and software versions.
Coach