How accurate is your running watch?
Trust a chest strap for heart rate and treat most other watch numbers as trends. GPS distance is typically off by 3–6%, watch VO2max estimates by roughly 4–16%, and watch lactate-threshold pace by about 26%. Watch threshold heart rate is a reasonable starting point.
What to do
- Use a chest strap if heart rate matters for the session, especially intervals and hills.
- Read watch VO2max as a trend over months, not as a lab result.
- Ignore your watch's lactate-threshold pace. Use its threshold heart rate only as a first estimate.
- Expect GPS distance to be a few percent off, and more in cities and forests.
- Don't change training or fuelling based on glucose monitors or skin-based core-temperature sensors.
Heart rate: chest beats wrist
A validation study found that chest straps measure heart rate within about 1–3%. Wrist sensors got worse at higher speeds, and accuracy varied by model (OSU validation).
For easy, steady running, a wrist sensor is often good enough. For intervals and hard efforts, a chest strap is the safer choice. Even an accurate heart-rate reading can mislead on short reps. In highly trained runners, 30-second reps showed more time at high heart rate than 3-minute reps, yet gave less time near VO2max (Fleckenstein and colleagues, 2024). See which VO2max intervals work.
GPS distance
GPS distance errs by 3–6%, and by up to about 9% in cities and forests (Gilgen-Ammann and colleagues, 2020).
An example of what that means: 3% of a marathon (42.2 km) is about 1.3 km, and 6% is about 2.5 km. Your watch may show a race as longer or shorter than the measured course. Our suggestion: on race day, pace by the official distance markers rather than the watch’s lap beeps.
VO2max estimates
VO2max is the most oxygen your body can use per minute. Many watches show an estimate of it.
- Garmin: close for moderately trained users, with a mean absolute percentage error (MAPE, the average size of the miss) of 2.8–4.1%. But it underestimated highly trained users by about 6 mL/kg/min (Garmin validation, 2025).
- Apple Watch: missed by about 16% in one study (JMIR study, 2024).
To make that concrete, a 16% miss on a true VO2max of 50 is 8 points, so the watch could show anything from about 42 to 58. That is an example calculation, not a finding.
These are single studies of specific devices. They tell you how those devices did in those tests, not which brand is best. The safest use for any watch VO2max is as a trend: is it going up or down over months?
Threshold estimates
Some watches estimate your lactate threshold, the point above which you tire within minutes. In one validation, the watch-estimated threshold heart rate was usable, but the threshold pace was overestimated by about 26% (Frontiers study, 2025).
Use the heart rate as a starting point and check it with the talk test. Ignore the pace. See how to set your heart-rate zones.
HRV and readiness scores
Nightly heart-rate variability (HRV) can be measured well by some devices. Against an ECG, the Oura Gen 4 ring and WHOOP strap agreed closely, with concordance correlations of 0.99 and 0.94 (2025 validation study). No proprietary readiness score has been validated as a decision tool. See does HRV-guided training work?
Glucose and core-temperature sensors
Two newer gadgets have little decision value for runners:
- Continuous glucose monitors (CGMs) in people without diabetes have no validated value for training or fuelling decisions (GSSI overview). Note that this overview comes from an institute funded by PepsiCo/Gatorade. During races, glucose readings also tend to run high (Kajisa and colleagues, 2026).
- Skin-based core-temperature sensors also lack validated decision value (CORE validation).
Look at the numbers if you are curious, but don’t let them set your training or your race fuelling. For fuelling, see how many carbs per hour.
Race-time predictions
Watch race predictions tend to be optimistic. See can you trust marathon time predictors?
A trust table
The research report behind this site sorts watch data into quality tiers. The tiers are graded B overall; the last row is graded C.
| Measure | How far to trust it |
|---|---|
| Chest-strap heart rate | High |
| Wrist heart rate | Medium on easy, steady runs. Low on intervals, hills and in the cold |
| GPS distance | Medium (3–6% error, worse in cities and forests) |
| Running power | Depends on the device. Start a new baseline if you change device |
| Watch VO2max | Trend only |
| Watch lactate-threshold pace | Don’t use |
| Watch lactate-threshold heart rate | A starting estimate |
| Glucose monitor, skin core temperature | Display only, no decisions |
The “low in the cold” and power rows are the report’s practical rules, not separate findings from the studies cited here.
What we don’t know
- Devices change fast. New models and software updates can be better or worse than the versions tested.
- Most findings come from single studies with specific devices and specific groups of people.
- Running power was not covered by the validations we cite.
- Highly trained runners may be misjudged more. The Garmin study found larger errors in the fittest users; other groups may differ too.
Sources
- OSU validation. Garmin and Polar heart-rate accuracy.
- Gilgen-Ammann et al. (2020). GPS distance accuracy of sports watches.
- Garmin validation (2025). Accuracy of Garmin VO2max estimates.
- JMIR study (2024). Accuracy of Apple Watch VO2max estimates.
- Frontiers study (2025). Validation of watch-estimated lactate threshold.
- 2025 validation study. Nightly HRV from consumer wearables compared with ECG.
- AthleteData guide. Comparison of wearable readiness scores.
- GSSI overview. Continuous glucose monitoring in athletes without diabetes.
- Kajisa et al. (2026). Glucose monitoring during endurance races.
- CORE validation. Validation of a skin-based core-temperature sensor.
- Fleckenstein et al. (2024). Time near VO2max in short vs long intervals in highly trained runners.
Last reviewed 13 November 2026.