By Raji Shah ยท Updated 2026-10-24
GPS accuracy isn't one number. The same watch can be close to perfect on an open road, read several percent short among tall buildings or under trees, and read long on a track or a race course. Knowing which way the errors run in each setting, and which settings and habits reduce them, matters more than any accuracy claim on a spec sheet.
How a watch measures distance
A GPS watch records your position again and again during a run and adds up the straight lines between the points. Every position has some error, a few meters in good conditions and more when buildings, trees or your own body block the signal. Two effects pull the total in opposite directions. Random scatter in the points tends to make a track zigzag, which adds distance: one mathematical analysis showed that the distance between recorded points is, on average, longer than the true distance. Smoothing, filtering and points dropped in poor reception tend to cut corners, which subtracts distance. Which effect wins depends on where you run.
What the research shows
| Setting | What studies found |
|---|---|
| Open road race course | Watches read slightly long: in a marathon on a mostly open course, finishers' watches averaged 42.33 km for 42.2 km, and 21.18 km in the half |
| City center with tall buildings | Mostly short, by up to about 9% in a validation study of eight watches |
| Forest | Mostly short, similar to city centers |
| Open running track | Long, by about 1 to 4% |
| Long road race (56 km) | Median errors of about 0.6 to 1.9% across categories of watch |
The eight-watch study, which covered models from Apple, COROS, Garmin, Polar and Suunto tested in 2019, found average errors of 3.2 to 6.1% across all settings, and only 44 to 81% of each watch's recordings fell within 5% of the true distance. Running made the errors worse than walking or cycling for six of the eight watches. The authors suggested that some watches' automatic corrections may inflate distances in open areas. That study used one runner, short distances and settings without multi-band GPS, so treat its numbers as a guide to the pattern rather than a ranking.
Watches also beat phones. In the marathon study, the average error in the half marathon was 0.6% for sport watches and 1.7% for phones running a tracking app.
Why your race distance reads long
Certified road courses are measured along the shortest possible route a runner could take, plus a safety margin of 1 m per kilometer, and almost nobody runs that line. Add GPS error on top and most watches finish a marathon showing more than 42.2 km, even on a correctly measured course. Our race-day wearable checklist explains how to pace from the course markers instead.
Does multi-band GPS help?
Multi-band (or dual-frequency) GPS adds a second, newer signal frequency (L5 on GPS satellites that broadcast it) to the standard one. Garmin says this improves positioning and reduces errors from reflected signals and from the atmosphere in challenging places, such as among tall buildings and under trees. Garmin offers it on higher-end models such as the Forerunner 570 and 970 and the fenix 8, but not on budget models such as the Forerunner 70 and 170. Those models also have an Auto Select mode that uses Garmin's SatIQ technology to choose the best multi-band setting while prioritizing battery life. Apple offers dual-frequency GPS only on its Ultra models; the Series 12 uses single-band L1 GPS. COROS offers a dual-frequency mode on many models, such as the PACE 3 and PACE 4, but not on all of them.
The peer-reviewed evidence for watches is thin: the watch studies above all collected their data between 2016 and 2019, from watches without multi-band GPS. The closest evidence is a study of phones standing still under forest canopy, where dual-frequency phones were more accurate on average, but the study cautioned that accuracy shouldn't be generalized from the number of frequencies or satellite systems a device uses. What is certain is the battery cost. Garmin rates the Forerunner 970 at up to 26 hours with GPS Only, 23 with Auto Select and 21 with All + Multi-Band, and COROS says its dual-frequency mode uses much more battery than its other settings. Use multi-band where it can help, in cities and under trees, if your battery allows it.
Why instant pace jumps around
Instant pace reflects only the last few seconds of data, so a few meters of position error becomes a big swing in pace. That is why it can show 4:30 per kilometer one moment and 5:30 the next on a steady run, especially downtown. Lap pace and average pace smooth the error out over a longer distance and are much better guides during a run. On a track, use your watch's track feature and set your lane: Apple Watch detects when you arrive at a track and asks for your lane, Garmin's Track Run activity takes a lane number for a standard 400 m track, and COROS's Track Run mode asks for the track length and your lane.
Settings and habits that help
- Get a fix before you start. Garmin says finding satellites can take 30 to 60 seconds in an open area, and suggests standing still for a few minutes if reception is poor; syncing the watch downloads satellite data that speeds this up.
- Use the right satellite mode. All Systems or multi-band in cities and under trees, GPS only when battery matters most. Avoid battery modes such as Garmin's UltraTrac for races or workouts you want to analyze; Garmin says it decreases the quality of recorded activities.
- Pace by lap or average pace, not instant pace.
- Use track mode on a track and choose your lane, rather than relying on GPS around tight bends.
- Keep the watch synced and updated, so its satellite data and software stay current.
How to judge your watch's GPS
Don't judge it on one run. Compare it over several runs against known distances: lane 1 of a standard 400 m track (as a normal run, not in track mode), a certified course, or a measured route you run often. Look at the map trace as well as the total; a trace that cuts corners or wanders off the road shows where errors come from. Expect it to read a little long on open courses and short among tall buildings or dense trees, and ask whether the error would actually change a training decision. A 1% difference on an easy run rarely does.
Bottom line
In race studies on open roads, GPS watches were typically within about 2% of the course distance. Among tall buildings and trees they often read several percent short, and on tracks up to about 4% long, so the errors run in different directions in different places. Pace from laps rather than instant pace, use multi-band where it helps and your battery allows, and judge distance by known courses rather than by your watch alone.
FAQ
How accurate is GPS on a running watch?
Typically within about 2% on open roads in race studies, where watches usually read slightly long. Among tall buildings or under dense trees, errors of several percent are common, usually reading short; one study found some watches up to about 9% short in those settings.
Why does my watch read a different distance than the race?
Race courses are measured along the shortest possible line plus a small safety margin, and runners rarely follow that line. Together with GPS error, that usually makes a watch read long; in one marathon study, finishers' watches averaged 42.33 km for the 42.2 km race.
Is multi-band GPS worth it for running?
It can help in cities and under trees, but published evidence for watches is thin and it costs battery: Garmin rates the Forerunner 970 at 21 hours with All + Multi-Band against 26 with GPS only. It matters less on open roads.
Why does my pace jump around during a run?
Instant pace uses only a few seconds of GPS positions, so small position errors become large pace swings, especially among buildings. Use lap pace or average pace instead.
Is a phone as accurate as a GPS watch?
Usually not. In a half-marathon study, the average distance error was 0.6% for sport watches and 1.7% for phones running a tracking app.
How should I measure distance on a track?
Use your watch's track mode if it has one and set your lane. Apple Watch can detect a track and ask for your lane, and Garmin's and COROS's Track Run modes have a lane setting.