The evidence scoreboard

GPS accuracy, device by device

One good study exists, it is six years old, and every watch in it has been replaced. Here is the record anyway, including every place the record is empty.

By Stephen V.Published August 11, 2026

#ad

We earn a commission when you buy through our Amazon links, at no extra cost to you. It never decides a ranking — the rubric that does is published in full, so you can check us rather than take our word for it. How this works.

The short answer

The only peer-reviewed multi-watch distance validation we found measured 3.2-6.1% error across eight sport watches against a trundle wheel, with only Polar's receivers under 5%. Every watch tested was single-band and none is still sold. No current-generation multi-band watch has been independently validated that we could find.

The one study, and what it actually did

Gilgen-Ammann, Schweizer and Wyss (2020) measured eight positioning-enabled sport watches against a trundle wheel accurate to one centimeter, across urban, forest and 400 m track conditions. Distance errors ranged from 3.2% to 6.1%, and only Polar's receivers stayed under 5% in every condition.

It is a good study and it is the best one in existence for this question. It is also from 2020, and every watch in it used a single-frequency receiver — the dual-frequency hardware now standard on mid-range watches did not exist in consumer form.

So the correct reading is: this is evidence about brand engineering history, not about any watch you can buy. We cite it constantly and we never let it stand in for a measurement of a current device.

Why the blanks matter more than the numbers

The table above has more "no validation" rows than measured ones, and that is the honest state of this category rather than a gap in our research.

Positioning is the one area where you might reasonably expect independent testing to be easy — no lab required, just a known distance and a wheel. It still is not being done at any scale for current hardware. Every claim you read about a 2026 watch's GPS being "excellent" or "rock solid" is either a specification restated or an impression from unmeasured runs.

What you can verify before buying

The specification, which is genuinely meaningful. A single-frequency receiver hears one signal and cannot easily tell a direct path from one that bounced off a building or a rock face. A dual-frequency receiver hears the same satellite on L1 and L5, and the disagreement between them identifies the reflection.

That mechanism is well established in the underlying science, and it is checkable on a spec sheet. What is not established is how much it improves any particular watch in the field — because nobody has published that measurement.

There is a second checkable thing, and it is a good proxy for a manufacturer's honesty: does the spec sheet publish a separate battery figure for its all-systems multi-band mode? Garmin does for the Instinct 3 — 24 hours in multi-band against 40 GPS-only. A watch with no such row generally does not have the mode.

What this means for choosing a watch

If GPS drift is your actual complaint — wandering tracks in cities or under canopy — buy a multi-band receiver. It is the one upgrade in this category where the mechanism is understood, the specification is checkable, and the fix targets the specific error mode you are experiencing. The GPS upgrade path lays out which watches qualify at each price.

If it is not your complaint, do not pay for it. Single-band watches produce perfectly believable tracks on open roads and trails, and the published error range for single-band hardware — 3.2% to 6.1% — is smaller than most people assume.

The record

What has actually been measured

Every row cites a numbered source below, or states outright that no validation study exists. There is no third state.
Device	Metric	What was actually measured	Evidence
Polar Vantage M / Vantage V / V800	Distance error vs trundle wheel	The only brand under 5% error across urban, forest and 400 m track conditions. Single-band receivers, tested 2020. [1]	Peer-reviewed validation
Apple Watch Series 4	Distance error vs trundle wheel	Within the study's overall 3.2-6.1% band. Single-band, tested 2020. [1]	Peer-reviewed validation
Garmin fēnix 5X Plus	Distance error vs trundle wheel	Within the study's overall 3.2-6.1% band. Single-band, tested 2020. [1]	Peer-reviewed validation
Garmin Forerunner 935	Distance error vs trundle wheel	Within the study's overall 3.2-6.1% band. Single-band, tested 2020. [1]	Peer-reviewed validation
COROS APEX 46mm	Distance error vs trundle wheel	Within the study's overall 3.2-6.1% band. Single-band, tested 2020. [1]	Peer-reviewed validation
Suunto 9 Baro	Distance error vs trundle wheel	Within the study's overall 3.2-6.1% band. Single-band, tested 2020. [1]	Peer-reviewed validation
COROS PACE 3	GPS track quality, 10-mile comparison	The independent tester rated its GPS favorably in a measured 10-mile run against other current watches. Not peer reviewed, limited dataset by the author's own statement. [2]	Independent measured test
Garmin Forerunner 970	Positioning accuracy	No independent validation that we could find. Multi-band GNSS with SatIQ is a manufacturer specification, not a measurement. [3]	Manufacturer specification
COROS PACE 4	Positioning accuracy	No independent validation that we could find. All-Systems and Dual-Frequency modes are manufacturer specifications. [4]	Manufacturer specification
Garmin Instinct 3 Solar	Positioning accuracy	No independent validation that we could find. Garmin does publish per-mode endurance for all-satellite and multi-band modes separately, which is unusually specific. [5]	Manufacturer specification
Every current multi-band watch	Distance or positioning error	No independent validation of any current-generation dual-frequency watch that we could find, from any brand, as of August 2026.	No validation study exists
Every Suunto currently sold	Positioning accuracy	No validation of any kind that we could find.	No validation study exists

Questions

Frequently asked

Which running watch has the most accurate GPS?

Nobody can answer that honestly for current models, because no independent validation of a current-generation multi-band watch has been published that we could find. The best peer-reviewed evidence is a 2020 eight-watch study in which only Polar came in under 5% distance error — on hardware two generations old.

How accurate is GPS on a running watch?

The one controlled measurement we found put eight watches between 3.2% and 6.1% distance error against a trundle wheel, across urban, forest and track conditions. On a 10 km run that is roughly 320 to 610 meters. Dual-frequency receivers should improve on that in reflective environments; nobody has published by how much.

Is multi-band GPS worth paying for?

If you run between tall buildings or under continuous canopy, yes — it targets exactly the reflected-signal error that causes wandering tracks there. If you run on open roads, you are unlikely to see a difference, and the money is better spent on a heart-rate sensor that will change your data far more.

Why has nobody tested current GPS watches?

We do not know, and it is a fair question — the test requires a measured course and a wheel rather than a laboratory. What we can tell you is that we searched and found nothing peer-reviewed for any current dual-frequency watch, which is why every current model on this site carries a manufacturer-specification evidence tier rather than a measured one.

Show your working

Sources

Every one of these was read in full before it was cited. Where we could not access a paper, we do not quote its numbers.

[1]Peer-reviewed validation2020
Gilgen-Ammann, Schweizer & Wyss, “Accuracy of Distance Recordings in Eight Positioning-Enabled Sport Watches: Instrument Validation Study”, JMIR mHealth and uHealth
DOI 10.2196/17118. Eight watches (Apple Watch Series 4, Coros Apex 46mm, Garmin Fenix 5X Plus, Garmin Forerunner 935, Polar Vantage M, Polar Vantage V, Polar V800, Suunto 9 Baro) measured against a trundle wheel accurate to 1 cm, across urban, forest and 400 m track conditions. IMPORTANT: this predates the dual-frequency receivers now standard, so it describes brand engineering history, not any watch currently sold.
[2]Independent measured test2023
the5krunner, measured heart-rate comparison of the COROS PACE 3
An independent measured test, not a peer-reviewed study, and the author explicitly cautions the dataset was limited. Compared against Garmin HRM-Pro Plus and Polar Verity Sense across easy runs, intervals and cycling. Found divergence during and after rapid intensity change. Tests the PACE 3, not the PACE 4.
[3]Manufacturer specification2026
Garmin — Forerunner 970 published specifications
Manufacturer-stated battery: up to 15 days smartwatch mode, up to 26 hours GPS mode.
[4]Manufacturer specification2026
COROS — PACE 4 published specifications
Manufacturer-stated: 1.2" AMOLED, 32 g with nylon band, 11.8 mm thick, All-Systems and Dual-Frequency satellite modes, 41 GPS hours, 19 days of daily training and sleep.
[5]Manufacturer specification2026
Garmin — Instinct 3 Solar owner’s manual, battery life information
Manufacturer-stated for the 45 mm Solar: 28 days smartwatch (unlimited with solar), 40 hours GPS-only (130 hours with solar), 28 hours all-satellite (50 with solar), 24 hours all-satellite plus multi-band (40 with solar), 36 days expedition GPS (unlimited with solar), 65 days battery-saver. Water rating 10 ATM per the same manual’s specification page. Solar figures assume all-day wear in 50,000 lux conditions, which is Garmin’s assumption and not a measurement of ours.