Garmin Forerunner 970 & Apple Watch Ultra 3 GPS Accuracy and Track Tests
I’m currently testing the Huawei Watch GT Runner 2.
I’d like to share the results of a controlled GPS accuracy DISTANCE test that pitted the Huawei against the industry’s titans: the Garmin Forerunner 970, Apple Watch Ultra 3, and Amazfit Balance 2. (OK, the latter isn’t quite a Titan, but it’s nice)
More: Detailed Huawei Watch GT Runner 2 Review
The Test Protocol. Running on a track offers a rare luxury in GPS testing: a known, pre-measured distance. I ran four laps in Lane 2 on the line (using Stryd‘s standard 405.8m calibration distance recommendation) in multiple configurations. Two watches were worn on my left (one on each wrist and one handheld, NOT as shown in the photos), and two on my right. The side of the body matters because tracking from the inside wrist (closer to the lane line) follows a shorter path, adding roughly 2.5m per lap due to a 40cm separation.
I tested both pure in GPS mode (where watches rely solely on satellite positioning) and track mode (where devices apply corrections based on lap counting and known track geometry). The tests were run in both clockwise and anticlockwise directions to ensure each watch was worn on both the inside and outside wrists.
Pure GPS Results
These are the important results. In pure GPS mode across multiple runs, the results were clear:
| Watch | Avg Error | Percentage |
|---|---|---|
| Apple Watch Ultra 3 | 8.22m | 0.50% |
| Huawei Runner 2 | 10.02m | 0.62% |
| Garmin Forerunner 970 | 15.02m | 0.92% |
| Amazfit Balance 2 | 46.78m | 2.88% |
The Apple Watch Ultra 3 delivered the most accurate GPS tracking, with the Huawei Runner 2 a close second. Both devices stayed well under 1% error—impressive given the geometric challenge of tracking curved paths from a wrist-mounted antenna. The Garmin Forerunner 970 remained highly competitive, coming in at just under 1%.
Note: Stryd distance mode was disabled on the Garmin.
Note: Interesting results happen when you truly use raw Apple Watch Ultra 3 data, which bypasses Apple’s GPS track smoothing. it would have been interesting to try that on this test.
Track Mode: Things Get Interesting
As a bonus, I also tested each watch’s track mode—a feature that uses baked-in knowledge of standard track geometry to correct GPS data. The results reshuffled the rankings:
| Watch | Avg Error | Percentage |
|---|---|---|
| Garmin Forerunner 970 | 11.78m | 0.72% |
| Huawei Runner 2 | 15.02m | 0.92% |
| Amazfit Balance 2 | 32.28m | 1.99% |
| Apple Watch Ultra 3 | 48.22m | 2.96% |
The Garmin improved modestly in track mode, but the Huawei Runner 2 slipped a little. Even the Amazfit cut its error significantly (outlier in the first test). But the Apple Watch—which dominated in pure GPS—completely failed in track mode, recording identical 1580m distances in both tests despite proper start/stop positioning and lane choice (conveniently evidenced, below). This suggests Apple’s track mode either miscounts laps, partial laps or uses incorrect lane assumptions, making it effectively unusable for Lane 2 running, at least in this test (n=1).
The Verdict
For pure GPS accuracy, the Apple Watch Ultra 3 remains a gold standard, delivering strong results in many scenarios, including ones with more challenging GPS reception conditions than here.
The Huawei Runner 2 exceeded my expectations, proving itself a worthy challenger in easy GPS conditions. The Garmin Forerunner 970 delivers solid, consistent performance across both modes, achieving the best track mode accuracy. The Huawei Runner 2’s overall consistency across both pure GPS and track mode makes it a strong performer. A highly surprising test result.
What impresses me most about this unnamed device isn’t just that it matches or exceeds established brands in GPS distance accuracy—it’s that the engineers behind this particular feature have demonstrated a level of polish that rivals, and in some cases exceeds, devices from companies with decades of experience in this space.
I’ll share more details later. Several brands are poised to release new watches. For now, these results speak for themselves: the competition in the GPS sports watch market just got significantly more interesting.
Remember: GPS and HR accuracy underpin pretty much ALL the features on your sports watches. Garbage In – Garbage Out. Significantly less garbage in these tests!
Edit: One data value updated as transcribed incorrectly, the interpretation also slightly changed as a result
Last Updated on 10 March 2026 by the5krunner
Reader-Powered Content
This content is not sponsored. It’s mostly me behind the labour of love, which is this site, and I appreciate everyone who supports it.
Support the site: Follow (free, fewer ads) · Subscribe (paid, ad-free) · Buy Me A Coffee ❤️
All articles are written by real people, fact-checked, and verified for originality. See the Editorial Policy. FTC: Affiliate Disclosure — some links pay commission. As an Amazon Associate, I earn from qualifying purchases.
tfk is the founder and author of the5krunner, an independent endurance sports technology publication. With 20 years of hands-on testing of GPS watches and wearables, and competing in triathlons at an international age-group level, tfk provides in-depth expert analysis of fitness technology for serious athletes and endurance sport competitors.
I’d recommend you take the mystery watch into the woods with curves in the trails, or switchbacks…
thank you for the headsup
I tried that a couple of weekends ago and it does parallel tracks under treecover (perhaps what you found?)
other difficult scenarios were excellent tho.
hr pretty good except a few raggedy bits in niche scenarios
I find with current firmware, it does offset/parallel tracks as noted, but also frequently cuts corners (akin to old-school Apple Watch Mario Karting). Also have had a few crazy-town bad GPS moments.
It does however handle deep/long tunnels super well, better then competitors, though I want to force equal dual-freq (vs letting Garmin SatIQ).
i’ve not looked at all my test results yet. (I think we are talking the same watch). a flat ‘switchback’ (red )with trees is below. but that’s ok. you probably have more and better examples than me from where you live.
yes tunnel mode is perhaps the best I’ve seen for my ‘special’ tunnel, which must be as a result of sensor fusion not gnss. i was going to post something else on the accuracy of the mystery watch. maybe i will next week.
(Replying to the first part/thread on offset results)
Yeah, in more and more testing, it’s super clear it really struggles anytime an incline is involved (or perhaps mountains, though it seems inclines without notable mountains do the trick). Left or right side of wrist/body doesn’t matter. Frankly, it’s pretty bad compared to any number of watches/brands I throw at it as comparisons.
As for tunnel mode, it’s sensor fusion. In theory the same logic/track that Suunto/Garmin started down years ago, though seemed to kinda stop going down for whatever reason.
maybe. you propted me to look deeper (ty). i dont think the tracks you hinted at are due to inclines but rather the turn (have some crazy test results to share on that soon ie crazy test rather than the results). it doesn’t always happen and other devices also do it. the track’s turn can happen before the actual turn…think about it, that can’t be simply explained. Overall gnss to me is “excellent …but…”
tunnel: i think you are right but the company presentation notes talk about gps points for tunnel entry/exits and tunnels when following tracks. the info dosen’t quite all gel to me (same with the wearing the band on a certain wrist argument). you know all this tho.
Apple Watches (plus Ultra 1, 2, and 3) have inaccurate distance measurements since OS 26. They measure over 30 meters per kilometer more, extending the distance by 1-1.5 kilometers for a full marathon (on a Korean Association of Athletics Federations-approved marathon course).
Even if you zigzag during a typical race, you’ll only get an additional 600 meters at most, with an average of 200-400 meters.
Other GPS watches have been stable.
A Korean YouTuber measured distances using an industrial distance measuring roller and tested the 970 and Ultra 3 on a riverside road. The 970 was significantly more accurate.
https://youtu.be/1Lcu5g3YSAQ?si=esd4Hak8-N0jBjDO
that youtuber had a case over the apple watch, i dont think i can take it seriously.
there are far superior accuracy test results from this site and others e.g. https://the5krunner.com/2026/01/21/apple-watch-ultra-3-review/ and https://the5krunner.com/2025/09/20/apple-watch-ultra-3-raw-data-accuracy-is-it-better-than-in-the-workout-app/ . I have literally hundreds of apple watch ultra 3 gps tracks and comparators.
Well, in Korea, during the second half of 2025, numerous Apple Watch users sold their watches on the second-hand market and purchased Garmin and Coros due to inaccurate distance measurements (most measured about 1.5km too long).
As a result, Garmin Korea was unable to hold its Black Friday event (due to insufficient inventory).
What’s more important is how well the actual distance is reflected, not how well the GPS route is calibrated on the map.
Agreed, the actual distance is very important, both on a small scale and on a total run scale. You can get what appears to be a decent total, but the individual distance components can be way off. If the individual components are correct, the total should be.
yes there are algorithms that smooth the tracks.
I was wondering, if it is even possible to account for different GPS length when wearing the watch on left or right hand. The difference might be several meters per round which can affect the results. According to the photos you wear Amazfit and Apple on the right hand and given standard direction it could mean the watches on the right hand might have travelled by about 20 meters more than the watches on the left hand during the 4lap test.
I specifically account for the extra distance. “The side of the body matters because tracking from the inside wrist (closer to the lane line) follows a shorter path, adding roughly 2.5m per lap due to a 40cm separation.”
what i dont account for: not running exactly with the inner watch on the line all the time and the deviation of the arm laterally across my chest (I dont d that uch). these are likely trivial omissions
I missed that sentence, I am sorry! Anyway, interesting comparison.
I have a mystery watch too.
It’s the one my wife don’t know I paid a fortune for and hides in the drawer.
I don’t suppose you’re at liberty to divulge any timing on the release of this mystery watch?
Amazfit T-Rex Ultra 2 ???
Looking at those two datasets apple’s gps only mode is more accurate than its track mode.
I thought track mode was supposed to increase accuracy on tracks, not reduce it. 🤔
quite!
something wierd happend with AWU3 in track mode. i did two separate tests so i dont’ think it was user error but …maybe it was (see image of correct lane choice, admittedly just for the photo)
At the Seoul Marathon (Platinum Label) in Korea last weekend, there were a few people who used both a Garmin and an Apple Watch Ultra. They were in Group S and Group A, positions that allowed them to run without bottlenecks.
Garmin showed a distribution of 42.3–42.5 km (the 970 was 42.25–42.31, and the 965 was 42.42–42.55).
The Apple Watch Ultra (versions 1–3) showed a terrible distance error of 43–43.3 km.
After finishing a workout, the route displayed on the app looking pretty is of no value.
What matters is the running pace and distance displayed on the watch screen while actually running (this is why we use GPS watches).
The Apple Watch has been completely broken since OS 26.
(Many Koreans hoped for a patch fix soon, but the problem has not been resolved for several months.)
In a full-course marathon, the numbers displayed on the Apple Watch screen cannot be trusted at all.
city marathons are going to suffer from multipath error.
I agree screen pace is vital for pacing and one of the key reasons we need sports watches. As I’ve said numerous times on this site, I would always use STRYD for pacing and distance – https://the5krunner.com/2023/11/16/stryd-review-discount-duo/. It adds expense but makes even the cheapest Apple Watch SE an accurate pacing and distance machine!