How accurate is the GPS of the watch? That’s often the question people wonder about. Here, let’s have a look at example tracks from a Polar Vantage V and a Suunto 9 Baro.
Two Watches, One New Sony GPS Chip
Comparing these two watch models released in the second half of 2018 is interesting – and odd.
Sports testers love to produce reviews based on their impressions; many people seem to search for watch reviews. They make less and less sense, unfortunately, as watch and especially GPS firmwares develop further – and increasingly, develop further after their release to the public.
New = New, but Better?
With the Suunto 9 and the Polar Vantage series, this aspect is a particularly strong one. They both utilize a new Sony GPS chip, which has not yet been developed nearly as far – firmware-wise – as the established SirfStar chips.
This often shows in the tracks recorded with watches using this chip. They have a tendency to be less “clean” that otherwise. It is a bit difficult to judge if the newness of the chip/firmware is the only reason, however.
Mainly, it all only means that the GPS chip is new. Chances are, it will still get better. And “reviews” make rather less sense if they don’t get updated to reflect firmware changes.
Sony GPS Chip Advantages
The advantage of this chip from Sony is that its architecture apparently uses less energy, meaning longer battery lives. In addition, this chip can integrate data from a watch’s accelerometer with GPS data.
This combination of inertial and GPS positioning data makes the FusedTrack functionality of the Suunto 9 possible – and some people claim that the Polar Vantage V achieves its claimed 40 hours of maximum GPS runtime in a similar way.
To the best of my knowledge and research, this is only a claim; I have not found any official confirmation. And I still have to have a look at current battery runtimes of the Vantage V with GPS, anyways.
Comparison GPS Tracks
Let’s start off with the map of comparison tracks you can peruse yourself…
… and jump into a discussion of the tracks.
Polar Vantage V, Firmware 1, GPS Tracks
Let’s make sure we don’t make the mistake of just looking at GPS tracks without considering the firmware being used.
First comparison tracks I can show here were recorded on the initial public firmware of the Vantage.
Nov 21 GPS Track, Road Running
On November 21, the Vantage V definitely had some issues at the beginning of the recording. There, it jumped around quite a bit and started the track off in a way I could not possibly have run.
One also sees that the Suunto 9 Baro track conforms rather better with the actual location of the roads I ran on.
All in all, though, one sees more of an effect of the watches always being on opposite wrists.
There is also a likely effect of different algorithms at work, where it looks likely that the different watches count/show more GPS fixes as location or filter them as erroneous points.
One obvious case is the place where I stopped and ran back and forth a bit to take (running selfie) photos:
The Vantage V recorded more points here at one place where I was not moving (resulting in the beginning of a “GPS flower”); the Suunto 9 seems to have filtered those more. Then again, the Suunto recorded a point farther away from where I was…
6.35 km on Polar Vantage V
6.36 km on Suunto 9.
Difference: 14.3 m, 0.2%
Funnily, the Suunto 9 recorded me as running between
176 and 181 m of altitude (avg 179 m; ascent 0, descent 3 m);
the Vantage V at between
182 and 189 m of altitude (avg 187 m; ascent 0, descent 11 m).
On the flat ground this was on, I could (and should) have manually reset the Suunto 9 to the initial 183-ish meters that should be correct. It’s interesting that the Vantage looked to be more correct with that although it does not offer a manual calibration (or anything at all to do with its altitude measurement, to set manually).
Dec 9 GPS Track, Forest Roads
On Dec 9, 2018, still on firmware 1, I went over forest/field roads and into a more trail-like condition. No leaves on the trees, but enough branches overhead in parts; some altitude changes.
Again, there are some obvious differences between the tracks, and with the Vantage V recording more of an offset from the actual location of paths than the Suunto 9.
Then again, the Suunto 9 track still shows more wobbling where it meanders in ways that I didn’t run. The track does not stray as far from the roads and trails, but it wobbles…
7.76 km according to Polar Vantage V;
7.84 km as per Suunto 9.
Difference: 82.5 m, 1.1%
Polar Vantage V min 181, max 260, avg 205 m; ascent 143, descent 151 m.
Suunto 9 min 163, max 240, avg 185 m; ascent 142, descent 143 m.
Again, but more noticeably now that there was a little of altitude difference, the two watches produced pretty similar altitude graphs, but started out in rather different altitudes.
In this case, I am pretty sure that the Suunto 9 measured too low an altitude. FusedAlti didn’t correct it and I hadn’t manually set it up correctly.
Polar Vantage V, Firmware 2.0.7, GPS Tracks
With the next run, we have come to the updated 2.0 firmware.
There is no specific mention of GPS improvements in the release notes, but Polar says that this update provides some fixes and improvements. These are said to include “Speed sensitivity improved (faster reaction to changes in speed)” and “Fixes for cases where distance is missing from the beginning of a session”, so maybe there are GPS-related improvements.
Dec 19 GPS Track, Field Roads
Funnily, talking of “distance missing from the beginning,” I had an issue with the Suunto 9 here. This resulted in the Suunto missing the first 200-or-so meters.
This run was one of those cases where it’s really hard to decide without some mathematical model – which would still have to be based on a perfect track to compare to (which we do not have).
For the most part, not looking too closely, both watches reflect the path I took quite alright.
Look closer, however, and differences from the actual tracks on the ground become apparent. Partly, these are the usual offsets from GPS watches with circular antennas, being worn on opposite wrists. Partly, they are probably errors…
Here, it is rather obvious that the Suunto 9 track wobbled sideways a lot, making its track look less nice overall. Both Vantage V and Suunto 9 have issues, though, when one looks that closely.
9.92 km per Polar Vantage V;
10.04 km per Suunto 9.
Difference 124.7 m, 1.2%
Vantage V min 145, max 179, avg 163 m; ascent 61, descent 62 m.
Suunto 9 min 148, max 181, avg 165 m; ascent 63, descent 60 m.
This is a noteworthy case because it nicely shows the workings of FusedAlti: The Suunto 9 again starts out at a wrong altitude, but then corrects.
The wrong beginning is easily explained, as altitude drifts when some barometric measurements are wrongly taken for altitude changes when they were just changes in atmospheric pressure arising from changing weather.
The Suunto 9 also shows barometric (hence, weather) trends, after all; and to avoid such issues, one should regularly enter the actual, known, reference altitude.
Anyways, it was corrected; the wrong beginning altitude gets thrown out from the information shown (except for the altitude graph drawn here, from the Quantified-Self.io tool), and the overall data given looks good.
Amazing, though, that the Polar Vantage V starts out well and overall shows a believable and similar altitude profile.
For good measure, I am also showing the GPS altitude recorded by the Suunto 9. Which, as one can see, helps correct barometric altitude via FusedAlti, but wouldn’t be quite so true.
That’s the tradeoff:
Sports/outdoors watches without barometric pressure sensor (like the Suunto 9 in its non-baro variant or the Polar Vantage M) are more likely to get wrong altitude readings from GPS alone.
But watches with barometric pressure / altitude sensors can see wrong readings from weather changes.
How the Vantage V does so, seemingly, very well… I don’t know.
Dec 23 GPS Track, Roads and (Field) Trails
That was a strange run.
The Sunto 9 still wobbles, but both watches’ recorded tracks show some errors where they go places I pretty certainly didn’t (since I don’t tend to run over house roofs, even with this being right before Christmas).
As always, it is odd to look at GPS tracks trying to find, by visual inspection, which one was better.
On the one hand, erroneous sections where there are roads and houses make it clear that they are erroneous – but in such places those happen for every device.
On trails, especially when there is any twisting and turning, GPS is just not exact enough to ‘paint’ the trail exactly. In such places, there is always a big influence of algorithms; some points do not get counted because they are too close and discounted as errors. Other points are counted as sufficiently far from each other, but result more from where the watch was pointing (e.g., inside or outside of the trail) than reality. Or not…
Dec 29 Mountain Trail GPS Track
On that note, let’s head into the mountains for a final look for now…
Look closer, and you’ll only find the usual problem: It’s impossible to say which track is more accurate. Both rather nicely reflect what the trail is like and where it is, though – and for the most part, in the mountains, that is quite enough.
There is a reason I rather like using the “terrain” view as the base map for tracks like that: Once zoomed in so far that terrain lines are not being shown anymore, chances become good that one is looking at it in so close a scale, the GPS error is actually rather greater than reality.
After all, what a detailed view like the above shows is not so much that there is a difference between tracks recorded from two different GPS sports watches worn on opposite wrists. Of course, there is.
Actually, what it also shows is that such a detailed view is quite unnecessary. Sure, there can be paths branching off that would make such exactness necessary for easy navigation along a route, but those are not the usual cases.
Most of the time, differences as the above don’t really mean much on the ground. There, both tracks reflect the actual path quite alright as long as one remembers to trust markings and visible trails and not blindly follow a GPS.
(Compare my experience at the BergeSeen-Trail over the Dachstein: I used two watches for navigation there, one left, one right – and the actual trail was right in the middle between the places where they showed me!)
Track Twists and Distances
It is interesting, though, to see that the Suunto 9 here (on a new beta firmware, not the one currently available to the public) smoothed the track and did not record as many (real or ‘imagined’) twists and turns. – And yet, in overall distance recorded, the Polar Vantage V came up shorter (see below).
Perhaps the most important thing to note:
Neither watch had issues here, even in sections where the trail goes along a rock wall, on a slope, with trees all around. This would be a part where GPS reflections and more-limited views to the sky could easily cause erroneous readings of GPS. But, they didn’t.
Suunto 9 total distance: 7.08 km
Polar Vantage V total distance: 6.55 km
I also used a Garmin Instinct on that run, and that measured 7.17 km.
An earlier Suunto Spartan Ultra recording ended up showing 7.83 km (on basically, but not exactly, the same path); a Suunto Traverse used at the same time measured 7.30 km.
Mountain Trail Altitude Recording
Here, the altitude recording is also rather more interesting, if only a snapshot:
Suunto 9 – ascent: 802 m – descent:831 m – max: 1045 m
Polar Vantage V – ascent: 805 m – descent:820 m – max: 1051 m
The highest point, the Grosser Sonnstein, is actually 1037 m high… But anyways, the two watches here show altitude measurements and profiles as they should look: Very similar ones.
There is only one justified conclusion to draw: Draw your own conclusions.
Remember that (consumer-grade) GPS easily has errors of 5-10 meters, and a lot of proclamations of bad GPS falls by the wayside. Pretty much all of the errors in the tracks above are well within that range…
(I will shortly add comparison tracks from a Garmin Instinct, partly for above runs – and they will only reinforce that conclusion.)
Furthermore, GPS works differently well in different circumstances and even on different days. It does not work entirely good enough for exact pacing, at least not always, anyways.
Still, we would all like GPS tracks to be more exact, more accurate, and certainly more consistent. Polar and Suunto are sure to be working on further improvements, and those are always great to have – and I will try and follow those further.