In this third installment of my series comparing the GPS performance of different sports watches, I ran in the Prague Marathon.
Modern sports watches develop while they are already being used. Case in point: Before the Prague Marathon, the Polar Vantage V got updated to firmware 3.2.10, which Polar says to include GPS accuracy updates…
Let’s see how it went.
This run proved nicely diverse in conditions; close to ideal for a look at GPS tracking performance of sports watches in one of the situations people would most want to use them:
The situation, obviously, a marathon. Kilometer markings and all…
The conditions: A start (and finish) in an old town’s city center with urban canyons, some time on open ground next to and over a river, several out-and-back sections, a course that uses some parts twice, some underpasses and a longer tunnel.
Perfect for seeing about GPS tracks and distance.
It was rather cold, so that should have made it challenging for optical heart rate measurement. My hands got warm enough later, during the run, but there was still enough cold wind at times for some chill…
Equipment-wise, I ran with the Polar Vantage V (on firmware 3.2.10) alone on my left wrist, in order to have it undisturbed by other watches in both GPS reception and optical heart rate measurement.
The Suunto 9 Baro again accompanied it held in / worn over the back of my left hand (and getting HR readings from a Polar H10 heart rate belt), running firmware 2.6.54 (the latest public version).
On my right wrist, I wore the COROS Apex (rather too low to the wrist for good oHR readings) on its latest firmware, v1.31.
Just above the Coros Apex on my right arm, rather too close not to get some disturbance between the two watches there, I had a Suunto test watch on firmware 2.8.
This time, I used all the watches (that allow for such a setting, i.e. all but the Vantage V) with GPS only, without GLONASS.
(If you didn’t know, I’m among Suunto’s “external testers” – that’s why/how I have those Suunto. The Coros Apex is also from its maker, for testing and reviewing; I bought the Vantage V myself from my own, largely non-existent, money.
My blogs might now be successful enough they pay for their hosting, but that’s about it. There are ways you could support me; just look in the sidebar…)
Back to the present business.
The Marathon “Pacing” Situation
To make sure everything was ready, I set all the watches to training mode soon after I had managed to get into my start block. Stuck between houses close by on two sides, the view of the sky was pretty limited there; I felt pretty cold; so problems with oHR are to be expected and GPS took a long time to get a rather tentative fix.
To avoid having to bother with four watches that need starting while already running, I also started the watches then and there, some 4 minutes and 200 meters before the start time and starting line.
Kilometer Markings and Autolaps
Given the earlier start, digressions from the ideal line, and different algorithms, some difference between the autolap at every kilometer and the official markings had to be there.
These differences felt consistent at first, but became more pronounced over time on the Suunto 9 especially, and less so on the Polar Vantage V.
The Coros Apex, which is usually described as undercounting distances compared to other watches, remained at 200 (to a maximum of 300, maybe 400) meters ahead of the race course’s markings.
At kilometer 17, for example, the Suunto 9 Baro had measured a distance of 17.7 km already, the Polar Vantage V was at 17.4 km and the Coros Apex showed 17.2 km.
At kilometer 24, it was 24.78 on the Suunto 9 Baro, 24.45 on the Polar Vantage V, and 24.26 on the Coros Apex.
This increasing difference also meant that, towards the end of the marathon, the Suunto 9 annoyingly marked an autolap for km 41 right around what was actually km 40, which is not what you’d want to pace – or even just motivate – yourself with.
The Polar Vantage V was ahead of the actual distance by almost half a kilometer, which is not ideal but within the difference I would expect.
The Coros Apex, however, marked autolaps well within visible distance and what felt like mere seconds from the official marking.
Eventually, the total distances recorded were:
- 43.35 km on the Suunto 9 Baro
- (42.98 km from the other test Suunto ),
- 42.83 km with the Polar Vantage V, and
- 42.50 km according to the Coros Apex.
The Stryd footpod, by the way, recorded a distance of 42.42 km.
(I used it connected as power sensor to the Suunto 9 Baro, so that its power number would show there, but it was not in use as a speed/distance sensor – but its internal memory still records that data for the offline sync which I used to get additional data.)
This makes for differences of max. 850 meters between the GPS devices (930 meters between Stryd and Suunto 9 Baro, if we also want to look at those). Compared to the total 42.195 or the values recorded by the watches, that’s ~2% difference – or not exactly horrible, yet again.
Compared to the Vienna City Marathon
If we look back to the Vienna City Marathon mere weeks – but also an older Polar Vantage firmware – back… well, it’s difficult to conclude anything just from my data. Except perhaps that judging watch performance from snapshots like these is difficult.
At the Vienna City Marathon, all the watches showed a few more (recognizably) erroneous bits in the GPS tracks. I would like to say this is likely because Vienna does have a few more tall buildings with glass facades which might have reflected GPS signals more.
Then again, there were more meanderings on course segments in Vienna which were on rather open ground, too…
And yet, the differences in total distance recorded by the different watches there had been a bit closer together, with a maximum difference ~1%.
- 42.92 km from the Polar Vantage V
- 43.19 km from the Coros Apex and
- 43.36 km from the Suunto 9 Baro.
- (42.74 according to Stryd)
Interestingly, the Polar had recorded the shortest distance here, the Coros Apex was in the middle.
It is still very likely, judging by what the marathon was like, that the (overall) higher distance was appropriate.
I had expected Prague, e.g. with the crossing of the Charles Bridge, to require at least as much weaving around other people as was necessary in Vienna. That was absolutely not the case, however. In Prague, I was able to run close to the ideal line much more often than in Vienna, and I did not have to constantly wend my way around other runners.
That this still made for more difference between the watches, though, is a bit of a headscratcher…
Same Sony GPS Chip, Different Algorithms
This whole issue of differences in GPS tracks is particularly interesting as all these watches use a – if not the – GPS chipset from Sony.
Polar Vantage V
In the case of the Vantage, it could be the new algorithm/firmware for the GPS that is responsible for the (now) greater distance recorded.
Perhaps the greater accuracy this is supposed to have brought with firmware 3.2.10 made the watch count a bit more distance (relatively speaking) now.
What is quite obvious from the start (of the watches) in Prague is that the Polar Vantage still, new firmware as old, counts nearly every turn or GPS signal that could be a turn as movement, even when it is received while basically standing still.
This was very visible on a track at the beginning of the year, in the mountains, which I have not (yet?) added to this series because it is, by now, with comparatively old firmware…
The Coros Apex is a bit hard to estimate.
Since this marathon, I feel that it may not actually undercount distances as people often think, but be closer to reality.
It certainly has been producing the “cleanest” tracks, and looking at e.g. how its tracks show in curves, that is not just because it is recording fewer points (which could give the appearance of little “wobbling” just because points in between are recorded by other watches but not by the Apex).
Suunto 9 Baro
With the Suunto 9, unfortunately, it is quite clear that it overestimates distances because it has a tendency to “wobble.”
Tracks that are recorded while moving in a straight line tend to shift ever so slightly left and right all the time, and my assumption is that this is what adds up to the increased distance.
It is actually very strange, though, because the Suunto seems to be much better at filtering out small, quite likely not-really-existing, turns in the track (which are visible in the tracks from the Vantage V) – but then, it still ends up showing a greater distance than the Vantage.
(I guess I will really need to put up the post from the hike in the mountains which showed that very well…)
Same as in the Vienna City Marathon, I want to focus on GPS here… and the reasons are both that I want to, and that oHR just doesn’t work for me except in perfectly comfortable weather, shorter runs, and when a few gods smile on me, apparently.
Just look at the chart, and know that the Suunto 9 recorded the HR as measured via the Polar H10 heart rate chest strap.
As I mentioned at the beginning, the marathon started out pretty cold, so my hands certainly didn’t get much blood flow. That got considerably better later on, but also only at times – and it clearly shows.
The only thing I find noticeable in the data from the Vienna City Marathon and now the Prague Marathon is that there is the same decline in heart rate as recorded by oHR as the marathon gets closer to the end.
This happened in both marathons, when my actual heart rate (as recorded by the chest strap) was shifting more and, at least eventually, trending upwards.
The HR strap is certainly correct with that, not just because that is a tried-and-true technology, but also because I had some (more) walking breaks and an eventual acceleration into the finish.
What else would you like me to have a look at?
I’m thinking power (between Vantage V and Stryd), when it comes to such comparisons…