Two GPS Watches, One Route, Different GPS Tracks

People often wonder why their GPS watches aren’t recording accurate tracks. They show single examples as proof that everything is good or nothing works right.

Here is an example (from a friend) of two watches of the same type – the Suunto 9 Baro – running the same software and GPS settings, worn on opposite wrists during the same run.

The recordings were started and ended within 3 seconds of each other, 3:32:40 PM and 3:32:43 PM. The run took 1 hour 59 minutes 50 seconds.

GPS Tracks

Of course, just seeing the overall GPS tracks does not say all that much… although some differences between them may be becoming noticeable immediately.

Zooming in, it’s a bit clearer, first of all, that the usual slight offsets from wearing the watches on opposite wrists are visible.

This alone may be something to note…

On these switchbacks, a typical issue becomes visible:

In general, any recording is good enough to say where the trail was, more or less well.

Over many runs, it’s likely that any erroneous points would just average themselves out.

In just one run like this, however, there are quite a few points where the one or the other watch did not accurately record the farthest part of a turn. Because of that, some corners are cut, some turns exaggerated.

The Misleading Exactness of GPS Tracks

It is a bit of a lie how GPS devices show such exact tracks, however, anyways.

There is still some positioning error inherent to the GPS system and especially in such weak receivers as GPS watches.

Thus, there is some “insecurity” in the positioning; what we are shown as one exact position should really be shown in a more fuzzy way as it is not actually that exactly measured.

GPS Reception Differences

One big reason for such (small) differences between watches that are from the same brand and of the same model are differences in GPS reception.

The above graphs show the number of satellites and the EHPE (estimated horizontal positioning error, a measure of the likely in-/accuracy of the GPS positioning).

Reasons for Differences

How come there are differences between two watches of the same brand, model, software?

Blocked View

One common culprit is a blocked view to the sky. If a watch cannot “see” (receive a signal from) a GPS satellite, it cannot use it in the calculation of the GPS position.

One reason why different satellites may be visible to a watch on the left vs. on the right wrist: the body of the person who wears the watch(es).

That’s why, only somewhat jokingly, people will say that you can get a much better track accuracy by wearing your GPS watch on the top of your head! 😉

The View Left-Right

The watches also point in different directions. Circular antennas are meant to get around that a bit. They are better at getting signals from satellites all around.

However, even with a circular antenna, the watch still points more to the left or to the right, and thus gets different signal strength from (some of) the same satellites.

Add in trees or buildings or mountains that block the line of sight (and thus signal – except perhaps via reflection, which adds to positioning inaccuracy) to satellites, and there is another potential reason for problems.

The GPS satellites themselves are also not always ideally ‘visible’ to GPS devices in different places – and when they preferentially ‘look’ in different directions.

Start thinking about these things, and it’s surprising – or rather, a sign of all the thought and technological progress that has gone into it – that the system works as well as it does!

No surprise that there are differences.

How Distance Diverges

In our example run, by the way, the easiest way to recognize a difference is by looking at the distance recorded (and how that developed):

The one Suunto 9 Baro recorded a total distance of 10.93 km. The other measured this same run at 11.17 km.

There was no sudden jump in how the recorded distance diverged, only a gradual divergence – as seen in the graph.

Therefore, it was probably not a sudden erroneous GPS positioning but the accumulation of small differences.

Again, there is a good chance that this relates to the watch position on opposite wrists, together with small differences in GPS positioning/recording (as visible when corners are cut or tracks extend a bit).

This run having been an out-and-back, at least it’s not the case that one watch was always on the outside of the track, so to speak, and therefore counted longer (as a matter of course, since that position would cover an ever so slightly longer distance).


Speed and pace are a bit of an issue from GPS (footpods are better at that, although they also have issues in mountain running), but for the sake of completeness…

Altitude Recordings

Altitude on a watch like the Suunto 9 Baro comes mainly from barometric readings, but there is also the calibration of it that will be done through GPS (under certain conditions), mainly at the beginning of a run – FusedAlti.

The Quantified-Self tool can also show the altitude data that the watches calculated from GPS – but GPS altitude is notoriously problematic; the EVPE (estimated vertical positioning error) is typically higher than the EHPE (the error in horizontal positioning).

What I find interesting here: It’s an example of two watches of the same model, showing very similar results – but again, as with positioning and distance, not the same results.

And there, altitude tracking is a lot easier to do and usually a whole lot better than position tracking.

Heart Rate

One last (data) point, while I have data from two watches of the same model to show: oHR.

Heart rate from the wrist, as measured by optical heart rate sensor.

oHR is another topic, just like GPS, that one could (and maybe should?) explain in a lot more depth. It’s convenient and therefore very popular.

So popular, in fact, that everyone had to go the oHR route, in recent years… even as the data one gets from these sensors is notoriously unreliable.

No, that does not mean that it never works or cannot have its use. But, it remains questionable.

Here, it worked pretty well, actually. It still has its differences and, at least two or three, as far as I can see, faultily divergent values.

Final Words

Of course, this is also just one example, not a trustworthy statistical analysis (which would have other issues in turn).

Still, I hope it helps understand that there are limits to sports tech like this which we have to accept.

Leave a Reply

Your email address will not be published. Required fields are marked *