How accurate is your GPS?

I went for a run last night and, at times, the track I looked at later was over 20m off. When I drive in London the car Satnav often has no idea where I am.

If you have experienced the same as me then you might be surprised to know that a well-designed GPS receiver can achieve an accuracy of better than 5m.
You might also know that GPS can be used to measure your elevation/altitude but the accuracy of that, in ideal conditions, is not quite so good.
But the error is not simply from the device you are using or the building you are near. There are also continually varying factors linked to the satellites’ positions above you. That’s why you can do the same run on different days and see differently plotted tracks of your route.

PDOP – Position Dilution of Precision

The term ‘dilution of Precision’ (DOP) is used to describe factors that worsen the accuracy you receive.
Complex maths derives a simple, single number for each kind of DOP. If that number is less that 1.0 then that’s “ideal” but even less than 2 is still “excellent”.  There are 4 kinds of DOP:
  • Geometric or Position (3D) dilution of precision (GDOP or PDOP – 3D);
  • Horizontal dilution of precision (HDOP);
  • Vertical dilution of precision (VDOP); and
  • Time dilution of precision (TDOP).

The maths behind a better spread of satellites means that you get less dilution and hence a better signal. Something like this diagram shows


Links to source at:

I’m most interested in PDOP and I almost always get 1.0<PDOP<2.0 for my runs. But NOT always.

Other factors are at play too

Refraction – in the atmosphere

The troposphere and ionosphere can make the effective distance from the satellite to your watch longer due to refraction (remember physics? age 15ish).

Different frequencies of signal from any one satellite can be used to minimise this effect. But not all satellites can do this.

More ion Dual Frequency (JBarbeau Via @Mirko) on

Reflection – near the ground (Multipath Effects)

If you run close to building you will sometimes see your post-run track veer away from the building and into the middle of the road. Your watch is picking up a reflected signal from the building which must have travelled further. Hence the maths puts you somewhere other than where you really are. I reckon this can give 1-3m of error in my experience.

Look here at the blue line in the bottom left as I run close to a building on Old Bridge St..

Garmin Fenix 5X 5 5S Forerunner 935 Review

Ephemeris – Satellite Location and Timing

Trilateration of distance (not triangulation of angles) is derived from each satellite’s time signals. However the atomic clock in the satellite can be slightly inaccurately synchronised. This can cause another 1-3m of error

However when you sync your sports watch you also sync the “ephemeris data” showing the satellites exact locations for the next week or so.

I *think* that sometimes when you see a post-workout GPS track that runs parallel to your real track only in one direction such as when you run South to North then this could be due to incorrect ephemeris data, most likely from it being sync’d incorrectly at some stage rather than an error in the source itself.

Something like the blue line being too far North here but ‘about right’ from East-to-West

Other Errors

There are other sources of SIGNAL error but those are the main ones for us.

Improving GPS Accuracy

GPS Chip manufacturers and those who integrate them into a sports watch can improve our accuracy. But we can help ourselves too.

  • Constellation level – GPS can be augmented with additional satellites from GLONASS and GALILEO. Simply there are more AVAILABLE satellites to choose from. The following chart ‘proves’ that GLONASS+GALILEO is better yet my experience tells me otherwise, a lot of you tell me the same. So maybe that is a theoretical improvement in ideal scenarios that don’t translate well to running watches?
  • I’ve recently seen as few as 2 GALILEO satellites available on some runs with the Fenix 5 Plus and 935. Maybe that’s why the accuracy wasn’t so great that day?
  • Satellite level – some satellites transmit dual frequency (this does NOT MEAN GPS and GLONASS). Dual frequencies will help both atmospheric refraction and multipath errors
  • Device level – a good antennae design (this paper came out top of my Google search: link to This is often bandied about as the be-all and end-all. Sure it must be important but…
  • Algorithm – from which satellites in which orbit to choose for the calculation through to how a swinging arm is accounted for. There’s a LOT of modelling that takes place.
  • User – Most of us know to wait until the watch beeps to tell us it has a signal. With Assisted GPS (A-GPS) if your current ephemeris data is up to date then you get a quick lock. If your data is not up-to-date it is ‘downloaded’ from the satellites. Which takes longer. However, I never knew until recently that the 3D position take even longer to acquire. So if you are using GPS for elevation and are bothered about its accuracy then, maybe, wait a bit longer.
  • User – wearing the watch on the underside of your arm WILL restrict the signal and increase inaccuracy.
  • User – even wearing on left/right wrists can make a difference if one side is more frequently closer to building.
  • User – wearing devices close to each other may cause interference and knock may affect signal reception?? Not sure about that but it can’t help.
  • User – For some reason I find that a good stationary ‘soak’ recording a 15 minute blank track in my back garden before heading off helps. That’s science behind that one. Give it a go.




Leave a Reply here

5 Comment threads
15 Thread replies
Most reacted comment
Hottest comment thread
5 Comment authors
paulBorisMirkoMirko Surf&Runthe5krunner Recent comment authors
newest oldest most voted
Notify of
Rui Pereira
Rui Pereira

I use a basic Garmin watch (Forerunner 25) that only has GPS (no GALILEO, etc). But between my friends I’m usually the one that gets the distance closer to the oficial race distance, even when there are tunnels and buildings involved. The guys with the Fenix and Spartan always have a bigger difference…

Mirko Surf&Run
Mirko Surf&Run

in the webinar of ublox “Super-E: low power and good performance” , that you can download from here
they found similar results as you: with a sport watch while running in their test they found that gps 2D-accuracy (95% confidence) in urban environment is 25 meter, in suburban environment 15 meter with a standard accuracy gps receiver while running.
Their test consisted in a comparison between a runner with a gps sport watch and a highly accurate gps receiver with centimeter accuracy that followed the runner.
They commented that the biggest problems in a gps watch are:
1- small dimension of antenna
2- battery problem : in a sport watch the gps must use very low power and this compromise performance
3- the swing of the arm while running
Usually to measure accuracy they don’t use always the 95% accuracy, but others methods, for example the RMS (root mean square) accuracy (that corresponds to a probability of 63-68%) or CEP (probability 50%). See


“it seems that gps accuracy of recent watches seems to be less than gps accuracy of old watches”

Yep. My (very) old FR610 generates more accurate tracks than my new(er) F3 or FR935 (better than the F3). But… I don’t see any real benefit from “super accurate” (read: precise & accurate) GPS tracks in casual sporting watches – so, I don’t care too much about it.

Better battery life means less of this annoying charging procedure. 😂

Even the bad GPS of my F3(HR) was good enough to see afterwards where I was running/walking/riding. 🤷‍♂️

“does it really matter what the post-run track looks like? for most devices it’s ok enough.”



The DOP stuff seems to be very very important, and maybe the biggest factor when we see big errors in gps data. In this article the university of Nottingham writes about the fact that the poor gps signal visibility and continuity associated with urban environments together with the slow convergence/re-convergence time of Precise Point Positioning (PPP), usually makes PPP unsuitable for land navigation in cities. With PPP they usually reach an accuracy under 1 meter, but in the article we see that in the test they made in one moment the HDOP value jumped to more than 90, and so the error increased to about 24 meter. So even with receivers that use high precision accuracy (in this example PPP), with accuracy sometimes of just few centimeter, we can have big errors when there are few satellites in view (in the case of this article in urban environment). “Fig. 8 and 9 show the HDOP and the horizontal components of the position error for the kinematic test when GPS L1 and L5 and Galileo E1 and E5a, linearly combined into the IF combination, are processed in kinematic PPP mode with the POINT software. It is possible to see how… Read more »


Looking at available data adding GLONASS to GPS will usually at least halve my HDOP/VDOP values so that it’s between 1 and 2, in reality turning it on seems to trigger some sort of offset and Suunto recommend only using it in urban environments (some tweaks in the algorithm maybe), so how the sources are integrated is a big issue even if they are available