Here are my Top 10 watches for GPS accuracy that have so far managed to complete my standard test route.

I have a 10 mile test route (16.6km) for accuracy of my GPS devices following a known track. It covers very many different types of environmental hazards to GPS accuracy such as buildings, trees and bridges/tunnels.

Below are the results of the RUNNING tests to date. All the raw files and the spreadsheet analysis that arrived at this table are linked to below.

Other watches will be added over time but, as I’ve now done more than ten, a TOP 10 seemed an appropriate post to write.

(Here) is my full methodology, source FIT/TCX files and link to the route. The methodology and route are generally good and challenging respectively, IMO.

The analysis obviously has flaws in that it is based on my subjective view of each GPS track and how it follows a precisely known route (although you are free to download the raw data and come to a different opinion). There are also other flaws in that I don’t have the opportunity to run this route repeatedly with the same device day-after-day to gain multiple data sets for each device. Tree cover and inherent GPS accuracy may well vary from day-to-day or month-to-month and indeed GPS is only supposed to be accurate to 5m in any case. (4.5m with GLONASS, supposedly)

I’m not a scientist, I only have so much time to do this. This is not supposed to represent a scientific method. Please take my views as well as those of others who post route images on forums to arrive at your own conclusions.

However there is some value to the data as most of it also broadly ties in with my perceptions of other runs I have performed with the same devices over different routes.

I wouldn’t get too focussed on, for example, the SPARTAN SPORT having the best score. Maybe it should be 78% or 77% or 76%. BUT it certainly should NOT be 60%. So allow there to be some “reasonable inaccuracy levels” in my subjective interpretation of the data. I appreciate that the SPARTAN ULTRA and SPORT SHOULD be the same and I appreciate a lot of early SPARTAN buyers were not happy with the accuracy of their devices – I can only report what I found.

Similarly the TomTom Runner 3 and TomTom Adventurer are essentially the same watch and, as you can see, they got the same score. If you look at the tracks, from the files provided above, you will see that there are obviously differences in the tracks but that the overall performance is broadly the same. Hence it’s no surprise they got a similar (identical) overall score.

I overlay the GPS tracks on top of each other to compare them that way and to the route I took, which I know precisely.

I would also add that I use my Garmin 920XT quite a lot for training. Even though it scored 63% I would say that such an accuracy level is good enough for me for much of the time to: record overall distances; to measure splits; to produce a track of where I’ve been. BUT I don’t navigate with it and I don’t rate it’s calculation of instant pace at all (a footpod is needed).

I would also point out the Suuntos: they should be the same as, like the TomToms they are essentially the same watch, but are clearly very much NOT the same. My only explanation could be variations in some of the hardware components or its assembly.

Comparing Total Recorded Distances: One surprising finding is that the total distances measured FOR EVERY DEVICE was within 1% of the average. Thus over the 16km route we are talking about 100-200m of variation. That’s not a lot, IMO. But this clearly illustrates that just because a device gives the ‘correct’ distance then it by no means indicates how well that device has recorded the nuances of the route.

Note: The best result was obtained when cycling with a MIO CYCLO 505HC and the worst was cycling with my smartphone. All the results above are arm-worn or wrist-worn whilst running.

Also of interest is the results of the V800. This indicates that when devices are worn in a sub-optimal wrist position, a notable reduction in accuracy can be observed.

Alternative data: For balance here is some alternative data from others. DCR states the 920XT to have “consistently impressive results when it comes to GPS accuracy with GLONASS enabled” which is clearly a much better performance than what I found. Also, the chart below, is from @fellrnr, Jonathan Savage, and the image clicks through to his site. Some of his findings broadly agree with mine eg the 920XT’s mediocre performance. Whereas others differ to my findings eg the M400 is found to be worse by Jonathan.

Also of interest on the chart is that Jonathan also indicates the manufacturer of the chipset and it seems fairly clear that the SiRFstar chips tend to perform better than MediaTek. I believe that Suunto use the latest SiRFstar chip. Jonathan does great work but what he does also lacks in that it does not address device-level variability (eg 2x Suunto SPARTANs could perform differently).

Clickable to source