Route & Test Methodology For Sports Watch GNSS/GPS Accuracy

I perform many different GPS tests. To the right is a map of one course that I repeatedly use for GPS tests of sports watches since before 2015. I put the full FIT/TCX/HRM files online for those who wish to do any further analysis and/or to verify/dispute results. I am happy to change any of my inferences if you correctly interpret the files differently. There is subjectivity.

What I Do

Before the test, devices will be synchronised online for technical reasons. The devices will then have a full 15 minutes to acquire the signal when turned on as a dummy run. This may allow the loading of additional positional information –  which should NOT be required by the device. However, this should give a level playing field. Until recently, GPS mode was typically used as it tended to be better, now we are in 2023 it seems that BEST ACCURACY mode means what it says! (It never used to)

I run to the start of the test recording a dummy run.

I wait for the signal to be acquired before starting out on the run and I run the course with +/-1m precision each time.

Course Description

• Estimated GPS Difficulty:
  • The harder side of average. Let’s say 60% difficult and AT LEAST representative of 80% of the usage of 80% of runners.
• Length: Approx 17km/10 miles+
• Elevation/Ascent: Flat, just above sea-level
• Types: Suburban-cum-rural. Riverbank. Parkland. Trail. Paved. Variable tree cover. Some large buildings. Close proximity to several 5m walls/buildings. A narrow alleyway. Tunnels long/short. Straights. Sharp turns. Long, sweeping bends. No power lines.
• ———————————————————————————————————————–
•     RESULTS: Public FIT/TCX File Folder: with spreadsheet analysis/results. RESULTS *
  
• ———————————————————————————————————————–

Points of Difficulty & Scoring

These will be assessed and marked out of 4; 4 being best, 1 being normal-worst, 0 being REALLY appalling. A 1-4 scale forces me to say above or below average rather than average (eg 3/5).

Being within +/-5m of the actual route WILL BE A 3 OR 4. So being more than 5m from the route means a 2 or worse. There is a degree of subjectivity to this.

1. 0.60km – test starts. Occasionally a device might accidentally not have properly fixed its GPS, despite the previous 15-minute dummy run. It has 0.6km to do so.
2. MED: 1.30km – double back/U-turn and sharp left turn.
3. MED: 2.25km – High sidewalls and tree cover. ‘Good effort to follow the road’ is expected
4. HARD: 2.39km to 3.98km – Fairly dense total overhead cover from high, deciduous trees on a gently curving path. ‘Hard to get a good track’ is expected.
5. MED: 4.36km to 4.53km – ‘Circle’ under trees finishing after going under a small road bridge
6. HARD: 4.68km to 4.86km – Sharp U-turn followed by going under the same small road bridge
7. EASY: 5.66km to 6.05km – Very easy, straight and very open. Perfect track expected.
8. EASY: 6.54km to 7.14km – Footpath lined with medium trees but open to the sky, good track expected
9. HARD: 7.22km to 7.39km – extremely difficult 2m wide track with a very high wall. ‘Cut Throat Alley’
10. EASY: 7.43km to 8.99km – ‘Typical usage’, fairly open, some trees, some curves. Should be easy.
11. EASY: 8.99km to 10.96km – Typical rural usage, fairly open, more trees than the previous section. Should be easy
12. MED: 10.96 to 11.76km – Typical tricky urban tree cover. Fairly dense deciduous tree cover at the start and high building at the end.
13. HARD: 11.76 to 12.01km – High, one-sided building preceding a sharp turn into an impossible 100m tunnel. This part of the test is partly to see how quickly the GPS location is corrected after the tunnel and partly to see how the watch handles the sharp turn.
14. EASY: 12.22km to 16.00km – Typical suburban usage – fairly straight road, limited tree cover. Some close buildings, not too high, and some buildings a front garden length away. A couple of 90-degree turns. Should be easy

Omissions

• I don’t particularly want to run through lots of wooded, hilly areas and don’t have that access in any case. There are long sections under trees nevertheless.
• Lots of frequent 90-degree turns could happen on your routes. I have included a few.
• A limited amount of buildings taller than houses have been included, some up to 7 or 8 stories. Clearly, this is a suburban-cum-rural run rather than urban. I also run next to buildings that are immediately next to the route to force errors caused by the GPS signal being reflected off the building and appearing to come from somewhere else.
• I don’t think there are any overhead power lines
• Repeatability – I will repeat the test for some devices where their score on the tests seems notably lower than my general experience. But I just simply have nowhere near enough time to test for repeatability with 10s of devices, 10s of times. TO DO THE TEST PROPERLY ONLY ONE DEVICE CAN BE WORN ON THE OUTER LEFT WRIST even changing the wrist position by a few centimetres affects results from multipath interference patterns (albeit in urban canyons)

Other reviewers who are wearing multiple watches on the same arm or wearing the tested watch strapped to their cap are performing invalid tests. Such scenarios do not replicate real-world RUNNING where the usage of a wristwatch and moving arm will introduce other potential sources of error.

Some other reviewers have no methodology, and some claim to emulate real-world usage which typically involves THEIR real-world usage rather than anyone else’s. Some repeat a relatively short run multiple times. Others seem to go for very short 1 or 2-mile runs around random parks. Each method, including mine, has problems but when I add in observations from MY real-world usage outside of the test then I think what I do is the best. But my best is certainly FAR from perfect. Ideally, if GPS accuracy is important you will want to consider 2 or 3 sources of GPS testing from testers you trust. It’s up to you whether you trust me or not. I try my best as often as I can.

General Test Route Notes

Parts of the route are on unlit, muddy tracks which might cause problems in the winter. A repeatable, planned pace is 5:00/km give or take 15 seconds/km. That seems reasonable and one which many people could emulate over the same route if they had the opportunity. Part of GPS INaccuracy comes from pounding the ground and vibrating the watch, so a reasonably consistent pace must be kept each time.

I deem the most favourable wearing position for this test as on the left wrist with the watch face on the outside of the wrist (ie antennae pointing upwards!). The results spreadsheet notes if the watch is worn in this position or not. It seems that when worn on the right wrist ON THIS ROUTE slightly worse results are obtained (closer to buildings on this clockwise route). So I will endeavour to give all watches a test whilst being the only watch on the left wrist.

There is almost no chance of traffic that would cause any deviation from the route and hence total distance should be highly similar from test to test. Accepted total distance will be estimated by an average of non-outlier results over time. Results between tests will be compared, however, this can introduce factors affecting the performance of GPS signals eg sunspots or weather or leaf cover on one particular test.

Satellite coverage varies for all of us all the time and will also be shown on the results spreadsheet along with a range of Dilution of Precision (DOP) info. Edit 2024: With multiple constellations in use, DOP is a significantly less issue than in years gone by.

Bearing all this in mind you might also look at @fellrnr’s results. His tests are very differently executed from mine and yet our results are scarily similar much of the time.

I have occasionally used parts of this course for previous, shorter tests. Every single device has had some problem or other so far on the shorter tests. Yes even the AMBIT3 and V800.

The Route from a Tourist Perspective

This is a beautiful route to run if you live in the area or visit the area, especially if you like river views and the odd historic building. The route starts and finishes at St. Mary’s University which is very well-respected for its sports-related studies & research, for its running club. and for the Sir Mo Farah Athletic Track. If like Mo Farah, you’ve run in the famous ‘Cabbage Patch 10 miler’ or the beautiful Richmond Marathon then this route follows quite a bit of the Cabbage Patch course, deviating to add points of difficulty. It also uses a part of the Kingston parkrun 5k course.

If you are coming to the UK from overseas and want a tourist run to keep you busy then this is a pretty cool run in a pretty cool part of London. It has Hampton Court Palace (King Henry VIII), the site of Richmond Palace (where Queen Elizabeth I waited for the Armada to be defeated…or not), Ham House, Eel Pie Island (Rolling Stones), Twickenham Rugby Stadium England Rugby), The Stoop (Harlequins Rugby), Petersham Nurseries/Meadow, Rowing Clubs, the first-ever canoe and hockey clubs – Royal Canoe club & Teddington Hockey, Ham Polo Club, Ham Lands/Common, Teddington Lock, Open Water Swimming & Aquathlons at Thames Young Mariners (RG Active), paddleboat trips, Strawberry Hill House, Weirs, Marble Hill House and Richmond Park for Sunday morning cycling with the lycra masses (London Duathlon – world’s largest). These locations are all VERY close to this route. There are only a handful of legally ‘protected views’ in the UK and the Richmond Hill end of this route is on one of those protected views. The uber-run-famous running mecca of BUSHY PARK is very close to the shopping centre at Kingston (where England’s early kings were crowned) and the Old Deer Park & Richmond Park parkruns are close to the shops at Richmond. There are now parkrun tourists who stay at the Travelodge in Teddington or the somewhat posher Lensbury which has a gym and indoor pool (guests & members only) –  my course would be ideal for your Sunday morning long run after your Saturday parkrun in Bushy Park.

Recommended Running Watches

• Wahoo Elemnt Rival Review
• Garmin Forerunner 945 Review
• Garmin Forerunner 245 Review
• Garmin Fenix 7 Review
• Suunto 9 Peak Review
• Polar Vantage V2 Review
• Best Bike Computer
• Best Running Watch
• Best triathlon Watch