Race 2: GPS, cadence and distance accuracy vs previous Suunto models

[jjpaz]

@Ger43 You can analyze information from your raw data and compare (export JSON or FIT files). That’s the best way to know which GPS points have recorded your watches and what’s the raw distance vs the official ditance showed in watch.

[jjpaz]

@Ger43 You can also review your activity data in https://www.sports-tracker.com/ (the old Suunto App website).
In my case, Sport Tracker analysis: Summary shows 10,20km (official distance showed by Race 2) but GPS data and Laps distances show 10,34km as total recorded distance.

With this data:

Official Distance (Device Summary): 10.201 metros
RAW Distance (GPS Points Summation): 10.334 metros
Algorithmic Reduction (Clipped): -133 metros (1.29%)

Official Pace (Reloj): 5:56 min/km
Actual GPS Pace (RAW): 5:52 min/km
Pace reduction: +4seg/km

Why Race 2 is shortening/filtering the distance? Other Suunto watches are more faithful to GPS data. That’s my question, because it seems obvious that Race 2 records GPS raw data correctly, at least is very similar to other watches but, postprocessing is more agressive.

[Ger43]

It’s raining in Milan today, so I thought I’d test it by placing both watches on my car’s dashboard and setting a cycling activity on both so I could travel at least 30 km/h

This test also confirms everything I said above, that the Race 2 measures meters…here we’re roughly 30 meters out of 5,000.

Amazfit: 5,070 m
Race 2: 5,040 m

This time, however, it seems that the difference is also in the raw file. I uploaded in “Quantifier” website the Race 2’s json file and the Amazfit .FIT file, and it seems that the 5,040 m are also in the json file this time.

My fear is that if I were to return it to Amazon and upgrade to the Vertical 2, I might find myself in the same situation. At the same time, I don’t want to buy the Race1 or Vertical1 models, which are still a step behind the Race2, and aesthetically, I really like the Race2.

This flaw, however, is unacceptable, so I don’t know what to do: wait, perhaps, for an update that permanently fixes this problem, or return it?

[EzioAuditore]

@jjpaz Just speculating here but there might be some other metric taken into account. Usually when dealing with signals there’s the Signal to Noise Ratio (SNR) and maybe there is no difference in FW but the Race 2 has lower SNR and discards more points?

[jjpaz]

@EzioAuditore Good point, because preciselly I noted that distance difference is higher in bad conditions (turns, buildings, forest,…) and lower in good conditions. Analyzing GPS quality data in several Race2 and Vertical JSON files, this is the result:

1. Analysis of Turns and Trajectory
   By comparing the GPS coordinates second-by-second during change-of-direction points:

Wismar (Race2 17,181 m): It tends to slightly “clip” or “cut” the corners. Its smoothing algorithms seem to interpret rapid direction changes as potential noise, drawing a flatter line between data points.

Rostock (Vertical 17,252 m): It records a trace that is more faithful to the actual path. In 90° or 180° turns, the Rostock accumulates between 2 and 4 extra meters per major turn compared to the Wismar.

2. Signal Quality in Critical Zones
   I looked for moments where the EHPE (Expected Horizontal Position Error) rose above 4.5 meters on the Wismar device:

“Accordion” Effect: In sections where the pace dropped sharply or there were technical stops, the Wismar activated more aggressive “auto-pause” or point-reduction filters.

Sampling Frequency: The Rostock (Orca hardware) appears to keep a slightly more consistent sampling rate during high-intensity moments, allowing it to “draw” the micro-variations of the route more accurately.

Technical Conclusion
The Rostock (Orca_RevC1) proves to be slightly more precise in this specific test because:

Its peak EHPE was lower during maneuvers.

Its Path Reconstruction algorithm is less aggressive, avoiding the “corner-cutting” effect seen in the Wismar (Sailfish_RevA1).

So, according to all my tests and analysis, seems that Race 2 algorithm behavior is different… Or is better and other watches overstimate distances or is poorer and Race 2 is underestimating…

[wakarimasen]

An interesting read for all of those suffering GPS distance errors which equate to 1-2% of the total distance travelled.

GPS mileage discrepancies

Even Garmin explains that discrepancies can occur!

Top FAQs About GPS Distance, Speed, and Pace Accuracy With Garmin Devices

I work in metrology and know that comparing different devices can lead to all sorts of confusion and doubt, if a traceable standard is not used as a reference. Unfortunately, neither Garmin, Polar, Suunto or Corus make (or even claim to make) such a device.

I guess it would be interesting to understand how the new watches calculate distance compared to the old ones. It seems that most devices tend to overestimate distances, and maybe some logic has been applied to try and reduce errors which occur due to this.

[Mff73]

@wakarimasen
You will break some dreams that some devices are having THE truth and their 10m more per 5km…

[wakarimasen]

@Mff73 said in Race 2: GPS, cadence and distance accuracy vs previous Suunto models:

@wakarimasen
You will break some dreams that some devices are having THE truth and their 10m more per 5km…

I can only apologise. Reality can be problematic

[jjpaz]

@Mff73 They are destroying our ego!

This topic is simply to learn a little more about how the gadgets work, and it’s very interesting.
Let’s use and enjoy the watches, that’s the important thing.

[duffman19]

@wakarimasen That is a good read (the first link), but it needs to be pointed out that it is from 2016. At this point, most of our devices have moved on from “recreational grade” to “mapping grade” since dual-band technology allows for accuracy to “tell us which side of the road we are on.”

I’d agree with those who argue that it isn’t important to quibble bout a 1-2% discrepancy in distance. The issue being discussed here, though, is that there is a clear and repeatable difference in distance calculation between devices from the same company that one would think are all using the same technology and algorithms (which apparently they aren’t).

[wakarimasen]

@duffman19 said in Race 2: GPS, cadence and distance accuracy vs previous Suunto models:

@wakarimasen That is a good read (the first link), but it needs to be pointed out that it is from 2016. At this point, most of our devices have moved on from “recreational grade” to “mapping grade” since dual-band technology allows for accuracy to “tell us which side of the road we are on.”

I’d agree with those who argue that it isn’t important to quibble bout a 1-2% discrepancy in distance. The issue being discussed here, though, is that there is a clear and repeatable difference in distance calculation between devices from the same company that one would think are all using the same technology and algorithms (which apparently they aren’t).

I think we may be ‘agreeing violently.’

From my previous post:
I guess it would be interesting to understand how the new watches calculate distance compared to the old ones. It seems that most devices tend to overestimate distances, and maybe some logic has been applied to try and reduce errors which occur due to this.

Nevertheless, I would be very surprised to learn that ‘official’ mapping is carried out by someone wearing a Suunto, Garmin, Polar or Corus device, and declaring the data ‘accurate.’

[jjpaz]

@wakarimasen Please, no “violence” here

The origin of the thread was “Have you noticed if the new hardware measures less distance in activities than the previous watches?”.
Just a question, just a comparison between behaviors of different watches with “similar” hardware and software.
The topic wasn’t (at least primarily) about discussing (or complaining about) GPS technology or professional map creation, although any information is welcomed to learn about it (Telecommunications Engineer here, I know some things about GPS technology ).

[sky-runner]

@wakarimasen said in Race 2: GPS, cadence and distance accuracy vs previous Suunto models:

An interesting read for all of those suffering GPS distance errors which equate to 1-2% of the total distance travelled.

GPS mileage discrepancies

That post is from 2016 when single band GPS was the only option and the accuracy was rather mediocre. Around 2022 there was a real breakthrough in GNSS accuracy, and now even cheap watches easily achieve 1 meter accuracy for a single position when using multi-system multi-band GNSS. A consistent bias of 0.5-1% in the distance measurement is no longer a random thing caused by GPS but a systemic error caused by the algorithm.

Furthermore, as has been explained above, the actual GPS distance remains accurate. The bias is introduced in post-processing, likely when fusing accelerometer and gyroscope data with the GPS data. I bet there is no such bias when doing a cycling activity and it applies only to walking or running. I know that Garmin does a similar thing for walking and running. They want to continue measuring the distance when GPS reception is lost, for example when briefly going through a tunnel or under an overpass, or near tall buildings. So the sensor fusion is continuously performed and that may make the measured distance different from the GPS distance.

[wakarimasen]

@jjpaz said in Race 2: GPS, cadence and distance accuracy vs previous Suunto models:

@wakarimasen Please, no “violence” here

The origin of the thread was “Have you noticed if the new hardware measures less distance in activities than the previous watches?”.
Just a question, just a comparison between behaviors of different watches with “similar” hardware and software.
The topic wasn’t (at least primarily) about discussing (or complaining about) GPS technology or professional map creation, although any information is welcomed to learn about it (Telecommunications Engineer here, I know some things about GPS technology ).

I bow to your greater GPS knowledge! As I mentioned, I’m wondering if Suunto have down something on the post processing side, to change the output from previous watches. The other point however, is that it’s not beyond the realms of possibility that this new data may actually be more correct than the previous one. This is the problem with comparing devices and not having an accurate reference.

No violence from my side - just curiosity, with a hefty slice of realism

[wakarimasen]

@sky-runner said in Race 2: GPS, cadence and distance accuracy vs previous Suunto models:

@wakarimasen said in Race 2: GPS, cadence and distance accuracy vs previous Suunto models:

An interesting read for all of those suffering GPS distance errors which equate to 1-2% of the total distance travelled.

GPS mileage discrepancies

That post is from 2016 when single band GPS was the only option and the accuracy was rather mediocre. Around 2022 there was a real breakthrough in GNSS accuracy, and now even cheap watches easily achieve 1 meter accuracy for a single position when using multi-system multi-band GNSS. A consistent bias of 0.5-1% in the distance measurement is no longer a random thing caused by GPS but a systemic error caused by the algorithm.

I know that Garmin does a similar thing for walking and running. They want to continue measuring the distance when GPS reception is lost, for example when briefly going through a tunnel or under an overpass, or near tall buildings. So the sensor fusion is continuously performed and that may make the measured distance different from the GPS distance.

Yes indeed, this is referenced in the (old) article. Given the over estimation that has been prevalent in devices, I’m merely wondering if this is the reason that distances have been reduced - perhaps with the introduction of a new ‘correction’ factor.

[MdzOtt]

@sky-runner said in Race 2: GPS, cadence and distance accuracy vs previous Suunto models:

bet there is no such bias when doing a cycling activity and it applies only to walking or running.

I think I reported very early in this thread that I didn’t notice any changes to the distance between S9B and Race 2 when biking (potentially<50 m on a 50 km ride, and this always varies from session to session).

[duffman19]

@wakarimasen said in Race 2: GPS, cadence and distance accuracy vs previous Suunto models:

I think we may be ‘agreeing violently.’

Agreed!

Nevertheless, I would be very surprised to learn that ‘official’ mapping is carried out by someone wearing a Suunto, Garmin, Polar or Corus device, and declaring the data ‘accurate.’

Only for OpenStreetMap

I made an interesting observation a few years ago when testing the SV1’s different GNSS presets (Performance vs. Endurance vs. Ultra). It seems that, as the GNSS accuracy/signal/capabilities (whatever you want to call it) increases, the measured distance decreases. Makes sense as fewer errors mean straighter lines and less distance. So it could stand to reason that the newer watches have even more accuracy, therefore measure slightly less distance.

However, I think this goes against all the work and analysis that @jjpaz has provided us here. It clearly looks like new watches have a different algo. My money would be on it having something to do with battery savings for those big, bright screens.

[Finnjf]

@duffman19 as a single data point, I have compared my multiple (two different Garmins, three different Coros, and three different Suuntos, one of them the Race 2) to my measuring wheel. All of them compare favourably except the Race 2, which measures short every time, in the error range we’ve talked about in this thread (usually closer to 0.5% and not as often up to 1.5%). My wheel is accurate (and technique with it is good), so from my perspective it is indeed that the Race 2 measures short, even with straight and flat lines. A great watch in many ways, one that seems to be a bit off where (imho) it matters most.