Race 2: GPS, cadence and distance accuracy vs previous Suunto models
-
I have a similar experience. After the November update, I have less distance than the Garmin FR 970. Before the update it was about the same. Now I only take what the watch shows on the display, not in the GPX viewer. We’ll see what happens after today’s update.
I have Race 2
-
@maszop said in Race 2: GPS and distance accuracy vs previous Suunto models:
@jjpaz In the past, with a Suunto watch, you had to run a bit further because Strava would slightly reduce the distance.
Strava doesn’t reduce the distance but just always rounds it down when displaying it in the UI. In contrast Suunto or Garmin properly round to the nearest significant digit. The difference never exceeds 0.01 km or mile.
-
@sky-runner Strava just analyze raw data itself.
Therefore there are different values for distance, speed, duration/pauses, etc. -
@jjpaz I think you may have uncovered something here. I do not own a newer generation Suunto (SR2 or SV2). However, I very frequently run with two or more watches at the same time and have hundreds of examples comparing different Suunto watches and competitor watches (mainly Garmin).
Suunto has traditionally used strictly GPX distance to calculate total activity distance. This can be verified (as you’ve noted above) by comparing the two statistics using a service like Quantified Self. For me, GPX distance almost always matches activity distance when wearing something like a 9PP or SV1. The one caveat is that certain types of activities will also factor in 3D distance using ascent and descent figures. Typically this will result in a very small increase in activity distance compared to GPX distance, usually something like 10-30 meters over an average trail run with mild vertical gain.
Garmin, like most other brands, uses a far more complicated algorithm to calculate activity distance. I have no idea what the exact parameters are, but it seems to weigh factors like stride length, cadence, and pace along with general GPX trace distance to come up with a final “distance.” This can be good and bad. It’s great because it can compensate for a poor GNSS signal. It’s bad because things like cadence and stride are very personal and always changing. For me, Garmin devices always calculate an activity distance that is shorter than the devices measured GPX distance.
I verified these behaviors with a test a couple of years ago comparing the recorded activity distance of an SV1 versus a Garmin 955 using different GNSS settings (Multi-band vs. all systems vs GPS only). In short, for Suunto, as the GNSS setting became less accurate, activity distance increased due to more deviations in the GPX trace. For Garmin, a lower GNSS accuracy results in a shorter activity distance, presumably because it is relying more heavily on factors other than GPX distance.
All that to say, it looks like the new Suuntos (at least your Race 2) are now using a more complicated formula to calculate activity distance. I don’t have one to compare myself, but judging from your results above, this looks an awful lot like Garmin behavior. This also squares with the fact that most reviews noted a general decrease in GNSS performance for the newer models. So all things are not equal across the Suunto lineup.
One last thought: It looks like you were doing laps in the activities above. Which wrist were you wearing the watches on when comparing them? There will always be a slight offset to the side the watch is recording. That is, a watch worn on the inside wrist will record a slightly shorter distance than one one worn on the outside when doing laps.
-
@duffman19 Thanks for your answer.
In the example activity I was wearing Race 2 in left hand, Vertical in right hand but I change direction over loops so I run every loop at least several times in every direction. I call it my “hamster wheel”…
The other days I repeated tests exchanging watches: Vertical on left, Race 2 on right. Same behavior.
-
@jjpaz No problem. Sorry that was a lot of words.
Changing directions is the way to go. I do the same and run a loop or path both forward and backward when comparing devices to offset any wrist bias.
-
I was comparing to my previouse suunto 5 after i purchased Race 2. Path on map looked much more strait (more correct) so thats may be reason for little shorter distance. I had problem with S5 that while hiking under trees route was too zig-zag and it extended real distance.
-
@maszop said in Race 2: GPS and distance accuracy vs previous Suunto models:
Strava just analyze raw data itself.
Therefore there are different values for distance, speed, duration/pauses, etc.Speed and duration - yes Strava recalculates these using their moving time algorithm. But distance comes as is without any change other than rounding when data is imported as a FIT file. If you do distance correction on Strava, only then it would recalculate the distance as a sum of distances between consecutive track points.
I received this information directly from a Strava dev. I am fairly active on Strava community hub.
-
@sky-runner Thanks for information. I thought Strava was processing everything itself.
-
I just compared some of my biking activities and I did not find any (significant) difference in distance between S9B and SR2. If there is a difference, it is less than 10-20 m in a 50 km ride. But such difference can easily be attributed to taking a slightly different course on a number of corners or curves.
-
@MdzOtt Cycling is handled differently than pedestrian activities. It is much easier to calculate distance from cycling as speeds are faster and any deviation in the GPX trace from a bad GNSS signal is minimized since the points are farther apart.
What @jjpaz has pointed out is that something has changed in the way running distance is calculated. It is no longer derived solely from GPX distance.
-
Yesterday’s test results, after the January update. ~10km flat run, steady easy rithm, linear route, out and back along the same route:
- Vertical 1. Watch showed 10.18km. GPX file distance 10,18km. Strava corrected distance 10,18km:
- Race 2. Watch showed 10.10km. GPX file distance 10,21km. Strava corrected distance 10,20km:
So, although difference is less than in previous activities, the behavior is the same. Race 2 shows less distance (~1%) than Vertical/RaceS and tracks are less accurate, with more difference/separation between lines (Vertical tracks are more “aligned”).
Race 2 is a good watch, GPS accuracy is good altough It doesn’t seem as accurate as its predecessors, and other competitors watches, in terms of GPS accuracy. Similar results regarding accuracy are also mentioned in some of the online reviews.
I’ll test again this weekend in hilly terrain, which seems to show more differences, and I’ll make a decission. -
J jjpaz referenced this topic on
-
@jjpaz
Yesterday I tested the behaviour of the new firmware on terrain with a bit more elevation:
Race 2 v2.50.26
Distance: watch: 13.15km, GPX: 13.40km, Strava corrected: 13.40km
Ascent: watch: 1023m, GPX: 1048m, Strava corrected*: 1003m -
After 3 activities after firmware update, it seems that both watches are closer in distance, altough Race2 keeps showing less distance (around 1% less).
Example from saturday run, ~12.4km, flat easy run.
- Vertical: showed 12.41km. GPX file 12.42km.
- Race2: showed 12.30km. GPX file 12:43km:
According to GPX file data, both watches recorded similar GPS points, laps and distances. If you compare GPX data track, very very very similar.
So, I use Garmin Connect to compare the FIT files downloaded from Suunto App, with the activities data from watches. This is the comparison result:
Distance is different (as shown in every watch) but, if you push the “play” button to show the 2 tracks in “real time”, they are practically equal in all the activity, similar positions, altitudes,…
So, it seems clear that every watch algorithm is calculating different. Which of them is more accurate? Is Race2 cutting distance or is V1 adding distance?
On sunday, similar behavior. 100m difference in 10K activity. Both tracks where similar and both GPX files very very close (I finished activiy in Race 2 several seconds after V1, so 20-30m more distance added in comparison).
-
@jjpaz Thanks for posting these results and continuing to dig into this topic. I find it quite interesting since I used to think that Suunto watches, in general, measured just a bit long. There was a time when Suunto’s GPX traces were a bit “jiggly” (some people referred to it as “zig-zagging”) and the assumption was that these wiggles added a slight increase to an activities overall distance.
However, over the past year, beginning with software 2.39.20, a fair amount of smoothing has been added to the traces and I’ve come to regard Suunto’s distance accuracy among the best. (I commented on this change over here. You’ll notice in the example I gave how the Vertical added slightly to the GNSS distance, presumably due to 3D distance, and the 955 subtracted from it, presumably due to the applied algorithm based on pace, cadence, stride length, etc.)
I find it interesting that, after working so hard to perfect their GNSS accuracy and tracing, Suunto has chosen to apply further distance calculations based on… who knows.
-
I may have to backtrack my observation that Suunto used to use GNSS distance alone to calculate activity distance. I just went back and looked at some older “Running” and “Walk” activities (using the new & improved Quantified Self) and noticed that sometimes activity distance was actually shorter than GNSS distance, even for older models like the SV1, 9PP, and even 9P. Not by much, perhaps only 20-40m over 5km in certain cases, but something nonetheless.
I’m mostly a trail runner, though. For these activities, GNSS distance and activity distance almost always align or activity distance is slightly longer due to elevation gain/loss.
Perhaps all Suunto watches do share the same distance algorithm (which makes sense), but the newer models are registering less-accurate GNSS accuracy and therefore applying more distance correction?
-
Hi all
I share my last test results. Yesterday evening, wearing Race 2 in left arm and, this time, Race S in right arm, to compare Race 2 against Race S instead of Vertical. Objective: ~10K flat easy run in one of my usual circuits.
Results:
- Race S: showed 10.46km. GPX file (and Strava corrected distance) 10.47km.
- Race 2: showed 10.30km. GPX file (and Strava corrected distance) 10.43km.
As on other occasions, according to GPX file data, both watches recorded similar GPS points, laps and distances. If you compare GPX data track, very very very similar.
Then, analyzing FIT file with data processed by watches differences appear:
As you can see it seems that Race S was affected by wind (up downs in altitude graph) but this elevation changes were perfectly filtered by the new ascent algorithm, so perfect in that funtion.
Then, you can see that both watches were almost identical until km 2. In this moment the distance (as tou can see in the altitude graph) starts shifting and the difference in distance increases every km during the activity.Results: after 10.3km (according to Race 2) theres a 170m difference between the 2 watches (1.5% difference). This is 6sec/km difference in average pace and 9-10W difference in average power. Is this too much? Is too less?
So, Race S shows a similar behavior than Vertical. Distance showed and measured is closer to the GPX file distance, and Race still remains calculating in the “short way”.
Is Race 2 more accurate then previous models? Are previous models more accurate? Is my Race 2 faulty altough GPX file data (GPX measured points) look OK? Am I a paranoid? What do you think? I think Race 2 is filtering data and “shorting” distances.
Some of you have compared accuracy between different devices? One of my decission points to choose Suunto was accuracy, prior to features. If accuracy now is at “Amazfit” level (joke) that point is lost.
-
I’ve been “playing” with ChatGPT and Gemini, comparing FIT files, GPX files and differences and discrepances between data in activities. GPX files show data recorded every 2 seconds (filtered), and FIT file record all data measured from watch sensors every second, so is the more accurate data.
It’s been very productive. I share some “higlights” when comparing FIT files:I've analyzed the RAW data from the two FIT files you uploaded. It's a very interesting comparison between two different models: a Suunto Race 2 (file 10f9...) and a Suunto Race S (file 439e...). Although they performed the same activity, there are key technical differences that explain the discrepancy in distance: 1. Summary of Final Distance Suunto Race 2: 10,371 meters (10.37 km) Suunto Race S: 10,461 meters (10.46 km) Difference: 90 meters, which represents an error of 0.86%. In the world of sports GPS, a difference of less than 1% is considered acceptable, but analyzing the data reveals why this occurs. 2. Differences in Cadence and Stride Length Here's the most revealing point of the analysis: Suunto Race 2: Average cadence of 87 rpm and a total of 5,128 steps. Suunto Race S: Average cadence of 85 rpm and a total of 5,151 steps. Analysis: The Race S recorded 23 more steps than the Race 2 despite registering a slightly lower average cadence. This indicates that the Race S's motion algorithm (accelerometer) was more active or interpreted the movement more sensitively. Since Suunto uses FusedSpeed (combining GPS with steps), these extra steps inflate the distance. 3. Altitude and Barometric Sensors The sensors also show different terrain readings: Race 2: Altitude range from 639.2 m to 645.6 m (6.4 meters of total elevation gain). Race S: Altitude range from 641.0 m to 645.8 m (4.8 meters of total elevation gain). Impact: The Race 2 detected more variations in the terrain. If the profile had many turns or gentle ups and downs, the watch that detects more "noise" in the altitude or positioning tends to add more meters due to the hypotenuse calculation (3D distance). 4. Temperature and Hardware Race 2: Average temperature 20°C. Race S: Average temperature 21°C. Although it may seem irrelevant, the Suunto Race 2 and Race S have different antenna sizes and case designs. The Race S is smaller, which sometimes affects the GPS antenna gain under difficult coverage conditions (such as trees or buildings). Conclusion: What is the error? The 90-meter difference is mainly due to: Step counting algorithm: The Race S counted more steps, which, in areas with less clear GPS signal, adds extra distance. Route smoothing: The Race 2 appears to have been more conservative and applied a more aggressive route "cleaning" filter, eliminating small zigzags that the Race S did include. What type of terrain did you run on? If it was an area with many buildings or tight turns, the Race 2 is usually more accurate because it filters the error better; if it was open countryside, the actual distance was likely somewhere in between. ----- Difference in Turn Sampling On tight turns, the key is how often the watch records a point: Suunto Race 2 (10,371 m): It appears to have applied a stronger smoothing algorithm. When detecting sharp turns, the watch tries to "clean up" the trajectory. If the filtering is too aggressive, it "cuts" the turn (cuts the string), which shortens the total distance. Suunto Race S (10,461 m): By recording more total steps (5,151 vs. 5,128), its accelerometer was more active. On turns, arm movement is more erratic; the Race S interpreted this extra movement as forward motion, while the Race 2 likely filtered it as arm swing noise. 3. FusedSpeed in Action As mentioned, Suunto combines GPS and an accelerometer. On a tight turn, the GPS speed drops momentarily (due to the abrupt change in direction). In that second, the watch relies 100% on the accelerometer. The fact that the Race S has a lower average cadence (85 vs. 87) but more total steps suggests that it recorded movement for a longer period or more sensitively during turns, which ultimately "inflates" the distance compared to the Race 2. 4. Hardware and Antenna The Race S is a more compact watch. The Race 2 has a different structure that can accommodate an antenna with a different orientation. On tight turns, the body sometimes blocks the view of some satellites (shadow effect). If a watch loses a satellite for half a second during a turn, the algorithm has to "guess" the trajectory, and that's where the distances diverge. Verdict summary: The Race S (10.46 km) likely overestimated slightly due to the greater sensitivity of its accelerometer during turns and arm swing. Race 2 (10.37 km) probably underestimated the distance slightly by "cutting" corners using its filtering software to avoid zigzagging. Which felt more realistic to you? Typically, on circuits with many twists, the actual distance is usually right in the middle, or closer to the clock with the cleanest line (fewer spikes) if you were to see them overlaid on a map. --------- Me: "In the previous comparison between the Race 2 and Vertical (10.12km vs. 10.18km), the Race 2 also showed 10.10km, meaning it also underestimated the distance measured in the FIT file." This is precisely the pattern of behavior of Suunto's algorithms. As in the previous case, the Suunto Race 2 applies a "post-processing" at the end of the activity that tends to be more conservative than the raw data it records in the FIT file. Here's an analysis of why the watch showed 10.10km even though the FIT file contains 10.12km: 1. The Algorithm's "Closing Adjustment" The FIT file records the accumulated distance point by point. If you add up those points, you get 10,122 meters. However, the watch (Race 2) performs one final validation when you press "Save": Micro-oscillation filtering: The watch identifies small lateral movements or GPS signal bounces that the FIT file has already recorded, but which the final algorithm decides are "noise." Result: The watch "cleans up" the trajectory and rounds down based on what it considers the most likely running line. In this case, it removed about 22 meters from the final measurement. ------- That 22-meter difference (from 10.12 to 10.10) is what makes you visually feel like the watch is "taking away" distance. 3. Why didn't the Vertical (10.18 km) underestimate as much? The Suunto Vertical has a physically larger antenna and generally has greater confidence in its GPS signal. The Vertical recorded 10,179 m on the FIT and probably showed you something very close to 10.17 or 10.18 km on the screen. With a more stable signal and a more accurate barometer (remember that the Race 2 experienced drift on the descent), the Vertical doesn't feel the need to "underestimate" the distance as much at the end because it has more confidence in the quality of the recorded points. Summary of the Race 2's performance: The Race 2 seems to be configured to be extremely cautious. It prefers to show you a slightly shorter distance (10.10 km) rather than risk giving you a distance inflated by GPS noise. On circuits with tight corners, this conservatism is more noticeable, because the watch interprets part of the turn as a possible positioning error and corrects it in the final calculation.So, it seems very clear that Race 2 sensors and algorithm are different from previous watches, maybe because is not so accurate regarding GPS measurement, is more cautious and less confident in GPS signal and apply filtering in postprocessing when saving de FIT File. In my activities Race 2 is “cutting” distance from the distance measured in his own FIT file (20, 30, 50m). RaceS and Vertical doesn’t cut/filter that FIT distance and show that as final distance.
I’ll try it again in non running/walking activities, like cycling, because in that activity I suppose that acelerometer and FusedSpeed is not mandatory.
Sorry for the big message…
-
Comparing several FIT files (Race 2 and Vertical) this is “The IA summary”
Summary of detected behavior In activities with pauses and repeated segments: The Vertical prioritizes raw accuracy: if you move, it counts it. If there's a curve, it tries to follow the path you took. The Race 2 prioritizes track clarity: it prefers to trim the track slightly (14.12 km) to avoid GPS noise giving you an inaccurate distance reading, especially after a pause or in areas you've already traversed. Final Verdict of your comparison: The Suunto Vertical performed like a precision "tank": there's hardly any error between ascent and descent, and it doesn't hesitate to count every meter you move, even during pauses. The Suunto Race 2 is an "optimizer": its priority is that the track on the map looks perfect and without spikes, so when in doubt (during pauses or sharp turns), it prefers not to add distance. -
@jjpaz Well, as I’m sure you do, I tend to take any AI response with a very large grain of salt. For instance, any reference to watch size and GPS antenna performance is bunk. Antenna design is far more complicated than “bigger is better.”
I suppose there is the possibility that there could be subtle variations in software from one watch model to the next. Although it looks like they all get the same base updates (e.g. 2.50.26 is the latest), each model actually gets a specific software package (My SV1 has 2.50.26.29840-… and the 9PP has 2.50.26.31604-…). However, I find it difficult to believe that the distance algorithms would be significantly different between watches as the AI response suggests.
We do know that the internal hardware is different between models, especially between the version 1s (Vertical, Race, Race S) and version 2s (Vertical 2 and Race 2). It’s plausible that, in an effort to conserve draw and increase battery life, the version 2s use a less active accelerometer. This would force more reliance onto GNSS accuracy and perhaps force the distance algorithm to “over compensate” and cut off more distance.
Or, perhaps it’s the opposite and the version 2 watches actually have a slightly lower-powered GNSS system forcing more reliance onto the accelerometer. After all, GNSS is the biggest sensor draw on the battery and the new watches need all the reserve they can get to power the newer, larger AMOLED screens. I could also see this causing over compensation and shorter distances. This theory would also squares with the slight decrease in GNSS accuracy that we’ve observed in the newer models.
@jjpaz It may be that few users actually care about such subtle differences, but I’m with you and would love to get to the bottom of this as I’ve always been a stickler for distance accuracy. Thanks for sharing your findings.
