# Max Speed is incorrect

The speed diagram seems a bit confusing. The "top speed" says I've done 75,6 km/h, but the diagram peaks at ca. 48 km/h. Why don't these to digits correspond, and what is the correct one?

### Kommentare

11 Kommentare
• Offizieller Kommentar

Thanks, Thomas. Strava can definitely do better with max speed calculation.

Strava calculates your max speed in a such a way that any GPS error can cause the value to be greater than expected. To calculate max speed, Strava looks for the fastest speed recorded between any two GPS coordinates recorded in your activity.

To complicate matters more, on the Analysis graph there is a little bit of smoothing because we do not plot every data point but instead every 3 or so to make a readable graph. If you look at the analysis graph, you usually won't find the same max speed as is reported in the activity stats because the graph is smoothed.

• Thanks for replying, that makes sense.

• "Strava calculates your max speed in a such a way that any GPS error can cause the value to be greater than expected."

Tell me about it. I'm an inline skater, and I regularly see Strava reporting that I have maxed out at 45+ mph doing my activities.

• The problem isn't with data smoothing applied to a graph, the problem is inadequate data processing. (GPS data have certain accuracy limitations and can jump if you are surrounded by high buildings, for example.) This is also linked to Strava overinflating elevation data. Max speed is another instance where this manifests itself.

I have no idea what Strava does internally with the RAW GPS data, whether it performs a rolling average and/or a plausibility analysis, but the fact remains many other apps deal with this in a much better way. Most rider won't accelerate from 30 km/h to 85 km/h and slow down back to 30 km/h on the flats in a matter of seconds. Strava should really fix its GPS processing, this is part of its core functionality and should not be treated as an ancillary feature. What point is there in recording my rides for training purposes when I cannot trust the data?

• A largo portion of GPS errors should be corrected by automatic internal checks and brakes. I just wonder why Strava system consider it normal if a guy becomes KOM with an average bike speed of 386 km/h. (Happened in segment 15880226).

• As there is clearly no algorithm to fix these issues - as there should be/as you'd expect there to be - then at least give the user the option to correct the speed downwards.  In this way no KOMs will be stolen, but the max speed will be a better reflection of the activity.  I have previously requested "max speed of all time" as a (fun) metric.  I guess one reason this isn't shown is that it would indicate I was going 100kmh (whereas I was just going through a tunnel and lost signal).  It is pretty frustrating that such core functionality isn't addressed.

• @Bryn Davies

That's not correct, there are algorithms that can filter out implausible GPS data. Such algorithms are used in science all the time, and other GPS trackers do a better job here. The biggest problem is that GPS data, out of the box, has an uncertainty of about 10 m by design (1 m for the military), but that may change. Buildings and mountains may reflect the signals and so the accuracy may be lower in mountainous terrain or the inner city.

So you have to process GPS data one way or another. You can use assumptions and prior knowledge for that: if you are in a car, it is reasonable that you will in all likelihood be on a road in the right direction, so navigation apps take that into account. Some of the early GPS tracking apps for sports were based off of the same code and tried to put my brother and me “on the road” all the time while we were mountain biking. Our tracking was off ;-) You can discard implausible data: for example, while waiting at a traffic light in the city center, the GPS put me in the building for a few seconds on the other side of the road, causing a spike in speed (and re-starting my counter). Such a motion is highly unlikely and should — and can — be filtered out automatically. Another thing you can do is take a rolling average where you smoothen out statistical fluctuations by averaging, using the fact that the statistical mean of the fluctuation is 0 and that you move with roughly constant speed. You can, of course, also use map data. Conversely, you can use the data obtained by Strava members to correct map data. These are just the simplest off the top of my head. (My back ground is in physics, not GPS data processing in particular, though.)

These fluctuations usually do not matter as much for quantities like average speed or total distance, but can matter a lot when it comes to top speed (which is not an average) or total elevation change (where the error just adds up and has no chance to average out). In my experience, most competitors to Strava do a much better job at processing GPS data.

• @Max - thanks for the reply, however, I didn't mean that an algorithm wasn't possible, I meant that Strava hasn't applied such an algorithm, when you would expect that they would!

• @Bryn Davies

Sorry, I misunderstood what you meant to say. :-)

• Well what I don't understand is the following.

I use an app called Cycle meter to record my rides. However, I use Strava to keep all my records. So at the end of the ride I upload my ride details from Cycle Meter to Strava. However, Strava seems to perform its own calculations on the raw data upload and come up with Bizarre results. See attached screenshots

• @elle you said: "on the Analysis graph there is a little bit of smoothing because we do not plot every data point but instead every 3 or so to make a readable graph."

Is it possible to take the max speed from the Analysis graph? Because the max speed there makes sense but the one Strava shows as "Max Speed" it doesn't.

Thanks.

Bitte melden Sie sich an, um einen Kommentar zu hinterlassen.