Losing faith in Strava accuracy

England shall surely fall....

[url=http://www.bikeradar.com/forums/viewtopic.php?p=19632183#p19632183]nothernerdownsouth[/url] wrote:

I did the same ride twice on strava and the first time it came out 23.1 miles at an average of 15.4mph and the second time it came out at 22.3 miles at an average of 15.4 despite being 4 minutes quicker second time round.hmmmmmm

I've done one of my training loops around 50 times, and it's never registered more than 0.1 mile difference (highly likely that's just from taking different lines due to traffic etc).

Also on the Garmin v iPhone, my mate uses his 810 and I generally use the app, rarely bothering with my Garmin. On segments we know, have both had a pop and ridden them pretty much together, I've never noticed a discrepancy when looking at our times.

[url=http://www.bikeradar.com/forums/viewtopic.php?p=19632407#p19632407]NorvernRob[/url] wrote:

[url=http://www.bikeradar.com/forums/viewtopic.php?p=19632183#p19632183]nothernerdownsouth[/url] wrote:

I did the same ride twice on strava and the first time it came out 23.1 miles at an average of 15.4mph and the second time it came out at 22.3 miles at an average of 15.4 despite being 4 minutes quicker second time round.hmmmmmm

I've done one of my training loops around 50 times, and it's never registered more than 0.1 mile difference (highly likely that's just from taking different lines due to traffic etc).

Also on the Garmin v iPhone, my mate uses his 810 and I generally use the app, rarely bothering with my Garmin. On segments we know, have both had a pop and ridden them pretty much together, I've never noticed a discrepancy when looking at our times.

It's probably down to my mobile phone network (EE) being a bit rubbish in my area as it takes ages to get gps lock before I set up and probably go through a few dody signal areas on my ride.going to get a cycle computer to measure distance and average speed

I sold my garmin 200 because it doesn't record ascent correctly so totally useless for any challenges now I'm finding that something is happening with local segments they seem to have shifted off the main road, so I'm on the road in my activities but now can no longer match various segments.

Meh I'm really only interested in improving on my own times anyway.

On a slightly different note but still regarding strava accuracy is the issue of 'elevation correction'. I had a ride yesterday with my garmin under reporting the elevation gain so hit the correct elevation switch and had the issue the other way with strava massively over reporting. (See below)

https://www.strava.com/activities/331113816

Interestingly the Garmin site also corrected the elevation incorrectly by roughly the same amount so maybe it's just maps that are wrong.

The strava and Garmin correction depends on the base map it is using to correct. I assume the starva / garmin have bought a good basemaps for popular areas like the UK and so the correction works better than barometer altitude (in my experience, garmin 500). But doesn't work so great in other countries, where barometer reading, whilst prone to some error, would be better.

EDIT: plus what he said above... if your gps track wanders up and down contours this will add to error after correction, not matter how good or bad the basemap.

The mobile phone height is way off, as that is based on GPS which is rubbish for altitude.

Barometric altitude measurement is extremely precise, in my experience. Clearly absolute values drift over time, but I've used Polar, VDO, Proscion, Cateye and Garmin units and they all manage to get acent and descent within 10-20m of each other over a variety of routes. If you are doing ca. 1000m totals, I'd say that's pretty damn precise.

Differences between units depend on their response time. My 510 is every so lightly faster responding than my Cateye, so reads maybe 5-10% more per ride, because the 510 registers fairly minor undulations and adds them to the totals. They both give perfectly good values, providing that you understand what they are reading.

Personally I'd like to be able to take my Garmin's barometric reading and process it to remove variations less than, say 10m. That would give a more realistic estimate of actual climbing.

Mobile phones can give misleading "GPS" tracks because they determine approximate position by triangulating 3G signals. So, you might think you have a GPS signal when you don't. A Garmin either has a GPS signal, or not, but when it does its very accurate, for position anyway.

I assume that Strava uses base map data for elevation. On my commute home there is a gentle and consistent climb up through a wooded valley - Strava shows an entirely non-existent big ramp up, then descent, half way up it, but then so does every other mapping service. The GPS track does indeed wobble off the road here quite regularly, but the spike is consistent every time.

One misunderstanding a few posts back - GPS has nothing to do with your mobile signal and will in fact work fine entirely without one, although most phones do use triangulation from masts to assist. Inaccurate or slow readings will come from poor satellite reception - even trees can stuff it up.

I would love to see a system that combines all sources of position - GPS and map data, bike computer, barometric altimeter, maybe even accelerometer (which is how military positioning systems worked before GPS). Don't see why it should be that difficult to program it. For a start, why on earth doesn't Strava have a lock to road function? (See Strava Needs a Polish)

Like FA, I'm really impressed at the accuracy of altimeters - my phone gives a pretty consistently accurate reading if you just raise or lower it by a metre or even less. I have no idea how much it would vary with the wind.

In short, the technology exists for a far more accurate positioning setup.


But apart from the geekish satisfaction of accuracy, why should any cyclist really care?

[url=http://www.bikeradar.com/forums/viewtopic.php?p=19632941#p19632941]Harry-S[/url] wrote:

Garmins and I-phones rely on single frequency GPS signals to fix their positions (not GSM signals), - this is only accurate to about 20 metres in plan, and probably over 100m in elevation (which is why there's a barometric function). So it's no surprise that Strava isn't 100% accurate either.

I wrote a bit more about it here:

viewtopic.php?f=40020&t=12990185&start=20

Thank you, I wish more people would realise GPS is not magic. That explanation is really all that needs to be said in this thread, everyone should read it. GPS inaccuracies can have a huge impact on your times, especially on short fast segments where a couple of seconds either way is very significant.

It's also pretty clear that Strava doesn't do the best job of calculating segment times (or speeds) from the data it is given - ever been given an average speed on a segment that's a few mph faster than you know you were doing?

steadyrollingman wrote:

And that's another bugbear of mine... Strava's all about testing yourself, comparing your own performances with previous ones and against other riders. So anything recorded on a chaingang, even 2-up, shouldn't qualify IMO - it's just like comparing a solo TT ride with one from a TTT. Drafting a JCB... slightly murkier waters as there'll almost always be some slipstream involved on any road segment and even the effects of a fast-overtaking lorry can still be evident half a mile down the road. But setting a KoM from a 10-minute draft is cheating in my book... What's the point in doing it if you're not gonna do it properly?

It's clearly not about testing yourself though, is it? Otherwise people wouldn't get so worried about what should count as a valid KOM! You can't control what other people will do, and there will always be those who are just out to get KOMs in any way they can. Even through "doping" If you start worrying about how other people are getting their KOMs you're just going to end up frustrated and annoyed.

[url=http://www.bikeradar.com/forums/viewtopic.php?p=19632978#p19632978]steadyrollingman[/url] wrote:

[url=http://www.bikeradar.com/forums/viewtopic.php?p=19632805#p19632805]DeVlaeminck[/url] wrote:

I got a KOM on our chaingang the other week - apparently I was travelling at 42mph on a flat road - now it's a decent group but even so! Well I say a KOM - I share it with Karol Domalsky of Raleigh - I doubt he was doing 42mph either. I got another KOM on a long steady climb because a JCB came past and I sat about 6 inches off the back wheel for 10 minutes - the only value in KOMs is your mates think how the **** did he do that !

And that's another bugbear of mine... Strava's all about testing yourself, comparing your own performances with previous ones and against other riders. So anything recorded on a chaingang, even 2-up, shouldn't qualify IMO - it's just like comparing a solo TT ride with one from a TTT. Drafting a JCB... slightly murkier waters as there'll almost always be some slipstream involved on any road segment and even the effects of a fast-overtaking lorry can still be evident half a mile down the road. But setting a KoM from a 10-minute draft is cheating in my book... What's the point in doing it if you're not gonna do it properly?

you can't trust the banks these days...

[url=http://www.bikeradar.com/forums/viewtopic.php?p=19632975#p19632975]bompington[/url] wrote:

I assume that Strava uses base map data for elevation. On my commute home there is a gentle and consistent climb up through a wooded valley - Strava shows an entirely non-existent big ramp up, then descent, half way up it, but then so does every other mapping service. The GPS track does indeed wobble off the road here quite regularly, but the spike is consistent every time.

Strava, Garmin, etc use an elevation basemap of the world created from data captured on a Space Shuttle mission.
The data is low resolution (data points can be 90 metres apart) and there are many artifacts due to this low res and such things as clouds obscuring the scanning signals. This is why the 'corrected' elevation can be wrong and why you can get elevation features that don't exist.

Of course accurate and high res elevation data is available for the UK but the Ordnance Survey charge a lot for it. Strava et al is not going to pay national mapping agencies for their data (if they even have it). The other benefit of the free data is that it covers most of the planet.

There is a lower res than the OS master version but still good OS elevation data available for free under the Open Data licencing but again it is more work adding in single countries than just using the worldwide data set.

[url=http://www.bikeradar.com/forums/viewtopic.php?p=19633349#p19633349]Harry-S[/url] wrote:

Nope.

Check out Google Earth for example, - the elevation data is there, and it's acquired (apart from in very remote areas,) by aerial LIDAR. Remote areas are usually covered by old LANDSAT data. The shuttle isn't a survey vehicle.

Yep! :-)

There was a specific shuttle mission to collect terrain data for the planet. It was the Shuttle Radar Topography Mission. Search for that phrase or SRTM.

The SRTM data is freely available for anyone to download and use. It is better than what was available before but still has issues in places.

On this page https://strava.zendesk.com/entries/20965883-Elevation-for-Your-Activity Strava state they use SRTM data outside of the US.

I have noticed that for some unexplained reason some of my rides where I was putting in 4500-5000ft of climbing have started showing as under 3000 at times! I have checked all the settings..and the rides are all recorded via Garmin 510..not iPhone and Garmin…My faith in Strava was always dubious…this only confirms what I thought…its a good tool for social riding etc…but all the KOM stuff etc..all a bit tongue in cheek to me..but each to their own.