Strava has changed their elevation algorithm.
#1
Senior Member
Thread Starter
Strava has changed their elevation algorithm.
Just found out that Strava has changed their algorithm for calculating elevation for devices without a barometer (ie iPhones etc). Looks like they have changed from using the various free digital elevation maps to creating their own database based on other riders with a barometer. I remember them talking about working on this last year and it looks like they have finally implemented it.
I just tried using the new elevation algorithm on a ride I did today. With barometer it gave 623m but with elevation correction it came down to 402m.
https://support.strava.com/hc/en-us/...s/115000024864
I just tried using the new elevation algorithm on a ride I did today. With barometer it gave 623m but with elevation correction it came down to 402m.
https://support.strava.com/hc/en-us/...s/115000024864
#2
Senior Member
Bad idea... now Strava will give me 1300 feet of climbing at the coast just like my Edge 510. I really wish both Garmin and Strava had a way to shut off the barometric altimeter. According to Garmin/Strava I have 79,907 feet of climbing this year. Now about 21,000 is real from a road trip to the mountains of Virginia but the other 58,907 is pure Garmin garbage. It's a real pain the a** to go in on every ride and click the elevation correction, such that I just stopped worrying about it.
#3
Senior Member
Join Date: Nov 2014
Location: Eugene, Oregon, USA
Posts: 27,547
Mentioned: 217 Post(s)
Tagged: 0 Thread(s)
Quoted: 18378 Post(s)
Liked 4,512 Times
in
3,354 Posts
It makes sense.
Ride a course once, and the elevation data is pretty bad.
Ride the same course 100 times, or 1000 times, and take averages, and the data becomes more representative.
Perhaps the biggest flaw will be short dips in the road that could get lost in averages.
Or, the places where Strava has troubles resolving which road a person is on.
Here is a local segment where most of the leaderboard is filled with people going under the overpass rather than over it. So, will Strava dock that extra 50 feet of climbing?
https://www.strava.com/segments/1062935
And this one, someone else posted, with a switchback, and the leaderboard being filled by people descending rather than climbing. So, will Strava resolve the data properly?
https://www.strava.com/segments/3267169
It all depends on how much care went into their programming. But, with the errors that seem to have been made, one has to wonder.
Ride a course once, and the elevation data is pretty bad.
Ride the same course 100 times, or 1000 times, and take averages, and the data becomes more representative.
Perhaps the biggest flaw will be short dips in the road that could get lost in averages.
Or, the places where Strava has troubles resolving which road a person is on.
Here is a local segment where most of the leaderboard is filled with people going under the overpass rather than over it. So, will Strava dock that extra 50 feet of climbing?
https://www.strava.com/segments/1062935
And this one, someone else posted, with a switchback, and the leaderboard being filled by people descending rather than climbing. So, will Strava resolve the data properly?
https://www.strava.com/segments/3267169
It all depends on how much care went into their programming. But, with the errors that seem to have been made, one has to wonder.
#4
Senior Member
Join Date: Mar 2010
Location: Seattle, WA
Posts: 23,208
Mentioned: 89 Post(s)
Tagged: 0 Thread(s)
Quoted: 18883 Post(s)
Liked 10,646 Times
in
6,054 Posts
Bad idea... now Strava will give me 1300 feet of climbing at the coast just like my Edge 510. I really wish both Garmin and Strava had a way to shut off the barometric altimeter. According to Garmin/Strava I have 79,907 feet of climbing this year. Now about 21,000 is real from a road trip to the mountains of Virginia but the other 58,907 is pure Garmin garbage. It's a real pain the a** to go in on every ride and click the elevation correction, such that I just stopped worrying about it.
#5
Senior Member
Join Date: Apr 2011
Location: Alpharetta, GA
Posts: 15,280
Bikes: Nashbar Road
Mentioned: 71 Post(s)
Tagged: 0 Thread(s)
Quoted: 2934 Post(s)
Liked 341 Times
in
228 Posts
Just found out that Strava has changed their algorithm for calculating elevation for devices without a barometer (ie iPhones etc). Looks like they have changed from using the various free digital elevation maps to creating their own database based on other riders with a barometer. I remember them talking about working on this last year and it looks like they have finally implemented it.
I just tried using the new elevation algorithm on a ride I did today. With barometer it gave 623m but with elevation correction it came down to 402m.
https://support.strava.com/hc/en-us/...s/115000024864
I just tried using the new elevation algorithm on a ride I did today. With barometer it gave 623m but with elevation correction it came down to 402m.
https://support.strava.com/hc/en-us/...s/115000024864
If Strava is really clever they'll leverage the power and speed data to further refine the elevation maps.
#6
Advocatus Diaboli
Join Date: Feb 2015
Location: Wherever I am
Posts: 8,639
Bikes: Merlin Cyrene, Nashbar steel CX
Mentioned: 14 Post(s)
Tagged: 1 Thread(s)
Quoted: 4737 Post(s)
Liked 1,533 Times
in
1,004 Posts
This thread has piqued my interest. It seems that at least a few here (eg. Floridamtb finds his altimeter has overreported elevation by 400%?) think that the barometeric altimeters in garmin units don't work? I generally at least see my start and stop elevation (ie. at home) tend to be pretty close (ie. Elev Gain = Elev Loss).
So.. what's the best way to capture the most accurate elevation stats.. with our without elevation correction? If using elevation correction and if data is sent to Strava (from Garmin connect).. is Garmin or Strava's correction better? fwiw i'm using an Edge 520.
So.. what's the best way to capture the most accurate elevation stats.. with our without elevation correction? If using elevation correction and if data is sent to Strava (from Garmin connect).. is Garmin or Strava's correction better? fwiw i'm using an Edge 520.
#7
Senior Member
Join Date: Mar 2010
Location: Seattle, WA
Posts: 23,208
Mentioned: 89 Post(s)
Tagged: 0 Thread(s)
Quoted: 18883 Post(s)
Liked 10,646 Times
in
6,054 Posts
In general a barometric altimeter is the best thing available. But there are situations where they don't work so well. In cold weather, barometers will under-report elevation. Also, a barometer measures air pressure, and tells you elevation from that, but air pressure changes, so a barometric altimeter can only work if it's calibrated to the current air conditions. If those change while you ride, it's going to get confused.
In some places the elevation correction database is pretty good and in some places it's not. Another problem with corrections is side-to-side GPS accuracy; if you're in a canyon and your GPS puts you 30 feet to the left, that could be 100 feet higher or lower than you really are. GPS accuracy tends to suffer in canyons where this is more of an issue.
The point is that there are a lot of variables, so there isn't a best way of knowing elevation, there's only a best way for the conditions. That's probably why sometimes Garmins work well and sometimes they don't.
In some places the elevation correction database is pretty good and in some places it's not. Another problem with corrections is side-to-side GPS accuracy; if you're in a canyon and your GPS puts you 30 feet to the left, that could be 100 feet higher or lower than you really are. GPS accuracy tends to suffer in canyons where this is more of an issue.
The point is that there are a lot of variables, so there isn't a best way of knowing elevation, there's only a best way for the conditions. That's probably why sometimes Garmins work well and sometimes they don't.
#8
Advocatus Diaboli
Join Date: Feb 2015
Location: Wherever I am
Posts: 8,639
Bikes: Merlin Cyrene, Nashbar steel CX
Mentioned: 14 Post(s)
Tagged: 1 Thread(s)
Quoted: 4737 Post(s)
Liked 1,533 Times
in
1,004 Posts
Thanks for the info.. So does a Garmin basically just self-calibrate when it's turned on? ie. baseline from start of the ride barometric reading? If so, would turning off and on the unit and creating separate smaller "activities" help? Obviously not ideal.
As to temperature effects, this theoretically could be easily corrected for by the unit, which also has a built-in thermometer, no?
I played with Strava's elevation correction.. it would generally remove about 10% of elevation to rides i tried it on.. Annoyingly, once you apply elevation correction in Strava, I found out you can't revert (unclick) to the Edge's originally reported elevation. I presume you'd have to resend or resync the data from Garmin again.
As to temperature effects, this theoretically could be easily corrected for by the unit, which also has a built-in thermometer, no?
I played with Strava's elevation correction.. it would generally remove about 10% of elevation to rides i tried it on.. Annoyingly, once you apply elevation correction in Strava, I found out you can't revert (unclick) to the Edge's originally reported elevation. I presume you'd have to resend or resync the data from Garmin again.
#10
Senior Member
Join Date: Mar 2010
Location: Seattle, WA
Posts: 23,208
Mentioned: 89 Post(s)
Tagged: 0 Thread(s)
Quoted: 18883 Post(s)
Liked 10,646 Times
in
6,054 Posts
Instead of breaking a ride up into several recordings, there's a way to calibrate manually in the menu. Of course where that is, is different in every model... But you could do that every time you go by an elevation sign or whatever, if you wanted to.
#12
Senior Member
Thread Starter
Thanks for the info.. So does a Garmin basically just self-calibrate when it's turned on? ie. baseline from start of the ride barometric reading? If so, would turning off and on the unit and creating separate smaller "activities" help? Obviously not ideal.
As to temperature effects, this theoretically could be easily corrected for by the unit, which also has a built-in thermometer, no?
I played with Strava's elevation correction.. it would generally remove about 10% of elevation to rides i tried it on.. Annoyingly, once you apply elevation correction in Strava, I found out you can't revert (unclick) to the Edge's originally reported elevation. I presume you'd have to resend or resync the data from Garmin again.
As to temperature effects, this theoretically could be easily corrected for by the unit, which also has a built-in thermometer, no?
I played with Strava's elevation correction.. it would generally remove about 10% of elevation to rides i tried it on.. Annoyingly, once you apply elevation correction in Strava, I found out you can't revert (unclick) to the Edge's originally reported elevation. I presume you'd have to resend or resync the data from Garmin again.
I'm not sure if it recalibrates during a ride. I wonder if I was to set a number of known points on my usual routes if the Garmin would recalibrate the altitude when I pass over them.
As for the temperature effects, I'm not so sure this would be a major problem with cycling GPS as most cyclists aren't out in sub zero temperatures. For aviation it is something to be aware of when flying in cold weather.
#13
Senior Member
I don't. I start my ride from the house and it might say 13 feet above sea level and when I return a few hours later I could be anywhere from -57 feet to 150 feet. Like I said, I stopped caring, I know my "real" elevation from the 1 or 2 road trips I take to the mountains and that's all that I care about. At least the mileage is correct...7,312 so far this year.
#15
Advocatus Diaboli
Join Date: Feb 2015
Location: Wherever I am
Posts: 8,639
Bikes: Merlin Cyrene, Nashbar steel CX
Mentioned: 14 Post(s)
Tagged: 1 Thread(s)
Quoted: 4737 Post(s)
Liked 1,533 Times
in
1,004 Posts
I don't. I start my ride from the house and it might say 13 feet above sea level and when I return a few hours later I could be anywhere from -57 feet to 150 feet. Like I said, I stopped caring, I know my "real" elevation from the 1 or 2 road trips I take to the mountains and that's all that I care about. At least the mileage is correct...7,312 so far this year.
#16
Advocatus Diaboli
Join Date: Feb 2015
Location: Wherever I am
Posts: 8,639
Bikes: Merlin Cyrene, Nashbar steel CX
Mentioned: 14 Post(s)
Tagged: 1 Thread(s)
Quoted: 4737 Post(s)
Liked 1,533 Times
in
1,004 Posts
It's not so much about "worrying" about it.. for me at least, it was just curiosity as to whether there is a simple and established best-practice that everyone should be doing to most closely capture elevation / climbing (ie. GPS correction or not).
#17
Has a magic bike
Join Date: Aug 2013
Location: Los Angeles
Posts: 12,590
Bikes: 2018 Scott Spark, 2015 Fuji Norcom Straight, 2014 BMC GF01, 2013 Trek Madone
Mentioned: 699 Post(s)
Tagged: 1 Thread(s)
Quoted: 4456 Post(s)
Liked 425 Times
in
157 Posts
Distance, I can understand. I want to know when I'm getting close to the finish line in a TT. Or make the correct turn, if I'm trying to navigate a new area.
Power- that must be accurate and reliable, Im making training & racing decisions with that data.
Speed, I can understand. Maybe I've made and modification to the bike and I want to get a sense as to how it impacted speed.
Even temperature has some usefulness, because I can compare how I feel today when it's 47 degrees and it can help me decide how to dress tomorrow if its forecast to be 42 degrees.
But elevation gain? There's almost nothing I would do with that information that needs to be more accurate than 15% +/-. So I can't really see troubleshooting it or taking any steps to try to make that info more "accurate". Knowing that I deviate systematically from Ride with GPSs totals or my friends' Garmins is good enough.
#18
Advocatus Diaboli
Join Date: Feb 2015
Location: Wherever I am
Posts: 8,639
Bikes: Merlin Cyrene, Nashbar steel CX
Mentioned: 14 Post(s)
Tagged: 1 Thread(s)
Quoted: 4737 Post(s)
Liked 1,533 Times
in
1,004 Posts
Well I didn't claim you were worrying about it. It's just that I can't imagine why it would really matter to be "accurate" about it.
Distance, I can understand. I want to know when I'm getting close to the finish line in a TT. Or make the correct turn, if I'm trying to navigate a new area.
Power- that must be accurate and reliable, Im making training & racing decisions with that data.
Speed, I can understand. Maybe I've made and modification to the bike and I want to get a sense as to how it impacted speed.
Even temperature has some usefulness, because I can compare how I feel today when it's 47 degrees and it can help me decide how to dress tomorrow if its forecast to be 42 degrees.
But elevation gain? There's almost nothing I would do with that information that needs to be more accurate than 15% +/-. So I can't really see troubleshooting it or taking any steps to try to make that info more "accurate". Knowing that I deviate systematically from Ride with GPSs totals or my friends' Garmins is good enough.
Distance, I can understand. I want to know when I'm getting close to the finish line in a TT. Or make the correct turn, if I'm trying to navigate a new area.
Power- that must be accurate and reliable, Im making training & racing decisions with that data.
Speed, I can understand. Maybe I've made and modification to the bike and I want to get a sense as to how it impacted speed.
Even temperature has some usefulness, because I can compare how I feel today when it's 47 degrees and it can help me decide how to dress tomorrow if its forecast to be 42 degrees.
But elevation gain? There's almost nothing I would do with that information that needs to be more accurate than 15% +/-. So I can't really see troubleshooting it or taking any steps to try to make that info more "accurate". Knowing that I deviate systematically from Ride with GPSs totals or my friends' Garmins is good enough.
#19
Non omnino gravis
My wife and I have had rides of 40-60 miles where the elevation reported by my 520 and her 500 are within 10 feet of each other in reported elevation-- a few have been the same to the foot. The solitary upside to riding in wind is brought to us by the barometric altimeter-- climbing and strong wind are both accompanied by falling barometric pressure. I have a ~40 mile out-and-back ride that I know for a fact is ~800ft of climbing (almost all on the return leg) but on a particularly windy day, I got credit for 1,400ft. For having to ride face-first into 20mph wind for 2 hours, I'll gladly take those 600 bonus feet.
#20
Senior Member
As for your mountain trips, I'm just curious as to how you know the "real" climbing elevation over the course of the trip? Obviously you probably know your starting and ending elevation, but that doesn't account for ups and downs in the middle.. unless the rides are only going up up up..
#21
Senior Member
Join Date: Jun 2013
Location: NYC, duh Bronx.
Posts: 3,578
Bikes: Salsa Ti Warbird- 2014/ November RAIL52s
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 67 Post(s)
Likes: 0
Liked 1 Time
in
1 Post
My 510 calibrates for elevation at the beginning of my ride...it even prompts on the screen, something like "calibrating elevation at start..." or some such nonsense. I don't pay attention to it too much.
#22
Senior Member
I live in South California. I found the barometer elevation measured by my gamin 800 is more accurate than Strava's algorithm, when compared the known elevation profile (I'm talking about >2000 ft extended Mtn climbs, Strava's Cat 2, 1, HC climbs). Strava's elev. could be off a lot, particularly at 400-600 ft. rolling hills (Cat 4 climbs), such as, those coastal hills at Orange County, CA. Strava would inflate the elevation by >30%.
If one climbs a lot, one is less likely to worry about an over/under-pass, or anything less than a Cat 4.
If one climbs a lot, one is less likely to worry about an over/under-pass, or anything less than a Cat 4.
#23
- Soli Deo Gloria -
Join Date: Aug 2015
Location: Northwest Georgia
Posts: 14,779
Bikes: 2018 Rodriguez Custom Fixed Gear, 2017 Niner RLT 9 RDO, 2015 Bianchi Pista, 2002 Fuji Robaix
Mentioned: 235 Post(s)
Tagged: 0 Thread(s)
Quoted: 6844 Post(s)
Liked 736 Times
in
469 Posts
Where I live every ride is 2000+ feet and it is all relative.
A ride with a few long steep climbs can feel harder than a ride of the same length and rolling hills, even though the ride with rolling hills might have more vertical feet. Vertical feet numbers don't tell the whole story.
-Tim-
#24
Senior Member
Thread Starter
Bit bored today so I thought I'd compare the new Strava algorithm with what my Garmin 500 (with altimeter) records.
I did a short ride (11.4km) to a nearby hill, rode the hill up and down 5 times and then rode back home. The ride to the hill is mostly downhill (I start at 85m, get as high as 133m before rolling down to 21m) but it is a bit lumpy. The hill itself is 1.6km long with a rather steady climb and a total elevation gain of around 104m (topographical maps show the hill starts at around 25m and finishes at around 130m).
I recorded all these as separate rides for ease of comparison. 1 ride there, 5 rides for the hill up and down and 1 ride home for a total of 7 rides.
On Strava (and Garmin) the total elevation gain for the ride to the hill was 135m.
The 5 hill repeats were each either 103 or 104m so not much variation there.
The ride home was a total elevation gain of 196m.
The next step was to use the Strava elevation correction on each ride.
For the hills the Strava elevation correction consistently gave 101m (versus 103-104 for the Garmin altimeter).
For the ride to the hill the elevation correction gave 92m (versus 135 for the Garmin).
For the ride back home the elevation correction gave 157m (versus 196 for the Garmin).
So there is a bit of a difference for the rides to and from the hill so I wanted to figure out what was the reason for this. I took the raw data from the Garmin fit file (I loaded the file into Golden Cheetah which lets you look at the actual ride data and it is a simple matter to copy the elevation data) and put it into a spreadsheet. I then added each increase in elevation and summed these. These totals matched the Garmin figures.
The next thing I did was download the GPX file from Strava which has the elevation corrected data. I took that data and put it into the spreadsheet to compare it against the Garmin data. For the ride back home the sum of the elevation points was 190m (versus 196m for the Garmin data).
It's interesting looking at the individual entries for the Garmin as the minimum elevation gain is 0.4m (in fact the Garmin data was only one of three possible values, either 0, 0.4m or 0.6m) while the Strava elevation correction data would have variations as low as 0.1m.
So, in conclusion, it looks like the Strava algorithm is doing something different when it sums the elevation data. The data from the elevation corrected GPX file sums to around the same as the Garmin fit file data but for some reason Strava is adding this data differently. I wonder if Strava is using some sort of minimum when summing the data.
EDIT: Done some more testing to see how Strava is adding up the data. By changing my algorithm from I have found that if the elevation change between each data point (one per second) is less than 0.15m then Strava does not count it. For the ride back home I am able to take Strava's elevation corrected GPX file and only add up the points when the elevation change is greater than 0.15 I get an answer of 157.2m which is close enough to that being reported by Strava (157m with no decimal specified). I still think this might be part of the Strava algorithm but it just happens that my Garmin 500 is not writing an elevation change if it is less than 0.4m.
I did a short ride (11.4km) to a nearby hill, rode the hill up and down 5 times and then rode back home. The ride to the hill is mostly downhill (I start at 85m, get as high as 133m before rolling down to 21m) but it is a bit lumpy. The hill itself is 1.6km long with a rather steady climb and a total elevation gain of around 104m (topographical maps show the hill starts at around 25m and finishes at around 130m).
I recorded all these as separate rides for ease of comparison. 1 ride there, 5 rides for the hill up and down and 1 ride home for a total of 7 rides.
On Strava (and Garmin) the total elevation gain for the ride to the hill was 135m.
The 5 hill repeats were each either 103 or 104m so not much variation there.
The ride home was a total elevation gain of 196m.
The next step was to use the Strava elevation correction on each ride.
For the hills the Strava elevation correction consistently gave 101m (versus 103-104 for the Garmin altimeter).
For the ride to the hill the elevation correction gave 92m (versus 135 for the Garmin).
For the ride back home the elevation correction gave 157m (versus 196 for the Garmin).
So there is a bit of a difference for the rides to and from the hill so I wanted to figure out what was the reason for this. I took the raw data from the Garmin fit file (I loaded the file into Golden Cheetah which lets you look at the actual ride data and it is a simple matter to copy the elevation data) and put it into a spreadsheet. I then added each increase in elevation and summed these. These totals matched the Garmin figures.
The next thing I did was download the GPX file from Strava which has the elevation corrected data. I took that data and put it into the spreadsheet to compare it against the Garmin data. For the ride back home the sum of the elevation points was 190m (versus 196m for the Garmin data).
It's interesting looking at the individual entries for the Garmin as the minimum elevation gain is 0.4m (in fact the Garmin data was only one of three possible values, either 0, 0.4m or 0.6m) while the Strava elevation correction data would have variations as low as 0.1m.
So, in conclusion, it looks like the Strava algorithm is doing something different when it sums the elevation data. The data from the elevation corrected GPX file sums to around the same as the Garmin fit file data but for some reason Strava is adding this data differently. I wonder if Strava is using some sort of minimum when summing the data.
EDIT: Done some more testing to see how Strava is adding up the data. By changing my algorithm from I have found that if the elevation change between each data point (one per second) is less than 0.15m then Strava does not count it. For the ride back home I am able to take Strava's elevation corrected GPX file and only add up the points when the elevation change is greater than 0.15 I get an answer of 157.2m which is close enough to that being reported by Strava (157m with no decimal specified). I still think this might be part of the Strava algorithm but it just happens that my Garmin 500 is not writing an elevation change if it is less than 0.4m.
Last edited by smarkinson; 12-31-16 at 02:56 AM.
#25
Senior Member
Join Date: Jan 2010
Location: Eugene, Oregon
Posts: 7,048
Mentioned: 10 Post(s)
Tagged: 0 Thread(s)
Quoted: 509 Post(s)
Likes: 0
Liked 9 Times
in
8 Posts
Just found out that Strava has changed their algorithm for calculating elevation for devices without a barometer (ie iPhones etc). Looks like they have changed from using the various free digital elevation maps to creating their own database based on other riders with a barometer. I remember them talking about working on this last year and it looks like they have finally implemented it.
I just tried using the new elevation algorithm on a ride I did today. With barometer it gave 623m but with elevation correction it came down to 402m.
https://support.strava.com/hc/en-us/...s/115000024864
I just tried using the new elevation algorithm on a ride I did today. With barometer it gave 623m but with elevation correction it came down to 402m.
https://support.strava.com/hc/en-us/...s/115000024864