Accuracy difference between smartphone GPS trackers.

[D.P.]

I recently started running longer distances and since I tend to run in a forested park with lots of small loops and trails it is near impossible to estimate distances purely through a map. I have a fitbit one and have been using the Android fitbit app's GPS tracker which seems reasonably accurate from some tests I made but the trouble I find is the app crashes a lot, e.g, 1 in 3 runs it will crash before saving the track, which is a a sperate issue I will post about.

Anyway, that made me install endomondo app, which I really like the interface and flexibility. It also doesn't crash which is great.
However, endomondo consistently under reads the distance. E.g. The fitbit app today said 10.0 miles, endomondo 9.1. Last week the fitbit read 13.2, endomondo 11.9.

Looking at the GPS tracks it is clear that endomondo samples the GPS at a lower temporal resolution. My runs go through fairly curvy and twisty forest paths and the larger distances between GPS points misses a lot of the detail. For runs that go along fairly straight streets or gentle curves it works ok but not for my typical runs. There is a lake that is makes as 0.5 miles around which fitbit agrees with (0.49 on average) but endomondo is nearer 0.42m. I'll do quite a few loops around the lake because it is flat,wide path and pretty. Not particularly twisty but even there you can see the straighter lines of endomondo's track vs fitbit.

Anyone experienced similar issues? Any recommendations for alternative apps?

 

[D.P.]

Some apps will allow you to set the polling frequency. IpBike is one of those apps and I believe it can be used for running (I've only had any experience of using it for cycling).

Thanks, will try it out. I had a good look in endomondo and didn't see such an option but that is exactly the problem, endomondo must be polling every 5-10 seconds or some such which is just not enough

 

[D.P.]

I have noticed a difference between phones as well. My Moto G, girlfriends HTC one and my mums Iphone 5 all produce significant differences over the same run. Runkeeper and runtastic are the 2 apps Ive used. Will probably take some trial and error to find the one that will match your fitbit

that could be due to different GPS receivers giving different accuracies and is probably to be expected.

I don't necessarily care about matching my fitbit app, but I don't want an pp that under reads my runs by significant margins because it could affect my training schedule and fitness. E.g., many people say that for marathon training you should run more than 20 miles on the long runs because you just damage your muscles, if I was to use an app that consistently under reads t then I may end up doing 22 miles or more given the error rate of endomondo. that may well lead to a pleasant surprise on race day if I am significantly quicker but it could lead to higher chances of injury. Also if If in my training i come to the conclusion i wont be able to hit a certain time in the race because I have to run an extra 1-2miles to hit my perceived goal distance, then my pacing will be set at a more relaxed pace and I will get annoyed if I miss a faster time because I paced too slow.

Plus there is the whole feed back thing, I can feel how tired my muscles are and how far I have pushed myself so can know how much to increase my distance the following week but if the distances are wildly inaccurate then it get confusing. Running an extra mkle when you are close to your distance limit is not comfortable.

 

[D.P.]

Don't think my Samsung s5 does?

Just tried runtastic on a shorter recovery run. Fitbit gave a distance of 4.74 miles, runtastic 4.57. Not a huge difference but this was road running so mostly just straight roads and the odd 90* turn. I think on my regular cross-country runs the difference will be similar to endomondo.

If I have time I will see if I can export some of the paths and examine them on more detail but the review feature on the phone makes the differences fairly obvious.

 

[D.P.]

Could you not just walk the route with a 20 quid surveyors wheel and not for future exactly how far it is?


Even sections so you can add to the run as needed

If I really wanted to I guess but it is a lot of miles to add up, e.g. my 13 mile run consisted of many loops and side paths, most of which I never ran more than once or twice. the forest park i covered in little wlaking and boke paths and I end to just randomly run around different paths and explore the area. I get very bored running the same paths so always try to vary things up.

I don't need massive accuracy but I wont want to run 14.5 miles when i am aiming for 13 miles. i think when it comes down to finding a realistic race pace i will do it on some straightness roads, maybe even drive out into the countryside. I was just hoping one of the many GPS trackers has a faster polling rate than edmondo. I will try IpBike tomorrow.

 

[D.P.]

Does the S5 have ANT+? Then you might be able to use a footpod for distance rather than GPS.

I think it does does.
A straight pedometer is not what i am after though, I have a fitbit that does that. They are OK when walking but massively under read when running

However, I might be able to get some kind of ANT+ GPS footpod thing which I would hope can by used by apps at a faster polling rate. However I expect most apps will poll at the same rate.

 

[D.P.]

How about buying an old Garmin Forerunner of the bay? Imagine they;re quite cheap and with an HRM are a very good training tool.

I am thinking of this solution but wanted to know if other people had these problems and if there were any apps they use for running.


Today was a classic example. At the 5.4 mile mark the fitbit crashed, I started it again but it had crashed again after about 0.2 miles. I started it again and it went OK for another 3.6 miles when I finished my run. So I did at least 9 miles, more like 9.2m according to the fitbit app. I used runtastic in parallel and it didn't crash but the final distance was 8.5 miles. Not a bug deal for casual running but annoying if I want to work out paces etc. Looking at the track the runtastic polling rate looks similar to endomondo.


I uninstalled fitbit and reinstalled it again to see if I can get it stable.

 

[D.P.]

Thanks for the links, some good reading there. Found some other links as well with detailed comparisons and analysis of GPS watches.

Th foot pod does sound very good for a pedometer device but from some reading people say only when running on regular roads, twisty forest paths can make it no better than GPS.


I know the fitbit GPS tracker is also slightly under reading but by much less, I just wanted a different tracking app that used the same polling frequency. Most apps seem to use 5-7 seconds but fitbit seems closer to 2 seconds, the phone itself can do 1 second.

Started thinking about getting a Harmon forerunner. Some of the older models are quite cheap and supposedly more accurate than the newer models that have optimized for battery and size instead of accuracy. This means I won't have to run with my phone and can quickly glance at the watch to get distance.

 

[D.P.]

DCRainmaker is a great source of info.

Did you get a chance to try out IPBike or from what you've read are you going straight to alternative/dedicated devices?

I've i stalled it,haven't figured it out yet. Didn't see a setting for polling rate.

 

[D.P.]

Did the same recovery run as Wednesday, albeit at 8min/mile

Fitbit this time gave 4.70 miles (as opposed to 4.74). I used IPBike which gave 4.737
Both Fitbit tracks look very similar at a quick glance but there must have been some subtle differences.

I measure the route with Google maps and it said 4.7 miles, I used hillmap.com to click out the the path and it gave 4.7 and since the path wasn't exact on the curves that should eb rounded up very slightly, about 4.1 to 4.2 would be my guess.


Bit of a mixed bag here, the runtastic track is clearly poor but only one sample. There was a bit of a variance in the 2 fitbit tracks. IPBike does indeed seem better than runtastic with a faster polling rate and a distance very similar to the fitbit's first track.
however, this is a road running course so the curves are wide and gentle and the corners hard 90* turns and infrequent, so relatively easy for most trackers. Also when running in straight lines is the kind of condition that makes GPS trackers over read the distance since any noise/deviation from the true straight line will add to the distance.

I might test how ipbike does on my forest trails, it could be similar to the fitbit if it uses the same polling rate. The UI on ipbike definitely sucks and since it is biking app it doesn't really have the features I want for running (e.g. 1 miles paces etc).


I still think endomono has the best UI/features, just wish the polling rate was higher.