Data analysis of step count accuracy

[lmacmil]

I have 3 different walking routes that I have measured with Google Earth.  (I have found my iPhone GPS tends to underestimate the distance and I generally don't carry my phone on my exercise walks.)  I recently analyzed the data from 9 walks (in which all 3 routes were used) and found a pretty significant variation in step count.  My pace is pretty consistent, ranging from slightly under to slightly over 14 min/mile.  When calculating step length from the step count and distance walked, it ranges from 36.5" to 46.4"!  Since I am below average height, there's no way my step length could be 46".  I saw variations of 800 steps over the same route while the time varied only 20 seconds total (indicating a pretty similar pace.)  In another route, there was over 400 steps variation.  FWIW, I wear my tracker on my non-dominant wrist, generally but not always, at the same strap tightness (this will be my next experiment.)

BTW, I'm a retired engineer with too much time on his hands!

[Rich_Laue]

Pace and distance is not a measure of step count. 

Have you walked a mile or two, counted the actual steps taken, then compared to what the watch says. 

This is really the only way to check the step count. 

I count every left step and multiply by 2 

[lmacmil]

I have not done that.  The steps counts when I run are much more consistent over the same router, varying only about +/- 100 step from the 4200 step average.  I contend that it would be unlikely for a route walked at the same pace to have a step count vary by 25%.  That's my story and I'm sticking to it!

----------

I did a 2 mile walk today (2.06 on the phone GPS but 2.26 per Google Earth) and the tracker measured only 3194 steps which would be over 40" step length.  Twice during the walk I looked at the step count, walked 400 steps and looked again.  First time the tracker showed 317 steps, second time it showed 307 steps.  For reasons I can't explain, the step count when I walk at a brisk pace (~14 min/mile) is undercounted by about 20%.  If I adjust the step recorded step count by the amount it undercounts, I get 4030 steps and a 33" step length which is about right.

----------

Two identical walks on a route known to be about 1 mile.  First one at a leisurely 18 min/mile pace, 2nd one at an exercise pace, about 14:40/mile.  GPS enabled and shows virtually identical maps (included below.)  First walk:  1.02 miles, 2254 steps (equates to a 28.6" step length) and 79 bpm average heart rate. (FWIW, my resting heart rate is about 55 bpm.)  Second walk:  .87 miles (same route as first walk), 1466 steps (equates to 44" step length!!!) and 96 bpm.  So there's something screwing up the step count and distance tracking when I'm walking at a fast pace.  As noted above, I do not see this discrepancy when I run, only when I walk at a brisk pace.

[Triletics]

@lmacmil mind that you lost the GPS connection at least twice:

If you share TCX files I can tell you more.

[lmacmil]

I don't know what TCX files are or where they are stored.  The distance discrepancy is over 200 yards.  Surely I didn't lose the GPS for the length of time required to walk that far.

----------

I've got the TCX files for both walks and have links below.  An online viewer I found shows the distance of 1 walk is 1.602km and the other 1.568km which is .021 mile different.  The Fitbit app reported a difference of .15 miles which is 7 times more than the TCX file.  Interesting!

https://www.dropbox.com/scl/fi/95smzviqnb0q9qj95is3s/58508342792.tcx?rlkey=96lbm66052vzjb9fkrnbok4b4...

https://www.dropbox.com/scl/fi/qn5oc9udctmdmdvz7914x/58509907541.tcx?rlkey=nfn8xeznzexxdcfosbhhkhcye...

[Triletics]

@lmacmil the TCX files indeed show a difference which is normal and acceptable. In one you lost GPS for 6 seconds and in the other one for 10 seconds total. The difference you see is most likely caused by the fact that Fitbit doesn't correct the distance using GPS data on the go. For distance, GPS or not, Fitbit uses computation based on steps and stride length. It is easy to reproduce inaccuracy by changing cadence or stride length during the walk/run. There isn't much you can do about it. It's been like that always. You may try to track your walk using the "Bike" activity as this one uses the GPS for distance and still get steps and step-based distance in your daily total (the distance from steps will be different from the activity distance, one will use steps, the other GPS).

[lmacmil]

What's the point of GPS tracking if it's not to give an accurate measurement of the distance?  Then it's only good for that map of the route, which in my case, I don't need.  On the plus side, that means I don't need to take my phone with me when I walk because it doesn't give me any useful information!  Obviously my pace (cadence) and stride length will be different between a leisure walk and an exercise walk.  So unless I change the stride length depending on the speed I am walking, I'm not going to get an accurate distance measurement from the app either.  And I still think there's something off on the step count when I'm walking quickly.  At 5'5" there's no way I have a step length of 40+" which is what the known distance/# of steps gives me.

Anyway, thanks for your help and introducing me to TCX files.

[Triletics]

@lmacmil the way Fitbit tries to estimate distance is quite common for other sports watches with one little difference. Fitbit doesn't seem to compare in real-time the GPS and steps based distance and then apply corrections. GPS has low accuracy for slow moving objects. If runner wants to know real-time pace, GPS is useless. Hence using inertial methods. So in theory, Fitbit tries to do it the right way but the implementation is faulty as it doesn't force sensors to correct each other. You can observe the problem when running and seeing pace jumping all over the place. Usually, watches can correct that quickly and reduce pace error. Fitbit doesn't seem to. So it's a good thinking to involve inertial methods for measuring distance but implementation lacks other elements which would make such measurement reliable.

[lmacmil]

Removed.

[lmacmil]

Removed.

[lmacmil]

Today I did one of my regular walks which is 2.14 miles per Google Earth.  The Fitbit reported 1.9 per the GPS so pretty close.  I wore my old $25 tracker from Amazon also and it recorded 4259 steps.  Using 2.14 miles that calculates the step length at 32.8" which I think is about right.  The Fitbit reported only 3004 steps which calculates the step length of 40" using the Fitbit distance.  The step count discrepancy is close to what I previously discovered by manually counting steps and comparing to what the tracker reported.  For some reason, my arm movement during fast walks (today's pace was about 14 min/mile) is not registering every step.  Oh well...