---
author: ''
category: ''
date: 2013/05/15 19:46:44
description: ''
link: ''
priority: ''
slug: endomondo-lied-to-me
tags: ''
title: Endomondo Lied To Me
type: text
updated: 2013/05/15 19:46:44
url_type: ''
---
A week ago I re-started my diet and exercise plan. Since I am grossly overweight, the
exercise plan is basically "walk around 5km every day, fat guy". Since I am a 
nerd, I wanted data so I could stop lying to myself about how much I was walking.

I had seen Endomondo mentioned in my tweeter timeline a bunch of times and the featureset
looked pretty much exactly as I needed:

* Track my walking
* Keep history
* Show it in google maps (because it's nice)

It even did things like tracking calories burnt and so on.

The only problem was... it really **sucks** at figuring out how much you walked. It consistently
overestimates by around 50% the distances, and since it calculates the average speed based on
time and distance (and the time measurement is correct) it overestimates speed by 50%, which
then means it overestimates calories burnt by (I am guessing) 125%.

How did I verify that Endomondo is wrong, and avoid the obvious explanation of "your GPS is broken"?

1) I tracked myself using Endomondo and Google Trails at the same time.
2) I counted steps roman-mile style (count every "left-right", multiply by 1.6)
3) I measured the path I walked in Google Maps and Bing Maps

All those measurements tell me a walk of 1100m +/- 150m is measured by Endomondo as 1.68 km

.. raw:: html

   <iframe width="425" height="350" frameborder="0" scrolling="no" marginheight="0" marginwidth="0" src="https://maps.google.com/maps?f=d&amp;source=s_d&amp;saddr=25+de+Mayo+300,+San+Isidro,+Buenos+Aires,+Argentina&amp;daddr=-34.4719353,-58.5149842+to:Av+Centenario+700,+San+Isidro,+Buenos+Aires,+Argentina&amp;hl=es-419&amp;geocode=FcIK8v0dqTeD_ClfbI6pNLC8lTEbBX5_uRk3FQ%3BFQEA8v0d2CGD_CkzC1umNrC8lTEoIwheov32Kw%3BFZwM8v0d0haD_CmTuBJVMLC8lTE_XZ2JL1tVRg&amp;aq=&amp;sll=-34.468876,-58.513591&amp;sspn=0.006315,0.013078&amp;t=h&amp;dirflg=w&amp;mra=dpe&amp;mrsp=1&amp;sz=17&amp;via=1&amp;ie=UTF8&amp;ll=-34.468876,-58.513591&amp;spn=0.006315,0.013078&amp;output=embed"></iframe><br /><small><a href="https://maps.google.com/maps?f=d&amp;source=embed&amp;saddr=25+de+Mayo+300,+San+Isidro,+Buenos+Aires,+Argentina&amp;daddr=-34.4719353,-58.5149842+to:Av+Centenario+700,+San+Isidro,+Buenos+Aires,+Argentina&amp;hl=es-419&amp;geocode=FcIK8v0dqTeD_ClfbI6pNLC8lTEbBX5_uRk3FQ%3BFQEA8v0d2CGD_CkzC1umNrC8lTEoIwheov32Kw%3BFZwM8v0d0haD_CmTuBJVMLC8lTE_XZ2JL1tVRg&amp;aq=&amp;sll=-34.468876,-58.513591&amp;sspn=0.006315,0.013078&amp;t=h&amp;dirflg=w&amp;mra=dpe&amp;mrsp=1&amp;sz=17&amp;via=1&amp;ie=UTF8&amp;ll=-34.468876,-58.513591&amp;spn=0.006315,0.013078" style="color:#0000FF;text-align:left">Ver mapa más grande</a></small>
   
Why does this happen? I could assume Endomondo is just crap, and probably be right, but trying
to come up with a "interesting" explanation, I am leaning towards noisy measurements. For example, 
if Endomondo saw my position shifting randomly 10 or 15 meters left or right it would probably add
enough noise to make the path 50% longer (for a much more fun example of this, `read this paper (by no other than Benoit Mandelbrot!) <http://en.wikipedia.org/wiki/How_Long_Is_the_Coast_of_Britain%3F_Statistical_Self-Similarity_and_Fractional_Dimension>`_ but this doesn't explain why Google Trails works
so much better (unless Trails does something smart with antialiasing and interpolation).

If you use Endomondo, care to share your experience? I am reluctant to 1-star it in Google Play without independent confirmation.