Running Distance Usually Short?

Ɛpsilon said:

That's a huge discrepancy. 2.5 miles on a 50 mile ultra!

Actually it's not huge. It's only 5% and whether we like it or not that's within the realms of reasonable accuracy for a wrist-worn GPS device. However, while it is only 5% in this case. more often than not I get less than 5% high or low.

AndersB said:

If 3D was On or Off = not so big difference

Don't agree. There are only a few with 3D distance off but if you compare the standard deviation between those with 3D distance and those without you will notice a big difference. This is also my conclusion from trying 3D-features years back, you will get less consistency and accuracy.
There are a lot of runs with less distance using 3D distance compared to those without which is of course an impossibility.

Dr Phil said:

It's only 5% and whether we like it or not that's within the realms of reasonable accuracy for a wrist-worn GPS

But this still misses the point... the discussion here isn't about the accuracy of the Fenix's GPS chip.  The GPS chip measures position (and therefore distance) reasonably accurately. 

The point is that this measured distance (viewable in the .gpx file) isn't used.  Instead, an error is introduced for some reason that renders the distance used in activities reporting erroneously short.

The issue isn't about GPS accuracy; it's about Garmin breaking the data via some sort of processing.  I just want to be able to turn it off. 

TobiasLj said:

For some reason one device doesn't compare with another, we all use the same software so it is probably minor difference between the sensors and assembly.

I don't agree. It is clearly not a hardware issue but a software issue. The difference is that depending on satellite signal strength or perhaps a number of available satellites Garmin seems to vary its algorithms. It is quite accurate on an open terrain. But when there are any obstacles (e. g. a few trees) it seems to start buffering GPS positions and drop its "current"  position further behind. At the same time it starts losing distance and the current pace drops noticeably (by up to 2 min/mile or 3 min/km). I've observed all of three issues many times. I ran with Suunto Ambit 3 Peak on another hand and watched numbers from the two watches side by side. Ambit 3 has a massive GPS antenna which makes it one of the most accurate GPS watches ever created, along with Polar V800.

One strong evidence of increased buffering and smoothing is in how my Garmin tracks live segments. When there are trees around it is often too late to detect segment start points. Sometimes as I run by a segment start it tells me I still have 50-100 feet (15-30 meters) before it. And then as I run away from segment start the distance to segment start on the watch actually decreases as if the current position was trailing behind. Also because of smoothing my Garmin often drops off segments at sharp turns or cannot finish a segment as I run through its end. I believe all of these issues are related to inaccurate distance and pace. 

None of that happens on an open terrain when it is easier for the watch to have a good GPS reception. But if you look at the recorded track, the positions are actually reasonably accurate, which tells me that somewhere there is a fundamental bug in the software that Garmin isn't willing to admit. 

"that everyone else using different watches"....there we go with the actual answer to the problem. just get what works

That was just a selection of my statistics. And the majority of my runs are with 3D On because I did not notice any big differences on the Pace and/or Distance when I tried the setting out earlier. In the beginning I thought that it had negative impact on the Pace algorithm but I could not verify it in my statistics. The problematic Pace is due to the accelerometer, especially when running with varied cadence, varied arm movement force etc.

Here is some more statistics for "3D Off". As you can see the device placement on the arm have much more bigger impact than if 3D setting is On or Off.

From my 6189 m course:

From my 10633 m course:

As a coach watching how people run is something I do all too often. I do wonder if the Garmin algorithm is tailored to specific arm movements. Should be pretty easy to test the difference.

so those individual that don’t move their arms at all, those the run with their hands pointed down, those with large movements across the body or even the drive the arms forward get “poorer” performance rather driving the arms backwards, wrist  passed the hip at right angles and not crossing the body.

im sure someone is up for a testing challenge 

@philipshambrook 

I think you slightly mis-understand the expected accuracy for GPS.

If you start at point A and go to Point B and travel 1KM, distance from point A to point B could be recorded as 995m or 1005m and still be 5 meters accurate (within GPS spec) or 5% - but it's important to note that this is NOT cumulative.

To to re-inforce this, if you start at point A and go to point B and travel 2KM, point B could be recorded as 1995m or 2005m and still be accurate to the 5m accuracy target. What this means is that your position can be within 5 meters of your actual location - but you don't get to add this each possible error each mile, it's a point A to point B measurement. So again to re-enforce, if I start at point A and travel 10KM point B could be seen as 99995m or 100005m.- because each GPS fix should be within 5 meters of your actual position. 5% accuracy is applicable to the fix at that point in time, it's not additive. Of course, this is only aspirational and not guaranteed, and can be thrown off by environment issues or simply a satellite constellation not being favourable. Note these examples assume an 'as the crow flies' straight line.

Contrast this with Stryd which is a device that I have always found to unreport distance (consistently over 3 devices now) - in this case, if it's off by 2 to 3% (quite likely), this IS cumulative as it's essentially a form of dead reckoning, so errors are cumulative, so if you come up 30 meters short on your first KM your second KM will be 60 meters below where you'd expect, and your 3rd KM 90m below and so on- - this does not happen with GPS because with each fix being within 5 meter accuracy typically you can self-correct. Ironically, the only time I've had the 'shortness' problem which other people report is when using Stryd...but that is probably a subject for another thread.

The Pace algorithm uses GPS, accelerometer data and foot pod data (if available) to calculate the Pace. It uses the source that it thinks is most accurate at the moment. And that is a problem for some people including myself. You can easily fool the algorithm by doing a slow jog and then move your arms like you are sprinting (but stays at jog Pace). The Pace algorithm will think that you sprints and will show a faster Pace than reality. So, when we run with a varied cadence and varied arm movement force the algorithm will sometimes fail.