SURVEY for Fenix 6 users : how accurate is your accelerometer?

I've only just started treadmill running last week and I use a Stryd Wind/Air footpod. Stryd matches my treadmill (Assault Air Runner) and the watch takes my Stryd distance data, so I have no need for calibration.

If you don't have the luxury of Stryd or similar, is it not expected that you will calibrate distance on the watch to suit your physical characteristics and running style and maybe treadmill? Is it useful to have data from default settings from a broad range of the population and treadmills, which themselves might not be accurate? I don't know, I'm just throwing out the question before people go to the trouble of testing.

Garmin instructions for treadmill calibration below.

www8.garmin.com/.../GUID-86541696-B60E-44BC-9A46-4349C86A1CD8.html

I know about the manual calibration, I did it last time. And it still gave me a shorter distance. So I'm not even sure if it can fix the problem. Maybe I can over calibrate, and it will be spot on? But at lower speed ? Then maybe only Stryd or another foodpod can really help, but it's another expense (especially Stryd !) and I don't even need a very accurate instant pace. A more accurate distance would be nice though. And the watch GPS is good enough to give me that. Most trail runners won't think about calibrating the accelerometer in a treadmill activity in order to have a more reliable distance. That's why I'm curious to see how the accelerometer is behaving. The problem is also that it seems to autocalibrate from the GPS. But how often ? How efficient is this auto calibration? That's a lot of questions I may never find the answer. But I'm still curious to understand how it's working and how to explain those shorter distances.

I just tried an experiment to confirm a suspicion. I started a treadmill activity on the watch and then jogged on the spot and got a speed of 4 kph. No actual treadmill involved, no Stryd. Then I tried keeping my hands by my sides and still jogging. Similar result. Then I tried swinging my arms as though running, but without taking steps. Speed was 0. Finally I stood still and just jiggled my arm up and down to simulate the impact of steps from arm movement alone. I got a cadence reading of 190 and a speed of 10 kph whilst standing still. In total I covered 110m without moving from the spot.

From this I conclude that steps (real or faked) are the primary driver for speed/distance. Therefore stride length and turnover, accompanied by arm motion, are going to be a massive influence on the data produced. I presume Garmin makes some assumptions regarding leg length and stride length, but without calibration how can the watch really know the truth?

Also, I imagine that a calibration at one pace and stride length may not work well for significant changes in these things. Maybe the watch has added sophistication in the accelerometer than merely counting steps, but also perhaps what happens in the legs is not mirrored in the movements at the wrist. e.g. if you lengthen stride but arm movement remains unchanged, can the watch interpret this accurately? 

Then how about Vloggers who might be holding a camera as steady as possible while they run. How would the watch handle such things?

Just guessing here, but based on my rather silly test, I would think that calibration is essential, and would struggle to keep up with complex changes in cadence, stride and arm movements in trail running. Maybe on a track at steady 5k pace, for example, would yield good results, but deviate widely from your calibration pace and I guess things could easily go downhill.

I often see values within 0.5% of treadmill distance. It's gotten more accurate with recent software updates.

Accuracy is affected by arm-swing. If I hold the rails on the treadmill, punch control panel buttons for too long or play with my phone, the accuracy drifts. It self-corrects when I go back to a normal walking or running gait. I'm legit impressed with how accurate it is without a foot-pod.

After my first run with 1.73km instead of 2km, I have calibrated the watch.
Today, 5 laps again, and it was too short, 1.85km. So I calibrated again, and for the 3rd test, it was too long this time, 2.17km.

First 2 tests, I was running at a quite stable pace.

For the 3rd one, I tried 3 laps at endurance pace, then one faster and one slower.  The 3 first laps were quite spot on. The faster lap was too long, and the slower one was even longer (500m instead of 400m !). It's not related to my cadence, I was slower for the last lap. I have no idea how it got so wrong.

Then finally, 5 more laps with GPS on, and that's how I got the best result, 1.98km for the 5 laps.

So from what I have tested, the accelerometer doesn't seem very reliable. It's lagging when you start, if you are still. It took 20m to start showing something. I tried to start while already running, and it was better. But it's still rarely spot on, or after 2 calibrations and a specific speed... If Garmin is using it to adjust the distance, especially for nature runs where the pace is changing often, it must be difficult to get something reliable from it... I'm not that surprised, it's not a footpod, can't be easy to measure a distance from a device worn on a wrist !

This is a useless "test" to do.

An accelerometer can't accurately measure "distance run" when you are 1) wearing the accelerometer on completely the wrong body part, and 2) running in place, ie, not actually covering any distance. How can an accelerometer have ANY idea how much distance you went, when you in fact went NO distance at all, and weren't even wearing the accelerometer on the part of your body that was supposedly covering the distance?

It's like asking people to verify the accuracy of their compass by holding a giant magnet next to it and seeing if it points toward Jupiter, then acting suprised when it doesn't. Test was done perfectly, so it must be a crap compass.

All the watch can do is guess how far you went, based on your number of armswings and the average stride length you typically have when you run. There is nothing scientific about it whatsoever. It is not using the accelerometer to measure distance. It is using the accelerometer as a pedometer. That's all. When your stride length goes up or down on a treadmill, how THE HELL would a wrist-worn accelerometer possibly know that?!? it makes no sense. PEople who are disappointed that it does a poor job of that need to understand why it's not working well, and why it never could possibly work well, and what you can do to resolve it (ie, get a footpod)

I run 15 laps on a indoor 200 m track, measured 2 260 m.

Then I run 10 laps, it measured 1 450 m.

Used the indoor running app in the watch, not treadmill were you can calibrate the distance. I can't find any way to calibrate indoor running.

C.sco said:

PEople who are disappointed that it does a poor job of that need to understand why it's not working well, and why it never could possibly work well, and what you can do to resolve it (ie, get a footpod)

Are you sure that with a normal Footpod you will have a better distance accuracy? Ok for Stryd but for normal Footpod I have some doubts because the watch match step cadence and pace like using the internal accelerometer. 

I havent done any testing, so like everyone else belive its spot on accurate!