Accelerometer, pace and GPS

I don't think you can deactivate it

You can calibrate the accelerometer, by doing a treadmill run of more than a mile. Then when you save it, there is an option. But I tried it, then went for some laps on a track, with various speeds. It was better than before, but still wrong and with some big variability. So in my view, the accelerometer is not very reliable, and it damages the pace estimation quite a lot. The more you search about it, the more you see that the best way to have a reliable pace with the Fenix is to get a footpod, ideally a Stryd. As much I love my Fenix for various reasons, the instant pace is simply very bad, and not reliable. No setting will make it fine in all conditions. 

As mentioned in the manual (in relation to indoor activities - but also applies when GPS is poor outside) https://www8.garmin.com/manuals/webhelp/fenix66s6xpro/EN-GB/GUID-3A4C7C6C-1FE3-4EB5-B38E-3F744A5C1F00.html the accelerometer based pace is calibrated during outdoor runs and walks and it can be messed up by arm movements that are inconsistent with the normal rhythm of walking/running:

"The fēnix® device can be used for training indoors, such as running on an indoor track or using a stationary bike. GPS is turned off for indoor activities.

When running or walking with GPS turned off, speed, distance, and cadence are calculated using the accelerometer in the device. The accelerometer is self-calibrating. The accuracy of the speed, distance, and cadence data improves after a few outdoor runs or walks using GPS.

TIP: Holding the handrails of the treadmill reduces accuracy. You can use an optional foot pod to record pace, distance, and cadence."

I've never been able to agree with people who look back with rose-tinted glasses at older watches and say that instant pace was better. It's a pity that we don't have access to the 305, 310, 910 etc forums because the same complaints were there.

Over the years Garmin have made improvements to instant pace by incorporating the accelerometer to provide some assistance to the GPS as Crispin notes but there is no escaping the fact that if you want an accurate reading of your pace at an exact moment in time then you should get a footpod.

Thanks. Not sure that I said anything about "good old days with old watches"...

Anyway, if there is no way to deactivate the accelerometer to see if it could be better for my needs or situation, i suppose that I have no choice. 

Yes, the 310, 405, 610 were surely not perfect for instant pace. But at least, they were quite usable. The GPS performances of the SirfStar chipset was better than the Sony. And the instant pace is unfortunately the same. The use of the accelerometer to smoothen it is really not working at all. I love my Fenix, I enjoy a lot of its features, but it's impossible to defend this indicator, I'm sorry. We are not talking about a difference of a few seconds per minute here. It was the error margin with the old watches, most of the time. I remember using them for interval training on a track, and I knew it was not crazy precise. But it was still quite usable. 

Now, it's not the case anymore, and the difference between the distance calculated and the instant pace is simply NOT logical, it adds a lot to the confusion. There are enough evidences from many runners to see it. And it's simple logic. Last example, from this week. I'm running with 2 watches to compare the data in the exact same condition. The Fenix 6 and the Suunto 9 (also using the accelerometer to smoothen the instant pace). 

I run the first kilometer in 5:50. I double checked that from the map, retracing my run (in perfectly straight streets, it's easy). And it was a very easy run, I know my pace well enough to know I was running around 10km/h. 

The F6 first lap says 5:50 (GPS trace was not perfect, but Garmin's distance calculation tries to counter balance it, and it did a good job on that first km).
The S9 says 5:38 (GPS trace was not perfect, adding to the distance)

Now if I look at the instant pace displayed by the Garmin, it just doesn't make any sense :

The Suunto was not great, giving me an average instant pace faster than what I was doing, 5:22 (I removed the first few seconds, since there was no instant pace, I guess it needs a few seconds of data from the accelerometer to give you a pace). The Garmin, 7:16/km ! I checked the chart, my pace is simply NEVER under 6:00/km and is mostly between 7:00 and 8:00. Mathematically speaking, it's impossible to have an average pace of 5:50 with those numbers. And I know how I was running, I was not running at 7:30/km. We are not talking about a difference of 5, or 10 or even 20 secondes per kilometer. It's 1 to 2 minutes. And it's easy to know I wasn't running that slowly. 

The idea to use the accelerometer is good. It helps when there is no GPS signal. It's properly done by Suunto from what I see, it's more reliable (still not perfect, far from it). But I'm sorry, Garmin is not really doing good with that. The old watches were significantly better, because their GPS was better. They were not perfect, but I never saw such a non sense in terms of data. At least, it was consistant with the lap distance. Now, it's NOT the case anymore. So I prefer to look only at the lap distance, and even this one can be significantly false, from the mix of GPS error and the poor use of the accelerometer (that's another topic, the wrong distance of the F6, compared to the old watches again. but probably also related to the use of the accelerometer to adjust the distance).

And I'm showing an example, but I have ran with 2 watches for a month already, to gather more and more data. Other watches are not perfect either, being the old ones, or the other recent ones (like the Suunto). But if there are 3 things that are really not performant on the F6, it would say it's the distance calculation, the instant pace, and the sleep tracking. The optical heart rate also. But there are tons of things I love and I'm keeping my F6, it's a great watch. It's simply impossible to really defend this instant pace. If AT LEAST it was consistant with the lap values ! But it's not. The new GPS chipset from Sony is wonderful for the battery. But it gives some headaches from accuracy, and thus pace & distance. And the idea of improving it with the accelerometer is good, but it's poorly implemented, unfortunately. From the wrist, it's understandable. It's easier with a footpod. So I wish Garmin was offering an option NOT to use it. It's an idea to improve things that seems to actually make them worse. And I tried to look at the data carefully, trying not to be judgmental from a feeling only. That's why I bought another watch to compare, in real time. And I'm recording tons of runs, to have enough data and not just a couple of runs. I have all the runs of my old 405 and 610, and at least the pace chart is consistant with the laps. It makes some sense, even if it's not smooth enough. Now, it doesn't make much sense, unfortunately. 

Great summary, Thanks! Keeping my F6X too, but always use chest strap for HR and and STRYD for distance and pace...

My first running watch was a SiRF Star chipped Garmin 405CX before moving on to Fenix 2, 3, 5, 5+ and 6x. I even dragged my 405CX out of retirement when I got the F6X (had to fit the fabric band as the rubber one had perished), to see if ‘rose tinted spectacles’ of the past were correct (the 405CX used to be much better than the Fenix 2 in track, distance and pace accuracy side by side tests) . I was surprised though; the 6X laid down a more accurate (less smooth, but more accurate) GPS track, total distance was closer to hand measured on the 6X and instant pace was not great from either anywhere near trees, buildings and other obstructions/reflections. So where does that leave us; Get a foot pod! (which is exactly what I did, back when I had the 405CX anyway)

If you can’t stomach the cost of a Stryd pod, a much cheaper calibrated Garmin SDM4 foot pod gives 98+% accuracy for instant pace, which is more than good enough for most people’s needs pacing during intervals or races (and unlike the accelerometer calibration numbers that are hidden inside the watch, the foot pod calibration is fully visible and user controllable via the sensor setting in the watch). The benefits of the extra cost of Stryd is lifting to 99+% accuracy of pace and distance as well as adding their running power stuff (up to you if that’s worth the extra cost to you).

There are moments when it's fine, but really, most of the time, it's really too low. And maybe your watch is using the data from the HRM Run. And I think it also depends of the GPS accuracy. I'm running in streets with not too tall building, but still, it's not perfect for the GPS. But if we have to rely on perfect GPS conditions to have a decent instant pace, it's not good. It should be acceptable in most conditions, and for instance Suunto is doing better. I didn't check for Polar / Coros. But it's possible to do better. It will also differ from the way you are moving your arm, possibly. An accelerometer on the wrist seems much more difficult to fine tune, compared to one on the foot. I had better instant pace with my old Garmin footpod, for sure. 

I'm very tempted to buy a Stryd, it seems to be the most accurate sensor for instant pace and even distance, hands down. But it's 200$++ ! That's a lot of money to have better numbers, that won't really make me run faster or further ;)