Current pace all wrong: possible cause

Intriguing, and news to me. Which video was that? Could you link to it? Thank you .

mcbadger https://www.youtube.com/watch?v=JxUPetG3JMI

Interesting video. This is just based on my skimming on the 7 minute video, so I apologize if I missed something.

I didn't watch the whole video, but I did watch the pre-run talk:
https://youtu.be/JxUPetG3JMI?t=114

I didn't interpret what he said the same way you did. He didn't mention battery saving behaviour, but what he did say that is that most manufacturers use the internal accelerometer to smooth out GPS pace (which is news to me). Meaning that the motion of your wrist / arm swing would be used to correct GPS discrepancies.

I think his point was that he thinks holding his arm steady (so you can see the screen) would make the current pace less precise.

I don't think he ever said that he would lift his wrist to keep the watch awake.

What I think he actually said was that he was going to "do some checks to make sure that he wasn't impacting [current pace] by trying to record this". So he was planning on monitoring his current pace to make sure that his lack of arm swing wasn't messing up the current pace.

IMO: he's not saying constantly looking at his watch makes the pace better. It's the opposite, where he's saying that not swinging his arms naturally is possibly making the pace worse.

(^ You can see this pretty dramatically if you run without GPS, of course, like on a treadmill. If you are constantly checking your watch, your pace is going to be off, since that arm isn't swinging naturally anymore.)

When he speeds up to about 6:30/mile pace, he mentions how he thinks that the current pace would've "caught up" more quickly, had he been swinging his arms:
https://youtu.be/JxUPetG3JMI?t=214

Further to the point, DC Rainmaker said nothing about that behaviour being applied only to Fenix 5. e.g. I would expect the 935 to be exactly the same, as the software is very similar. So whatever the behaviour is, I don't think it's unique to Fenix 5.

---

Anyway, I have no idea what the real reason for pace inaccuracies is (I don't own a Fenix 5, but I do own a 935), but I would guess that it's because of the metal body interfering with GPS signals, at least based on what people in the forums have said.

I raced a 5K in the city with a friend -- my 935 was within 2% of the 5.00 km distance, while his fenix 5 was 10% off.... We ran the same race under the same conditions, although we didn't finish at the same time. But I don't know how to account for such a dramatic difference, except that the "metal body interference" theory makes the most sense to me.

Thanks WillNorthYork for the interesting answer.
My bad about the video. Today, while I was running and noticing that particular behavior, this video that I saw some weeks ago came into my mind, but I wasn't remembering it right.
What I'm quite certain about though, is the "wrist lifting" thing that I tried many times in my 10km run of this morning. I was running with a friend who owns another, more accurate, watch so I had his to compare with. The current pace was, as always, incredibly wrong. It's something that drives me mad how bad it works. All the times that I tried to lift the watch 2-3 times in roughly 10 secs time, I could witness the current pace slowly adjusting to a probable value.
For example... while running steady I check the pace, it displays 5:30 min/km. I know that's impossible, as I know I'm running faster. So I ask my buddy and he reads 4:45. Then I do the lifting trick, and in 10-15 seconds time I see my pace going down till roughly the same value of my friend's.
This happened 4-5 times just today. I don't know what it means, but it's got to mean something.
I'd be very interested if someone else with a Fenix could try this and tell me if I'm delusional or not.
The metal body thing.. well, it's a thing. Perhaps it messes up the signal or just makes it worse but not beyond recovery. The point is we cannot change body, so maybe there is something left that Garmin could do on the firmware side.

Edit: I also tried Race Screen data field to be able to use a current pace calculated on more seconds but the results are the same (except for the time that it takes to reach the actual value).
I'll make more tests in the days to come anytime I got the chance to run with a buddy who has a solid GPS watch.

Yeah, that’s pretty interesting. Wish I knew what the problem was. The lifting trick could be giving you better line of sight with the satellites, maybe? Who knows. Or maybe if you hold your arm steady you get a better signal.

Yeah, DCR mentioned a lot of the “GPS magic” is really in the software, in reference to ppl constantly worrying about changing GPS chipsets, etc.

All I can say is that I have a 935 (supposedly the same or similar firmware) and I’ve never had problems with instant pace, except for when I didn’t have a “true” GPS fix at the beginning of an activity. (Current Garmins seem to lie about when a GPS fix is available, so I wait for instant pace to settle down to “—:—“ before starting to run. If it’s fluctuating wildly while I’m standing still, then I know I have to wait.)

And I have the one anecdotal data point from my race with my friend who has a Fenix 5. I think if I look at his Strava activity from then, I’ll be able to see him running on top of buildings and zig zagging across roads, according to the map....

Today I had a 7+k run and since I was alone and bored, I made many small tests on the subject. Some of them perhaps valuable, some not. What it’s important to say though is that every test result, wether useful or not, has been very consistent with the others.
I was running in a (mostly) flat trail in the woods. Sky clearance was very good tho, as trees were small, very sparse and with no leaves at all. Also, no clouds.
I’m a metric guy, so I use min/km for pace.
I selected three cases to show you. The first 2 have been repeated more than once, always giving consistent results.

Test 1:
a. I’m running normally at steady pace without looking at the screen in the last 5 minutes. While running, I peek at the watch screen without making any changes to the normal wrist movement. It displays 7:00, quite steady. I know that’s not my actual pace.
b. I lift the wrist and keep it still, facing the sky, 10cm from my chest. The pace begins immediately to adjust, and in 10-15 seconds reaches 4:45 and keeps that value almost steadily (this is also the pace I thought I was running)
c. I go back to running normally and peeking at the screen without modifying the way i move my arm, like in point (a). The pace starts to go back to the former value, but very slowly. After roughly 2 minutes I can see a pace of 5:30.
d. I repeat point (b) and the watch behaves exactly like it did at point (b).

Test 2:
a. I’m running normally at steady pace without looking at the screen in the last 2-3 minutes. I peek at the watch screen without making any changes to the normal wrist movement. It displays 6:00, quite steady. I know that’s not my actual pace.
b. I keep moving my arms the same way, but I rotate my wrist in order to have the watch face up (and singing “run like an Egyptian..” :-D). The watch behaves like point (b) of the 1st test, reaching and keeping 4:45
c. As in point (c) of the 1st test, I go back to running normally and peeking at the screen without modifying the way I move my arm. I notice the pace starting to (very slowly) go back again to the former wrong value.

Test 3: this one I want to repeat ‘cause it’s a bit tricky to perform (and also a bit awkward)
a. I’m running normally at steady pace without looking at the screen in the last 2-3 minutes. , I peek at the watch screen without making any changes to the normal wrist movement and I keep note of the pace shown.
b. I lift the wrist and keep it still, facing DOWN, 10cm from my head. While keeping it this way, I try and peek the pace. It seems not to change. Again, i’m not 100% sure like I am about the first 2 tests.
c. I rotate the wrist wrist in order to have the watch face up. Pace gets quickly adjusted.

Conclusions:
It looks pretty straitforward to me: cadence and wrist swinging seems not to make any changes to the detected pace. That just works well.
My thoughts about battery saving were just plain wrong.
Screen orientation seems to be the deal breaker here. Maybe the metallic body filters so strongly the signal that you really need to keep the watch facing up.

I believe this proves something.
I would be very interested if someone was willing to reproduce the same behavior, at least for the first 2 tests.

Any thoughts?

Conclusions:
It looks pretty straitforward to me: cadence and wrist swinging seems not to make any changes to the detected pace. That just works well.
My thoughts about battery saving were just plain wrong.
Screen orientation seems to be the deal breaker here. Maybe the metallic body filters so strongly the signal that you really need to keep the watch facing up.

That's pretty interesting, and makes a lot more sense than any of my guesses (one was sort of close, but not really).

I also sort of doubted what DCR said about the accel being used to filter GPS pace, but then again, he has inside information that we don't. But I never saw anything like that in any of my runs. I guess it's possible that other companies might do this, though.

That's pretty great tho, because you seem to have confirmed what everyone has been speculating, and also discovered some new information.

Wish I had an F5 to help you test this.

I guess it's possible that other companies might do this, though.

Suunto clearly advertise that they augment GPS speed with wrist based speed in their Ambit series watches; they call it Fused Speed https://www.suunto.com/en-gb/Support...es/fusedspeed/ It’s entirely possible that Garmin have their own version of this, which makes it reasonable for DCR to make the comments/observations he has. In my experience of running with the Fenix 3, Fenix 5 and now Fenix 5+, the pace is OK with a good view of the sky and not so good when I run nearer to trees or other obstructions to GPS; that would suggest to me that speed on these watches is purely (or at least predominantly) from GPS and not augmented by wrist derived speed. Only a Garmin software engineer could give you the definitive answer, but that information is probably commercially sensitive and unlikely to be released.

Crispin_Ellisdon yeah I've heard of Fused Speed and it sounds interesting. But you would imagine that Garmin would also announce their own version of "fused speed" as a marketing bullet point, if they in fact implemented it. Especially with all the complaints about GPS accuracy that pop up over the years. Even if they didn't tout it as a "feature", you might expect them to just say "our pace is better than GPS alone because ..."

As far as I can tell, the biggest user-visible change to Garmin made to pace within the last few years is to start rounding instant pace to the nearest 5 seconds. I'm sure lots of changes have been made under the hood tho.

As far as DCR's remarks go, it is not unreasonable for him to say that, but it's hard to interpret whether he is just talking generally (as in it's possible) or whether he has some special insider knowledge.

I agree that it's possible and it's hard for us to know for sure, but I've never seen any evidence in my own real-world running to suggest that's the case, either.

EDIT: Another interpretation of his remarks could be "If you lose your GPS signal, then the wrist accel takes over", which is certainly true. I didn't interpret it that way tho, especially when he sped up to 6:30/mile and made the remark about how he felt his instant pace would've reacted faster had he been swinging his arm naturally.