Differences in measured pace and report pace

Ah OK, thanks. What you are saying is I cannot expect it to be more reliable. I've noted that it tends to underestimate my pace when there are trees around, and sometimes overestimate it when there isn't. Could be do to the accelerometer then. A bit strange that garmin doesn't post-normalize the data by the actual time elapsed per the distance, that would seem quite easy.

Would this change a lot if I upgraded to e.g. the 970 or 965?

As an aside , is your HRM-PRO set to pace / distance INDOOR only?

Apparently this option no longer exists. On this version 8.90 it says it only transmits accelerometer data when the GPS is off

3684016 said:

I've noted that it tends to underestimate my pace when there are trees around, and sometimes overestimate it when there isn't

Yeah in that case I would say it's due to GPS error. It's easy to see issues with unrealistically fast instant pace if you start a run around tall buildings, especially if you don't wait for a solid green GPS fix. Even if I have a solid green ring, in my experience that doesn't always truly equal a solid fix - even if I'm standing still, instant pace indicates that I'm running super fast (like 0:59/km), and my position on the map is changing. I usually find that I typically have wait an additional 30 seconds after the green ring appears for instant pace to settle down to "0:00", then I can start my run. (Sometimes it takes up to 2 minutes)

Of course, this sort of GPS error tends to be less significant over the course of a long activity, as opposed to the specific short section where it actually occurs.

Ironically this makes the case that instant/current pace should not be based only on GPS data, which does mean that we can't expect the instant paces to always "make sense" when compared to the average pace for a single short lap (for example). I think I can say for sure that if you averaged all the instant paces/speeds, the result would probably not equal the activity average pace (which is based on total time and distance), even if you had access to the raw data in the FIT file with the max amount of precision.

3684016 said:

A bit strange that garmin doesn't post-normalize the data by the actual time elapsed per the distance, that would seem quite easy.

Well, as I said, at least on older devices (including my FR955), all of the average/max pace metrics aren't even recorded on the watch, but are calculated in Connect after the fact. So things like activity average pace and lap average pace are all based timer time and distance. I don't use the word "elapsed" here because that has a special meaning in Garmin land - it means time between the activity start and activity end, disregarding pauses. In contrast, "timer time" / "total time" / "time" means time excluding pauses, same as the Timer activity data field.

If you are suggesting that Garmin should somehow normalize the instant/current pace that was recorded by the device based on total time and distance, it's an interesting idea, but I'm not sure how it would work in practice.

For example, for very short laps, it's common (for me) to see obviously wrong data like max lap pace (based on instant pace) being slower than average lap pace (based lap time and distance) . In this case, how should Garmin normalize the data? Linearly increase all the instant pace values for that lap so that they average out to the lap pace? While that data might "look right", is there any guarantee it will be right? Should Garmin apply similar corrections to every lap? What happens if those corrections aren't uniform and/or they don't result in instant pace "looking right" for the entire activity?

Furthermore, if Garmin thought that instant/current pace could be based solely on time and GPS distance, wouldn't they have implemented it that way on the watch in the first place? In this way, there would be no need to normalize the data post-activity, since it would already be consistent with time and GPS distance.

But on the contrary, they apparently augment it with accel data, which says otherwise. Again, I don't think GPS is very reliable over very short distances, which is why GPS data alone is probably not good enough for instance/current pace .

I've sort of made my peace with the fact that instant pace isn't really 100% correlated with average pace, or perhaps even reality. Over the course of a long run, I think it's good enough.

On a related note, if you export your activity FIT file (Connect website > activity > gear icon > Export File) and open it in the free 3rd-party website www.runalyze.com, there's a "best sub segments" feature which shows you your best paces for a variety of times and distances in the activity, including very short times like 1 second. When I use this feature, I often see a "best 1s pace" which is faster than Garmin's calculated max pace in Connect for the same activity. I think the discrepancy here comes from the fact that runalyze best pace is probably based on time and distance via GPS (as it would have to be for the longer sub-segments), while Garmin's max pace is simply the fastest recorded instant pace. Again this shows that instant/current pace is not 100% based on GPS distance.

TL;DR I think that instant/current pace is not an exact science. 

3684016 said:

Apparently this option no longer exists. On this version 8.90 it says it only transmits accelerometer data when the GPS is off

Furthermore, I've seen user complaints that pace/distance from HRM-PRO (and similar chest straps) is only used for indoor running activities (and not team sports like tennis, for example).

It's almost as if they don't think the accel in the chest strap is accurate enough for many purposes.

In contrast they'll still accept pace/distance from a footpod under any circumstances. (Ofc this doesn't necessarily mean that the data from a generic footpod is any better.)

3684016 said:

Would this change a lot if I upgraded to e.g. the 970 or 965?

I doubt it. Just looking at the forums I see even more complaints about Fenix 8 (the latest generation of that line). FR970 software is pretty close to Fenix 8.

As for FR965, my watch (FR955) has similar software (which is also similar to Fenix 7), and I don't think there's anything especially great (or bad) about instant pace for my watch.

Probably the biggest obvious change Garmin made to instant pace over the years is to stop displaying it on the watch to the nearest 1 second, but to round it to the nearest 5 seconds instead. If they've made changes to the actual algorithm, they're not telling us exactly what those changes are.