Instant pace is not accurate and has bias towards slower than actual pace

Mattbest said:

The influence of the accelerometer in conditions of non-optimal GPS conditions terribly affects the instant pace.

The above is very easy to observe, and I observe it very regularly. On one of my local loops in a local park I have a place where the trail goes out of a densely wooded area into an open meadow, and then back into trees. The trail is flat and I run it with a reasonably constant pace. However every single time, without a failure, as I run out of the trees and into a meadow, within 30 seconds the pace on the watch speeds up by 2 min/mile, then as I exit the meadow the pace quickly drops by 2 min/mile again. That is about 20% variation in the pace.

Same here, have some places where it almost always ocurrs.

Anyone seen any pace accuracy measurements for the Venu 2 Plus? My girlfriend is interested in that watch but also in the upcoming Fenix 7.

Update: Here is graph showing device pace vs. GPS pace from a recent trail race.

The blue line shows speed measured by the device (in meters / sec). The red line shows 60 sec rolling average speed derived from the GPS track. The light gray line shows instant 1-sec GPS based speed, which are, basically, distances between points from consecutive 1-sec samples.

If the device speed / pace was correct, the blue line showing the device speed would be consistent with the GPS speed and on average, would be above the red line roughly half of the time. However, as you can see, the blue line stays under the red line most of the time, which considering that the red line represents a rolling average, doesn't make sense.

To give everyone a sense of the vertical scale:

"If the device speed / pace was correct, the blue line showing the device speed would be consistent with the GPS speed and on average, would be above the red line roughly half of the time"

I do not think this is true. We know for a fact that there are errors in the GPS signal. Suppose you are running on a completely straight line and also suppose the device speed is correct. The GPS samples will be scattered around you. If all would hit exactly on the line, device speed and GPS speed should be the same. If 1 or more sample would be on either side of the line, the GPS speed will always be higher, never lower. The watch however does not know where you are running and must use the (erroneous) GPS samples to make a path through them. This is why the device speed isn't correct either. Since most runs involves twists and turns, there will be various levels of corner-cutting and overshooting, both by the GPS speed and the device speed. But, it is extremely unlikely that the GPS speed is correct.

broyson said:

The GPS samples will be scattered around you.

No, GPS points aren't randomly scattered around the actual course. They are already corrected using the vector of movement and the data from compass and accelerometer. That is easiest to see in turns, as the plotted GPS points almost always cut through turns. If GPS samples were scattered the track would look way more wobbly. 

If anything, the path formed by GPS points tends to be over-smoothed and already too short, at least at running speeds.

silentvoyager said:

They are already corrected using the vector of movement and the data from compass and accelerometer.

No, they are not.

Ahh. I thought you where talking about the GPS samples, not the adjusted track.

gaijin said:

silentvoyager said:

They are already corrected using the vector of movement and the data from compass and accelerometer.

No, they are not.

Look at the screenshot below. Do you seriously think the below is a raw GPS output "scattered" around the path of my run?
In that case I ran fast downhill then walked slowly uphill on exactly the same trail. The path uphill (blue) is reasonably accurate while the path downhill (red) is quite wrong. To give you some scale, the track turns downhill about 25 meters too early (my path was right to left on the map before doing an out-and-back downhill and uphill section). FYI: Garmin map used by Garmin Connect isn't correct in this case, but OSM map shows this trail correctly, and it is much closer to the uphill part.

I can find dozens of examples like that where the track produced by the watch either turns too early or very significantly cuts through corners. What are the chances those would be random wobbles.

Furthermore if you look at how much the distance increases with each sample when running on a straight path, that varies from 1 meter to 5 meters when running at 3 meters per second. If the GPS track wasn't smoothed, positions would constantly wobble several meters either side. But that is definitely not the case. There is an algorithm where positions should land based on the recent speed and direction of travel, and that is used to do error correction of the path. Whether that is done in the GPS chip or in the watch software, I don't know, but I strongly suspect the watch does some additional buffering and correction, because I see behavior of Fenix 6X that I didn't see with Suunto 9, like the one shown above, despite exactly the same GPS chip.

Here a screenshot of the same part of the run from Strava to show that the above is not just a result of visualizing the track in Garmin Connect:

You are confusing correlation with causation - nothing in a GPS track rendering like you show above tells us anything about what, if any, "correction" of raw GPS data is being done, certainly nothing as definitive as "vector of movement and the data from compass and accelerometer."

If anything, you should be analyzing GPS track renderings to determine the efficacy and suitability of various GPS watch settings, e.g. Smart recording vs. Every Second recording, or GPS only vs. GPS + Galileo or GPS + GLONASS.

gaijin said:

If anything, you should be analyzing GPS track renderings to determine the efficacy and suitability of various GPS watch settings, e.g. Smart recording vs. Every Second recording, or GPS only vs. GPS + Galileo or GPS + GLONASS.

Please don't make assumptions. Obviously, GPS settings are as optimal as they can be. I have been using GPS watches for more than 10 years and I always squeeze as much accuracy out of my watch as I can. And I routinely analyze the underlying data using my own algorithms.