“How many miles did you get?” is a common question at the end of a group run on the trails in the heavily wooded sections west of the Cascade Mountains. A group of five different runners will likely have five completely different distances, even though the same route is done simultaneously. The distance shorted on a GPS watch is commonly known as the “tree tax.” Now that spring is coming up, the deciduous trees hidden among the evergreens will begin to regrow their leaves, causing even more of a canopy for a runner’s watch signal to penetrate and reach the satellites circulating the earth.
I’m definitely not a watch expert. I care about the app’s ease of use, battery life, distance, and ascent accuracy. I’ve owned five GPS watches since 2015.
- My first was a Garmin 310XT. This watch was a beast and existed before apps. Even though it’s one of the first generations of watches, it’s the most accurate of all the watches I’ve worn. The battery life slowly made this watch unusable; otherwise, I still might be using it. It also didn’t work with any Garmin apps.
- My next watch was the Garmin 235. I was intrigued by the wrist HRM, but that never seemed to work for me. Being on average 15% short on each run in Forest Park was very frustrating as well.
- I found a sale for the Epson ProSense for $100, so I decided to give that a shot. It was more accurate than the Garmin 235, but the app was horrible. Many of my runs didn’t upload to Strava, which is a definite no-go. It also had trouble finding a satellite on some runs.
- A friend of mine ran the Wildwood end to end with me wearing a Suunto Ambit 3 Peak and almost got the entire 30 miles on his watch. I bought one the next day and still use this watch. I find it about 5% short on most runs in Forest Park. However, I’m having trouble with the Suunto app and can’t change the settings the way I want on the watch.
- I also bought a Suunto 9 for the 2019 Mountain Lakes 100 for the 30-hour battery life. I stopped using it because it was continually 20% short on both distance and ascent. Please note that this Sunnto 9 did not have barometric altitude, which impacts elevation data. So I am back to the Suunto Ambit 3 Peak.
So I decided to research GPS watches and thought I’d share it. I looked up runners of Wildwood E2E in April and May of 2020. I only found 26 results on Strava, so I added four of my runs. It didn’t reveal anything significant. So I thought I’d compare the Portland Trail Series competitors on June 19, 2019, a race I ran. I found 34 results on Strava, yet no watch seemed to work better than any other. I wanted to try one more race, so I looked up the 2019 Stumptown Half Marathon and found 38 results. Still nothing. You can view the spreadsheet here.
The best way to analyze watches is to wear them all at once and compare them. I didn’t have that option, so I sent my information to a fellow trail runner and creator of runs4cache, Pete Carleson. Pete is an engineer, so I thought he would figure out how to analyze my data.
Pete ended up making a big table of all watch brands as a portion of the expected distance. So “1” would be the ideal distance, and 0.9, for example, would mean it measured short. Note Strava is just a phone app, but many runners use it on a wide range of Android and iOS phones.
Figure 1. Showing combined data by source as shown on Strava, in a box and whisker plot. The horizontal line shows the median value, the box shows the 25th percentile to 75th percentile, and the error bars indicate the range. There was one Max Outlier, lucky? Or perhaps a footpod user. Strava does not indicate the presence of a footpod. Data from 96 activities over 3 courses with 28 different sources. Sources with only 1 activity (looking at you TomTom, Polar, Samsung Health app) were omitted. The quantity in each source type is: Apple Watch (11), Garmin (43), Strava(30), Suunto (8), Coros (4)
“People talk about how lousy phones (using Strava app) are, but there is enough data there to show it can be better but also has a lot of variances,” added Pete. “I’ve nearly thrown up my hands on this because of the personal calibration that the watches do with the onboard accelerometer, making individual watch comparisons challenging.”
Pete started using the Stryd footpod to measure runs and used it on a Wildwood end-to-end last New Year’s Eve. “I got 30 miles on my watch from Stryd. I exported the manual lap data and showed it (in the table) below. I think the 3.0 (mile) marker is missing, and I missed 9.0. The 3.0 is on a tree just before the Walker bridge, and I may have been looking for it too late. I manually hit the lap button when crossing the trees with the diamond mile marker. The Walker Bridge re-route seems to cost about 0.14-0.15 miles. From the 4.0 marker to the 30.0-mile marker, the footpod reports 26.03 miles. Last year with the same shoes in the same direction, it measured 30.04 miles, but that included some walking around on the watch at 53rd ave and FL1. This year when I stepped off the trail, I also stopped the watch. A mile is 1609.344 meters, so I left that in the table so you could see how many meters it was off on each split, or perhaps how far the TREE was off.“
|🔷||split (meters)||split (mi)||total (mi)|
Table 1. Column 1 is the blue diamond mile marker, column 2 is the manually lapped distance value from Garmin FIT file in meters, column 3 is the split distance between markers in miles, and column 4 is the cumulative distance in miles.
“I did one of my neighborhood loops even with no Stryd and GPS off, so just the wrist accelerometer, and it came out accurate. I’ll need to try this no GPS set up on the trails, but the watch accelerometer calibration is from road pace, not trail pace. Maybe this is one of the contributions to the crazy distances in FP. One thing that is telling is to do the Correct Distance function in Strava, that recalculation is only on GPS points, and it’s always different from what the watch reports from its combination of onboard accelerometer and GPS.”
One E2E I did was with an Epson watch. The watch never caught a GPS signal, but I didn’t know that until I uploaded it to Strava. The wrist accelerometer got me 30.768 miles https://www.strava.com/activities/1482606070. So it calculated the distance better without GPS.
After running the Firelane Frenzy 50k last month, I noticed that almost all watches were very close to getting the 30.44 miles. My only guesses were that the firelanes are more exposed, and they don’t have any switchbacks, making access to the GPS satellites much easier.
So basically, we’ve declared there really is no watch that works best under tree cover. Stryd combined with a compatible watch is the only guarantee it will get the most accurate distance. Or wait until 2029.