I noticed a discrepancy between how Garmin reports a session and Athletic does, which, in my opinion, misleads the AI coach. Here is a Z2 run as an example: “The aim of this session is to develop your aerobic base. Easy/steady 38′ [8:12 - 9:18] min/mile, [113-130]HR”
Garmin report it like that with a 92% execution score
The raw file is probably close to what Atheletica displays, despite smoothing over 3 seconds. Should atheletica smooth over a longer interval or am I the only one having this issue? If so, I need to change my watch, Garmin Epic Gen 2, probably.
In addition, I have 2 proposals for improvement for the AI coach
1- Teach it to adjust its feedback if there is some elevation gain in a run, moreover, on a workout pace based.
2—Teach it to provide feedback based on how the workout was prescribed to be executed. If it is prescribed by HR, the feedback should not be about pace zone but HR zone. It never bases its feedback on HR zone despite the prescription.
Thanks @Jerome … I’ll let @Stef comment here as I believe his latest version soon to be released should address some of your comments. Great ideas also with thanks!
@Jerome thanks! You’ve hit the hearth of our new improvements indeed. When some elevation threshold is met, the AI coach will prefer commenting on HR, until we sort out a better solution for the GAP computation (if any is possible)
In general, for the time being, the coach will consider the best compliance between HR and pace. In your case, it’ll say you’ve nailed the session!
We want to move towards providing feedback based only on the user’s preferences - as that is the ideal one - but it’s not easy to do at the moment. We are discussing this now, but may need some other UI/UX changes as well. That’s the goal any way!
My running paces in the running pace time in zones chart today were way off as well. It included my non-moving paces (which obviously resulted in more time in zone 1 than I actually did) and it also included speedier paces in zones 3b, 4, 5 and 6 that I didn’t hit on my run. Strangely the sessions chart doesn’t show the lines hitting those zones when you add the pace zone overlay.
I had something similar on Saturday. Apparently I was hitting zone 7! As much as I would love to, I dont think I have ever hit 4.25-2.34 with a 128 hr…
Here you go…Athletica pace results on top as requested showing my pacing all over the place. Adding quite a few other charts from other sites to compare.
Athletica Pace zones vs planned chart details:
Z1: 14:10 to 9:55 /min.
Actual time: 51:14
Z2 9:55 to 8:45
Actual time: 55:36
Z3A 8:45 to 8:28
Actual time: 12:26
Z3B 8:28 to 7:58
Actual time: 15:45
Z4 7:58 to 7:12
Actual time 7:18
Z5 7:12 to 6:55
Actual time :40
Z6 6:55 to 5:46
Actual time :35
Athletica zone chart with pace overlay (not smoothed:
This was actually a VERY steady and easy zone 2 type workout for heart rate, which I stayed in just about the whole time except for a few zone 1 minutes to start. Pace wise I was pretty even the whole way with a slightly slower start to warm-up and a walk up a tiny hill. Was just one of those runs where every time you look at your watch over almost 2 1/2 hours it’s pretty much always around the same pace number and HR.
Thanks, we are well aware of this pace related issue. We will be releasing the improvements to the AI Coach feedback and then work on this asap as there are several possibilities on how to handle this and we want to devolve the right time to choose the best one. Unfortunately, we are very busy with a lot of new things at the moment (stay tuned!), but we understand the frustration
I realize this looks like the old AI coach feedback issue, but there also seems to be an issue with the data itself which is misleading the coach. I too got some surprising zone compliance feedback today after a 5x1km interval session, and decided to look at the data. The Athletica speed data is all over the place, going from 3:51 min/km to 4:58 min/km within the same interval, while I’m sure my pace never fluctuated that much and Garmin very much agrees. So the Athletica data is somehow different from Garmin’s data, even with smoothing disabled. See attached examples:
Nice and even pace in Garmin, fluctuating wildly in Athletica.
I’d say 99% of my intervals were in the correct zone 4 range, but Athletica thinks I spent less than 10 minutes there. This isn’t an AI coach issue, but a data issue.
Yes, working on this at the moment. Thanks for keeping the influx of data coming guys, really appreciated.
Please, do not forge to report your session date and activity type (run, bike, etc) when you write us. It is a very small thing that makes our job much much easier
I’ve been using Athletica for about 1.5 months now, and during that time I’ve never seen the correct pace displayed for any of my runs.
I normally run with Stryd on my Apple Watch, syncing the runs to Strava and then into Athletica. I’ve also tried running without Stryd—just using the native Workout app on the Apple Watch and syncing from there to Strava and then into Athletica. Regardless of the setup, the pace data in Athletica is consistently inaccurate.
The time spent in each pace zone doesn’t reflect my actual run efforts and differs significantly from what I see in all the other apps. As a result, the pace-based analysis and feedback currently feel pretty useless—and honestly, a bit frustrating.
This is indeed a delicate issue. Currently, we’re calculating pace using a method that’s theoretically the most accurate: we compute the distance based on GPS coordinates and divide it by the elapsed time to derive speed or pace. While this approach is mathematically sound, it depends heavily on GPS signal accuracy — which, in real-world conditions, tends to be noisier than we had initially anticipated.
Devices like Garmin often use proprietary algorithms that combine multiple data sources (e.g. inertial sensors, historical patterns) to produce more stable pace readings. As a result, the values they report can differ significantly from those we calculate, even though they may appear more consistent to users. Whether one method is more “accurate” ultimately comes down to user expectations and trust in the brand’s processing.
We understand your frustration and are actively exploring the option to use the pace/speed values provided directly by the device, rather than computing them from raw GPS data. This change should lead to more consistent results.
We’ll keep you updated as soon as we begin implementing this — likely within the next few weeks, as we’re currently focused on some higher-priority tasks. Thanks for your patience and support, and stay tuned!
Different ways of calculating pace from GPS data could explain some discrepancies—but yesterday I did my first indoor treadmill run using Athletica, and the pace is still way off.
It was a structured session (4 × 2 minutes at around 4:00/km pace), and all four intervals were run at nearly identical, steady speeds. Given the treadmill environment, this should be one of the easiest cases for accurate pace tracking.
The data flow was the same as always: Stryd on Apple Watch → Strava → Athletica. Both the Stryd and Strava apps show consistent, accurate pace data. But in Athletica, the pace looks almost random—showing speeds that are clearly unrealistic.
So at this point, it doesn’t seem to be a GPS or outdoor signal issue. It feels like something is going wrong in how Athletica processes or interprets pace data during import.
