Automatic Workout Import and Hiding the Workout Pill

Features Discussion
1

Is it possible for Arc to automatically import workouts rather getting me to tap to import each one, wait, then tap again to decide if want it in my timeline?

I feel like I should be able have Arc auto import all workouts, then I can decide if I want them in my timeline, or I just want them deleted/forgotten (from Arc). Could even put that flow into the unconfirmed items flow.

Also is there anyway to hide the little red workout pill? It sits there demanding my attention on each day, but there’s nothing for me to do, I’ve already imported (or decided not to import) each workout. It feels like after the import decision has been made, workouts should disappear into triple dot, or become some other UI elements that’s less intrusive.

P.S. love the new app updates. Huge improvements all round.

2

I think this should definitely be on the todos list for a future update. Though at this stage I’m stepping carefully, going for “minimum viable product” functionality for workout imports, for a few reasons.

  • The imported data is actually pretty large, at least for the Apple Watch. So to avoid bloating the database I want to only keep around imported data that’s definitely going to be used. I’m also going to make it auto-delete imported workouts that haven’t been added to the timeline, after some period of time. (They can always be reimported later).

  • The imported data isn’t necessarily superior to the data recorded by Arc itself. Although the sampling frequency is much higher (at least from the Apple Watch, which is faking up an extremely high samples per min, which I have a separate rant about), the data doesn’t include accelerometer data, so it can’t be auto classified by Arc’s activity type classifiers. The data is lat/long/altitude only.

  • The imported data also can’t (yet) be edited like Arc recorded data - no segment splitting, visit extraction, etc. I want to try carefully adding some of that functionality eventually, but it needs to be doable without breaking the ability to toggle back to the Arc recorded data (ie “remove from timeline”), which creates some timeline processing headaches.

Yeah that would be great! Something to work towards in future, if I do manage to get the imported data to work nicely with timeline processing in general.

I’ve been pondering this one too. There needs to be a way to get back to the workouts list view, so a button needs to continue to exist somewhere. It could be moved to the ellipsis/options menu, but it still makes sense to have an initial button in the current place, until each workout has been viewed/considered. I haven’t settled on a best way to handle that yet.

Thanks! Hopefully another update coming soon.

I’ve got a Dark Mode update almost ready, which will probably be in the next update. And there’s also a replacement of the activity types classifiers in the works, making us of Core ML instead of my homemade classifiers. I’ve got Core ML classifiers working really nicely in private beta builds, but need to do the extra work to get them integrated into the shared classifier trees,. That one might be a few updates away.

3

Oh interesting, didn’t know that. You got any guidance on when it’s worth importing vs when it isn’t?

4

Heh. You can’t know whether it’s worth importing until you’ve already imported :grin:

Well, there’s the distinction between “imported” and “added to timeline”. Importing it just adds it to the database, but in a “disabled” state. Then it’s a matter of whether you actually want to use it or not, ie add it to the timeline or not.

I’m mostly only adding to timeline the workouts that have a cleaner / more complete looking path line on the map versus the Arc recorded data.

In practice that means I add the short 2-3 minute walks from home to Starbucks each day, because Arc’s data is always a bit poor right after waking from sleep, so for a short trip like that the first half of the trip is sparse in samples and not terribly accurate. Where as the workout (I start a walking workout on my watch before leaving home) will have full data right from the moment I start the workout.

IMG_8824.PNG
IMG_8824.PNG1170×2532 104 KB

There’s a fairly typical example. The green being the imported workout, with (mostly) proper data from the start on the right, while the grey is Arc’s data, which is… okay in this case the Arc data follows the road more accurately. Perhaps not the best example.

But yeah, I find it especially useful in those cases where Arc waking from sleep meant the start of the trip is data-sparse, where as a manually started workout doesn’t suffer from that slow/gradual wakeup.

Outside of that, I’ve been finding that the Arc recorded data is usually better. Even though there’s more samples from the workout data, it’s more aggressively smoothed than Arc’s own data, so it’s often a less accurate path line, with sharp corners too aggressively smoothed out, etc.

I’m curious to see whether that’ll change with the Apple Watch Ultra and its dual band GPS. And other wearables that record workouts might be producing less aggressively smoothed paths, so might actually be beating the Apple Watch already on this. Though I’ve no evidence of that yet - just a hunch.

1 Like