Apparently Garmin added OAuth2 to their Connect API: New apps can already use it, existing apps can't migrate yet. More info "this fall".
As a developer, this makes me very happy (maybe more than it should, but OAuth1 really is a pain to work with).
Docs: developerportal.garmin.com/sites/defaul...

I do like the shoulder method from Spragg et al. but it needs a cutoff that is very duration specific, making implementation (especially in a wider duration domain) difficult/arbitrary.

Question: Which approach do you think gives the best fit for the 3-param CP model? (Does it matter?)

Friday Vibe Coding

The same approach can be extended to other power duration models (such as the OmPD model) on more detailed mean-max data, showing the same effect of tightly fitting around proven maximal efforts while ignoring submaximal efforts:

Mean-max data has a built-in asymmetry that least squares regression simply cannot handle appropriately. When your model predicts you’re weaker than your data, it’s not just mathematically questionable, it violates basic physiological constraints.

Thoughts? 🤔

I have managed to do this with an asymmetric cost function. This is the OmPD model but it works similar for other models.
The challenge is how "asymmetric" the cost function should be so it allows for "gaps" (i.e. sharp declines) in the mean-max data but still stays close to the data.

Incorrect units on the y-axis in plot F. Corrected version:

Playing around with the data from this paper about running economy: link.springer.com/article/10.1...

A easy-to-use Python client library is also in the making. It gives you quick access to all your data and allows you to spin up a JupyterLab instance that is authenticated with your account in just seconds:
(this is powered by uv)

I have been working on an application to help sport scientists streamline their Python data workflow. Here's a sneak peek at Sweat Stack: From signing up to Python in under 1 minute!
Currently in closed beta. Learn more at sweatstack.no