Huge spike in Training Load when I started using AW
complete
Quinn Comendant
On the Training Load screen, there is a huge spike at the very beginning of the history of my data, when I first started using the Apple Watch. Anyone else seeing this? Is this a bug? I don't see anything similar in the Apple Health app. Does this affect the calibration of Superset? (The app always shows me as “low” recovery, maybe because it is indexed to this initial spike of data?)
Grey
complete
Grey
in progress
Grey
Quinn Comendant This should be fixed in the last few updates. Can you check to see if you're still having this issue?
M
Mike Suiter
Grey are you saying the first few weeks should be fixed now? For me it isn't.
Quinn Comendant
Hi Grey! I'm using build 786, and the initial spike is still there. 🫤
In any case, my spike is about to fall off the left side of the chart, so you have about a week to fix this before I won't be able to see it anymore. 😅
Grey
Quinn Comendant, Mike Suiter, are you still experiencing this issue?
I've not seen this issues for other users. Here is an example: https://www.reddit.com/r/bevelhealth/comments/1e2dz1z/comment/ld0suy0/?utm_source=share&utm_medium=web2x&context=3
Quinn Comendant
Grey I can't tell because my data starts more than one year ago and the max scale of the chart is 1 year. 🤷♂️
M
Mike Suiter
Grey Not seeing it now!
Randell
planned (tbd)
Grey
Merged in a post:
1Y TSB charts have bad early data
M
Mike Suiter
After yesterday's betas, my TSB charts were missing data from the past several weeks so I did a manual reset cache. If I change the charts to 1Y, the early data is wrong. I wonder if maybe you have to fetch 58+ weeks of data on a cache reset?
Grey
planned (soon)
Grey
Quinn Comendant, this isn't necessary a bug. The way that TSB model works is that it will always spike up initially and then normalizes once it has enough data. More here: https://fellrnr.com/wiki/Modeling_Human_Performance
That said, I can pass in more data before the actual time scale so that it doesn't produce this effect. This should not impact your current training load at all.
Quinn Comendant
Grey
> That said, I can pass in more data before the actual time scale so that it doesn't produce this effect.
That's a great idea. Trimming the chart by a couple weeks might help avoid this confusion. If the user has less than a couple weeks of data, just show a message like, “Two weeks of data are required before a training load is calculated.”