In 2019 there was a feature request for pulling more non-TR data into the algorithm, and again in 2021 where Ivy said she’d pass this on to the team. I haven’t seen anything recently in the forums or in the podcast, so raising it again here.
Most modern wearables collect data on sleep and HRV. Would love to see my workouts updating based on those two things automatically. It’s my main use of ‘alternates’ in the calendar - not getting enough sleep the night before and choosing an easier workout.
Sleep’s impact on performance is massive, incorporating it into TR would be pretty slick.
I will leave this to @ZackeryWeimer to consider if this topic should be merged there or possibly left with an emphasis on the “Sleep” portion since I didn’t find an existing request on that part.
What research says you can’t train on less than normal sleep? What protocol would TR implement with this feature request? They should just trust what Whoop or Oura ring says and adjust your training to be easier?
Like red light / green light this is just common sense stuff to me. If your legs are fried and you didn’t sleep last night, then adjust your training as you see fit. People don’t need a million dollars of research and software development to make these decisions.
My thread you liked to was about TR importing the HRV during sleep data so I think this overlaps. Talking about HRV and sleep is confusing as I see 4 different data metrics you could be talking about:
1- sleep (the non hrv sleep metrics like time in different sleep stages)
2- a) hrv during sleep (what my post linked to is about)
2- b) hrv when waking up (See hrv4training @marco_alt , ithlete, etc) This represents the same type but better quality data than 2a, but requires work as you need to do these measurements consistently. This would be prefered but I’m pretty sure there are people who wear watches that automatically do 2a who wouldn’t do the effort for 2b
3- hrv during exercise (alpha 1 dfa)
See the link @mcneese.chad posted. I gave links in there to some of the research. It would be one data point of many that would influence the machine learning that influence AT and red light/green light. This is the advantage of machine learning, you don’t need to fully trust a single data metric, you can use lots of metrics together and let machine learning show how much weight it should be given.
As to trusting a single source, no. They should tag the method of getting the data and let that influence what machine learning makes of the data. Machine learning can tell which devices should be given more weight. (guessing elevate 4 hr garmin devices should be given more weight from 3 devices)
I realize you are talking about multiple metrics, but studies have shown similar to what @AJS914 was talking about. RPE can be a better indicator than some of these other metrics. Likely because it is complicated, and taking even a few metrics doesn’t tell the whole story.
The other study linked is interesting but I don’t see how (or why) TR would base a major software development effort on it.
One of the metrics in the other study was leg soreness. That metric is free without software development. Are your legs fresh? Do a little more. Are your legs trashed? Take it easy.
Also in the other study, the individualized group’s 10k time suffered because they didn’t do the programmed rest week like the control group.
The problem with the other study is that the control group trained like robots on a schedule no matter how they felt. Most people in the real world do train by feel to a large degree. The control group should have similar guidelines to adjust training up or down based on how they feel. Then you’d see whether having senors and software do this for people is really adding something or not.
Absolutely - I track HRV, resting HR and sleep quality and, while there might be some trends medium / long-term (particular with resting heart rate raising during a training block), I see little / no relationship between these metrics and the ability to complete a tough workout on any particular day
RPE, nutrition and feel are much more reliable indicators, for me anyway