Which is the point people have been trying to make…there is no scientific data for workout recommendations based on HRV, sleep, etc.
Except I never said that. I said I often find the recommendations from my Oura ring to be 180* from how I am actually feeling / performing.
Let’s try something that might get through the shell- most cycling users do not have this data and/or it is coming from enough different sources to justify the development costs associated. That might change. TrainerRoad has provided a resource for all their users, not the snowflakes and turtles.
I don’t understand why you would rely on these types of “measurements” to adjust your training. You know how you feel. If my Apple Watch is telling me I had a great sleep and great recovery, but I feel like 
, I’ll adjust my training because I feel like . If my Apple Watch tells me that my sleep was horrible and I’m totally not recovered, but I feel good, I’ll leave my hard workout alone, because I feel good. I think people should listen to the podcast pointed out by @Flashpoint51. They go into detail explaining the limits of these types of “readiness scores”. They might be ok for looking at long term trends or confirming how you actually know you feel, but they’re nowhere near ready to be useful for actually adjusting a training plan based on the score they generate. Remember, you know how you feel, you don’t need a device to tell you.
This only works if the data is good, the accuracy of wrist based monitors (have not seen any concrete info on finger based ones) is very poor. The best you get from the is sleep time.
You are better off learning to listen to your body rather than some device that is much less accurate.
Everything that I’ve seen is that this data may work well for some people, but conversely it works horribly for others.
I have what I’d call a very predictable and reliable Heart Rate. I can use it as a pretty objective metric for how I’m doing aerobically on the day, in a workout, etc. HRV generally tracks well for me for how I’m recovering and feeling.
But HRV and sleep metrics are also a pretty big miss for me in terms of predicting training readiness. Interesting to follow, if I sleep horrible or HRV is all of a sudden way low I know it’s a bad day, but that’s about it. Especially going the other way - HRV and sleep being good does not mean I’m necessarily “Green”.
My opinion, take it as a data point and follow it, but that’s it.
And like everyone else said - bad or inconsistent data doesn’t help. Data quality is the bane of AI / ML and literally one of the first and most important things that needs to be addressed in development.
I get this all the time. My normal training is back to back hard days on Sat. and Sun. Not too surprising I usually have my best sleep on Sunday night. Longest hours, deepest sleep, good quality. For whatever reason, I also usually have a fairly high HRV on Monday morning. After a weekend with 400+ TSS, I am definitely not ready to train, but these metrics would likely put me in the green.
My Oura ring had my readiness for today at 86 (Optimal) and the highest it has been all week.
Hopped on the bike and legs immediately felt like . Was planning on the full 4 hour BMTR ride and pulled the plug at about an hour because the legs just weren’t responding.
I see some replies pointed out the lack of scientific base for some of these metrics.
Can somebody point me to the science behind the RLGL approach used into Adaptive Training?
Being purely based on workload this is no different than the basic Performance Management Chart that we’ve been using for the past few decades, based on Dr. Coggan models.
To date that model has several limitations, most of which were already highlighted by Coggan, in regard to the determination of the time constants that are driving the decay and response of fatigue and form.
The models, as we have them today, are based on parameter fitting based on “black box” statistical analysis.
Today we have the opportunity to start exploring another level of optimization that would allow to identify specific time constants for recovery given a specific amount and type of workload (because let’s not forget that TSS/CTL approach does not really differentiate between work in different zones, while our body certainly does)
An optimization of these models is only possible if we look at the output of the training impulse and we measure directly the level of recovery for a given subject.
While “Adaptive Training” sounds pretty cool, I would argue that the algorithm in place is hardly adaptive in nature (whereas the algorithm would change its form to adapt to the specific user) and at best a “tunable” algorithm.
The only feedback signal collected as result of training impulse is the RPE survey after a workout (and the initial survey at the beginning that asks how motivated you are to train today)
I often get RED days when none is needed.
The PMC suggests I am not in a ‘red’ training state today.
Yet TR prescribed a Rest day today and tomorrow (red and yellow).
My morning ride was excellent, thanks. Very hard to take these recommendations seriously, and questioning the adaptations on the training plan as well at this point.
Honestly this is not a question of integration with other platforms or data complexity, these are trivial tasks, it’s a matter of advancing training in the right direction, and if TR is not looking at it someone else will.
I fail to see how the problems with RLGL will be solved with HRV, which you have acknowledged many of us have found to be mostly meaningless in the real world, and science has not proven to be a worthwhile metric. I’m not arguing with your RLGL concerns, just stating that it seems obvious to me that HRV isn’t the magical solution so many marketers want you to believe it is.
Plenty of really smart people are……doctors, PhD’s, medical researchers, etc……and they have yet to find any scientific evidence supporting using HRV as a reliable training metric.
Exactly. I think @ZenTurtle is right that TR should do things to advance training in the RIGHT direction. I also respect them for not just giving in to the fad of the moment if it’s not the right direction.
You seem to have some real beef with HRV, I get it, you made it very clear, I don’t blame you (yet there is ton of supporting research if you care to look for it)
Let’s talk about sleep for a second:
if I was up all night because my child was sick vs if I slept 10 hours like a baby.
My capability to execute an intensity workout will be impacted way more than whatever workload I did last week. Period. This has nothing to do with my HRV, it’s simply sleep tracking.
Right now the algo they are using is completely blind.
If I trained 16 hours a week, the determination of my capability to train (aka readiness score) would be based on 10% of my total time, ignoring what happened in the remaining 90%.
On the other hand, my experience with products like Whoop was underwhelming because they focused entirely on my recovery and poorly tracked my workload.
In order to have a good model we need to have a holistic approach, a comprehensive and integrated perspective that considers all aspects of a system or individual, rather than focusing on isolated parts.
How we get there is up for discussion, that’s where all the smart people should focus on.
I have no beef with HRV….there are a lot of good uses for it. A training metric however, isn’t one.
What I object to is advocating for unproven technologies to be used in ways that are not yet backed up by data.
The world is full of people who have done great athletic achievements with little or no sleep. There are studies that show sleep deprivation doesn’t affect physical performance for 72 hours.
More importantly, if you have been up with your kid all night and you feel tired, listen to your body. You don’t need an app or a device to tell you that.
It seems to be a common and understandable misconception that RLGL is an indicator about how you feel on that days the warning is given.
My undestanding is that the feature is there to prevent long-term fatigue and is not the same as PMC nor is it todays training state.
This is not to say it is perfect or working for everyone. I’ve found it has taken a little time and adjustment in the user settings but is now working great.
https://support.trainerroad.com/hc/en-us/articles/20753792083995-Red-Light-Green-Light
