Has TR ever considered adding coaches to the program? I currently see two options. 1.) Hire a coach. 2.) use TrainerRoad or similar. I am yet to find a coach that actually leverages AI, including brining in wearable HRV data etc. If anyone knows of one, let me know 
Wouldn’t that massively increase costs?
Good coaches generally start at the yearly cost of tr but on a monthly basis.
And why would AI necessarily make coaches better in addition to the tools they have now? (i.e. TrainingPeaks and WKO, with wearable data that already imports daily.) Have you thought that “Using AI” might not actually be better than what’s out there now?
Point being - the coach option with good tools already exists. It’s just not through TR. (And, it is orders of magnitude more expensive)
I don’t know what I don’t know. Won’t argue that. But TR does not currently ingest wearable data. I also like having a real person on other side. I also love TR. I believe there is a future where this can all live together.
You can do all of this now, except ingest wearable data into TR. And the reason for that is that the data around wearables, HRV, etc. has been shown to be unreliable as a performance / recovery metric. Here’s TrainingPeaks as an example for sharing between you and your coach that gets auto-imported from Garmin every morning:
Also - AI isn’t necessarily better, in some cases it’s worse. While I’m sure new tools will get developed in the future, a good coach with the right tools is still going to outperform any AI solution. The catch is - a real coach will cost you $150 - $1000 per month on top of TR and the tools you pay for.
This was discussed on a Podcast a while back and the short answer is it made no difference.
This is an interesting point to me, TR in podcasts can talk pretty decisively. A few years ago HR was not used by the model as it made no difference, apparently now they use it, and it plays a signficant role. I wouldnt be surprised if in the future wearable data makes its way in there. This also highlights that AI is not perfect… yet. The model keeps being updated and refined to better fit the population and individuals.
They are starting to incorporate it, but I haven’t seen anywhere saying it plays a significant role.
I think there are actually bigger more fundamental issues with it being used in a significant capacity because of the inherent variability for the same effort / power…
I’m don’t want to go back and forth on this, but they mentioned training is adjusted based on measured HR data. The fact they decided to included in contrast to earlier statements leads to my interpretation that this is significant.
We don’t have to go back and forth on it, but I would dispute anything significant based on the threads I’ve seen here.
Yes they’re including it, but Heart rate is innately very very variable, even for the same power and fitness which makes it unreliable without context that TR doesn’t always have. My take is it’s used in a limited basis if at all, and that they’re more testing it to see if it’s reliable / predictive enough in certain scenarios that they can use it more in the future.
My take is that they have identified a statistically significant signal and therefore included it in the analysis. I do believe they stated in one of the podcasts this much, I disagree with the testing statement, that was done when they recommended recording it but explicitly stating it wasn’t used for anything.
Anyway here we go… the back and forth has started
I don’t think it’s a thing. But it definitely should be. Let ai pick the workouts, coach can do the rest. Literally the best of both worlds.
Joe