This seems crude, but I wonder if maybe the AI determines what the appropriate watts and duration are and then goes through the list of appropriate workouts alphabetically?
1 Like
That would explain Cloudripper as well.
1 Like
I had something similar and it might be even better than that. Do you wear an HR strap? It might be picking up something there that’s telling the system you’re not as fit as it thought.
1 Like
Yep. I do wear a strap. Between my pegged HR and me rating it on the right hand side of the curve, I think it adjusted exactly the way it should have.
1 Like
A thought to add to @aydraper’s opening post…
Going by one of the recent podcasts, it seems TR have focussed on the ‘F’ in FTP: “functional”, as in, functional for training. Naturally this means, within the TR ecosystem. The way I see it, TR’s FTP prediction is simply saying, this is what we’ll train you at next month. That’s it.
The number offered is not an assessment of general bike fitness.
Now… if we could have a power curve prediction, that would be mint!
I personally am not using the AI prediction, because the 5% bump after the old system resulted in threshold workouts feeling unnecessarily cruel. (I’m currently training 6 watts lower.) Once I’ve swallowed some cement & HTFU I might let it have its head…
8 Likes
I had the same feeling. Now that I have been doing more threshold etc at my old FTP maybe, just maybe I can bump my number up but I dont know that I am willing to do the level of work that TR thinks I can do. They are just a level of effort beyond which makes sense for myself. 10watts is just a bit too much for me at this point.
Ok. So, when we see our completed w/o and look at workout levels I guess it’s very hard to understand why a level could be lower then expected with no pauses then? Here I assume the lower watts cased this?
I’m looking into this now. The only thing I can see is that you ended your workout a bit early, but that shouldn’t affect your threshold points from what I understand.. 
It sounds like Scott is working through this via your support ticket, so he should be able to help you out there..
1 Like
I don’t think that it is true that changes in the AI FTP prediction is a straight line assuming each workout follows the most likely RPE response.
It is true that RPE response is the only opportunity for the AI to get subjective data from the user. But it has all kinds of other time series information gathered during a workout – cadence, HR, power – that help it understand how your body actually responded to the workout. You can’t change those data once the workout is done, and the TR folks have certainly dropped hints that they factor into the AI’s calculation.