I’ll let the team know that there is some confusion around the sub-5-minute workouts in the library. 
Thanks for clarifying!
So, if I successfully complete 10 ERG mode workouts, how does it figure out my FTP? Those ten could have been really tough or really easy, but the resulting power profiles would be the same either way since the power is prescribed by the workout and controlled by the trainer. What does it go by? Is it solely the responses to the difficulty survey at the end of each workout? If I checked easy for all 10, would that jack my FTP way up? Does it use HR in some way?
I don’t know what TR app observes but there is lot of information in pedal stroke alone.
Have you looked someone riding bike and subconsciously understanding whether rider is experienced/strong or not? If you analyze why you think this one way or another, then you can separate those signals (“round” strokes, consistency, mashing, bouncing, body rocking, cadence shifts depending on duration or current load).
This is just one obvious indicator. I imagine ML should be able to find more of those 
I will say that when it comes to data, quality in = quality out, and feeding the system as much quality data as you can will end with the best results.
Doing the same 15-minute workout 10 times might not result in the same precision as those who follow their TR plan and knock out their first 10 prescribed workouts.
There’s obviously a lot that goes into it and a good chunk is a part of the TR special sauce. 
5 Likes
I’m skeptical that FTP can be inferred from that. Any correlation would have to be pretty loose IMO, certainly too loose to give a reasonably accurate quantitative FTP. It would also be influenced by the response of device measuring it. Lately, I’ve been recording my workouts on two devices using the smart trainer for one, and the Assioma pedals on the other. The fluctuations in the pedal data is at least 2-4 times that of the trainer power data.
It’s not really logical that the 10 rides is a meaningful limit at all. After all, it works on any data after that point (potentially years afterwards) and if you do 10 Z2 workouts that offers 0 useful data anyway.
The limit is likely a way to hold new users in the ecosystem for a while before letting them cut loose, as well as ensuring riders have at least some data to go at if they are brand new or haven’t synced any ride history.
The problem is the 10 limit makes no sense for riders that have imported years of training history.
There is 0 chance that this is happening.
Oh sorry, I wasn’t clear – I meant it in context of profiling athlete, not detecting FTP.
1 Like
I think I did ten 10 minute workouts, or a lot if them anyway to get it started. AIFTP was very accurate.