I have no idea if AI FTP detection takes this into account or not, but adjusting the intensity of the workout down, ending it early or taking breaks during hard intervals might come into play. I’m a relatively new TR user, but I’ve found that these things definitely affect your progression levels.
Just stumbled across this previous thread. I just had a similar experience, with the AI FTP detection suggesting a very modest (4 w) FTP increase. So, I took the ramp test and got an FTP that was >20 watts higher (8% higher than the AI estimate).
I think what tripped up the machine learning algorithm is that I’ve been working with a dietician to better fuel my workouts and recovery. That meant that I was adapting much more strongly to the workout stress compared to my training history.
So I think one lesson here (apart from completing the post-workout survey!) is that AI FTP is likely accurate as long as there aren’t any substantive changes in what you are doing. If you do make substantive changes to your training, it’s probably a good idea to do the occasional FTP assessment to help the algorithm out.
Now I have to deal with the fact that it reset my progression levels to ridiculously low levels. I’ll experiment with more challenging workouts to re-calibrate things (‘approaching them with curiosity ™’).
1 Like
You learnt the wrong lesson. Ramp test is a bad estimator of FTP. I’d bet money the AI is way better….If you want to do an assessment, go ahead and rip 30-40min hard. Or at a minimum do the Coggan protocol, for the latter don’t forget to go ALL OUT for 5 min!
1 Like
agree to disagree. I’ve done both ramp tests and 20 minute test and they always align for me.
Maybe (anaerobic contribution) they both overestimate your FTP? 
Now that it has been a couple of weeks with your new ftp, how are you doing with the workouts? Where are you in a training plan also? You must have felt like superman before deciding to do an actual ramp test.
Let us know if AI FTP was way off the mark for you.
Good question. It took a bit of trial and error to find the right progression levels, because at first they were waaaaay too low (like level 1!). I hit the mark perfectly with the first threshold interval, picked one slightly too challenging (but doable) for a 90 minute sweet spot, and then aimed a bit low for the following VO2 max workout. But nothing that the trainerroad algorithm can’t deal with
So generally, my fitness has handled the change well.
What hasn’t handled the change well: my knees. One knee (the “problem” knee) had a flare up of patella-femoral pain. In retrospect, it would have been better to increment upwards the upwards change (in total 26 watts) over the span of a month. I’m now holding back and doing some endurance and tempo workouts until the soreness subsides.
1 Like