Pacing and New AI ftp detection metric vs physiology

TrainerRoad Software
1

Hey all! Like many users, I have been trying to wrap my head around the new AI FTP and workout scaling structure.

I will start by saying that I feel like the focus on progressive overload, rather than just scaling workouts to zone is hugely positive and I believe the workouts I have been getting are much closer aligned with my abilities in each category.

When the new AI dropped, I got a pretty devastating 20ish watt drop in FTP. I have since emotionally recovered and I do value the new AIs ability to pick progressively scaled workouts as previously mentions. I do have some lingering questions about the new AI ftp detection that I would like to understand a little better.

On one of the initial podcasts, Nate said something to the effect of the new AI prediction is based on setting your athlete abilities roughly at a level 3 in the threshold category. From my experience using trainer road, the only thing that increases threshold level are threshold workouts. This would make it seem like the best way to increase your AI FTP prediction would be doing a bunch of threshold intervals. Conventional sports science seems to contradict this and we hear Hannah Otto often talk about raising FTP by pushing from the bottom ( endurance ) and pulling from the top (vo2). So my question is really two fold. Does the AI FTP detection register an increase from progressing the threshold athlete level? Does the progression of the threshold athlete level occur from DOING threshold work or from how you PERFORM and report on threshold level workouts? Like, if you theoretically did a ton of endurance and vo2 work, and then completed a threshold workout with a low heart rate and and easy rating, would the system recognize that your FTP was likely set too low?

My other question is related to pacing events. Trainer road has previously released articles on FTP based pacing. Some of the initial podcasts about the new AI FTP seemed a little ambiguous about whether the new AI FTP detection was really the same thing as what a traditional 20 minute or ramp test might reveal. Does trainer road still endorse using the new AI FTP prediction number for event pacing purposes?

Thanks in advance for any feedback. Sorry if this is a redundancy in the forum, but from what I’ve read through I haven’t found these points directly addressed.

2

I haven’t kept up with the forum after the new AI system dropped, it’s been off-putting. If I were following a TR plan I would compartmentalize AI FTP and FTP if pacing a race via FTP or IF. For me those two numbers are close enough to not be much of a concern, but others may differ from what has been discussed in the forum. In a race setting RPE will be king with power serving as guardrail (or anti-lollygag motivation)

3

I can anecdotally answer part of your query. Right after TR AI dropped, I completed an 8 week Z2/3 Base Building block of my own design. Without any threshold or higher workouts my AIFTP went from 289 to 317. So it clearly detects progress beyond just completing threshold workouts.

4

Thanks. That is good information.