AI FTP Detection without using Plans

Hi everyone, I’m curious what people think about the AI FTP detection who don’t use the TR plans. Seems like the algorithms would be based around their workouts and ratings but I could be wrong on that.

I was a long time plan user, but hired a coach about 4 months ago and am loving his programming and hitting tons of PRs (both in training and races) but I still take a sneak peek at the AI FTP detection when it says it detects a change. I still do my interval sessions in the TR software and often time the classifications of the workouts from TR are a bit “interesting” to say the least. I’ll stick with what my coach says for zones but my ego really wants to believe the 9W FTP increase over the last two blocks haha so just curious about other people’s experience. When I used the FTP detection in TR plans it was always crazy accurate to set the zones in their plans.

1 Like

For me AIFTP is merely a data point to guide my TR training. I use it as benchmark purely to reference how I feel each workout feels and how I’m progressing.

I then use a variety of other data points as in input to validate the AIFTP number.

As an example I use outside rides and inside rides as a benchmark - for me, my inside AI 'FTP" feels like its about 10w lower than what my outside 'FTP" is in races. I’ve found that to be consistent over time.

That said I don’t up my inside FTP to match my outdoor FTP as again I use the ‘FTP’ number as a guide for my training.

I also never look at my watts while racing I just race.

1 Like

I don’t follow TR plans, and AI ftp is bang on


While AI FTP Detection requires 10 completed indoor TrainerRoad workouts before it can begin to be used, it doesn’t depend on TR workouts to function after that. In other words, AI FTP Detection isn’t based around TR workouts or Workout Levels – instead, it works by analyzing all of your cycling power data – TR or not!

Workout Levels assigned to custom/imported workouts (i.e., non-TR workouts) can sometimes be a bit funky, as you mentioned. This is something we’re working on improving!

AI FTP Detection, however, can function independently of Workout Levels/Progression Levels, so the data you feed it from the workouts your coach prescribes you will still contribute to accurately detecting your FTP.

Hope this helps clear things up – feel free to let me know if you have any additional questions!


I self coach but use the software because it works better than zwift or other glitchy stuff I’ve tried. I just do Free Ride workouts and then everything on resistance mode.

I do the detection every once in a while just from curiosity. It’s actually more accurate now that I don’t do TR workouts/plans and just do free ride and do the intervals I plan myself (software probably also improved in that time interval). I don’t test- but just use RPE for sweet spot/FTP work. Right now I’d say its within 2 watts of the last time I did sustained FTP intervals.

I’m also not currently following a TR Plan - coached as well - but I check AI FTP every 28 days (or close enough if I forget for a day or 2) and record that in WKO5 where I’ve built a chart that tracks:

  • mFTP - WKO5 modeled FTP
  • My Nominal (set FTP) - the FTP I have set in TR / TrainingPeaks / WKO5 / Etc.

Comparing all of these (see screenshot below) gives me a good triangulation for the FTP I’m using. Where this breaks down is if I haven’t “fed the model” (WKO5 FTP model) with hard / all out efforts, which usually only happens early in the season