Stingy AI FTP after strong base block? Or maybe looking at my history to project an end point for build?

I just wrapped up my final base block and I thought AIFTP would indicate a pretty big jump in FTP based on some of the numbers I’ve been doing. It only gave me a 1.7% increase and that has me wondering what the heck it looked at since my ftp was sitting pretty close to my off season low and I’m doing better numbers (in my mind) than I was doing this time last year when my ftp was about 7 watts higher (that’s about where I thought it would land). One theory is that it may like to give out bumps in small increments and maybe I should have let it bump me up a month ago. The other possible factor is that I’m doing a lot more outside training this year and I’m not typically assigning a TR workout to those rides (but I didn’t think AI ftp relied on TR structure and the rides do have intervals). My last theory is that maybe it sees that I’m already pretty close to post-build highs in previous years and it’s projecting a limit somehow based on my history.

I was “due” for a FTP update about a month ago and would have expected a small increase at that point, but I had a hand injury and I decided to wait until I saw how the injury would affect my training (it was a little disruption, but just allowed me to “re-do” base 3 stronger). Anyway, I felt like I was crushing this last 3 week block and ramped volume from 800tss to 1000+ (more volume than last year). On the last day of the block (today) with all that fatigue in the legs, I did a 6+ hour 371 TSS ride that started with ~2.5 hours at .7 IF for warmup, then I did a 60 minute interval at .98 IF (basically my current FTP, but felt more like sweet spot), then followed by ~3 hours at .65 IF. I rated it as a “hard” ride based on duration, but it was very manageable. With my new FTP today, that 60’ interval is still .96 IF. Similar ride last Saturday, just an hour shorter. I know the FTP accuracy doesn’t matter that much in the context of adaptive training, but I just thought AI FTP would have picked up those long intervals and given more credit for the effort. I’ve done zero vo2 max this year, so maybe it wants to see some of that? I usually question (and sometimes reject) AI ftp because I think it’s optimistic, so it was a surprise today.

Anyone have an idea what type of intervals/efforts AI FTP prioritizes? I know it’s designed to not require specific protocols, but it must be keying off some time/power metrics (especially if you don’t use TR structured workouts to give AI ftp context for a given effort).

I think I read somewhere else that if you don’t assign your outdoor rides to a TR workout than it can impact how the AI calibrates your fitness progression.

I hope I got that right as I don’t want to give you a bum steer

I myself have been thinking about the same thing and my plan, note my stupid idiot stubborn lizard brain’s plan, is to do a ramp after each major block to cross reff with the AI overlord.

A TR employee told me there’s no need to if your data is matched and sweet, but you know, stupid lizard brain can’t deal with reason

1 Like

No clue how to game the AI to get it to notice what you want it to notice, but if you just did all that, give yourself your rest week then do a long form test and see what your FTP is. It sounds easier than what you did today.


I wonder if the AI FTP detection has changed. My last detection suggested dropping 2 watts, progressed in 4 different PLs (End, SS, and Threshold, VO2) and didn’t fail a workout (had failed one in the previous block). Not upset but kind of surprised.
Will keep progressing through the PLs and see what it says next time but hopefully not another suggested drop (that would make 3 in a row, I didn’t accept the 1-2 watt drops advised with the last 2 detections).

1 Like

I’m experiencing similar (lack of) gains in FTP as I hit my Build….it seems that AIFTP responds “better” to threshold / VO2 Max work vs. what is primarily endurance work of the Base period.

Despite lots of base miles and weeks routinely between 500-750 TSS, I’m basically picking up 1w / month. But I haven’t had a whole lot of higher intensity work yet, so hoping it starts to increase a bit faster now. But my fitness feels good and my endurance is definitely improving.

Remind me not to try and sit on your wheel at Unbound, ‘k? Thanks. :crazy_face:


I’m in the exact same position. Three strong blocks of base, completing SST workouts with a PL of 10+ and seeing a 2w increase…

I went back and looked at last year’s base, and I did nothing but sweet spot and endurance and none of last year’s sweet spot workouts were as hard as the outside rides/intervals I’ve been doing (even ignoring the extra volume around those outside intervals). The big difference was that I was almost exclusively doing TR workouts inside last year.

I think this is really about AI FTP doing a much better job of recognizing interval/workout efforts when it’s tied to a TR workout vs. just going out and riding (with intervals worked in). AI FTP works for both, but appears to work much better when there is a TR workout (which makes total sense). Maybe when Adaptive training supports unstructured stuff, better AI FTP recognition for unstructured will come along with it.

No hard intervals at Unbound for me after the first ~30 minutes to sort things out. After that, I hope to be smooth and steady all day.

1 Like

The gains feel less this year. My FTP is lower this year at this point, which really was part of what I was trying to manage better as I felt it went up too quick last year. On the other hand I had never done a VO2 workout in the 8s for PLs. The last VO2 workout prior to AIFTP reassessment was an 8.9PL. I was surprised that the bump was 3 watts.

Dont really question where I am as I will be roughly where I expect to be in May.

It just feels different this year.

AI FTP doesn’t need a workout to be assigned to outdoor rides. It analyzes the work done, inside and outside, whether there’s a workout assigned or not.

You’re probably thinking of Workout/Progression Levels and Adaptive Training, which do need an actual TR workout assignment (for now).



Ah yes, that was it. Thanks for the clarification

1 Like

The temptation to remind you all that you’re all a year older than you were this time last year and should be comfortable with slowing down a bit is almost irresistible



My bad.


1 Like

Just do an FTP test.

1 Like

Just to update on this, since my AI FTP on Jan 2nd 2024, the next 2 detections suggested a 1 watt drop and then a 2 watt drop, this week finally got the same AIFTP detection despite having much higher PLs than I did in December before that Jan 2nd detection.

This most recent block included Endurance progression to 4.5 (Seathwaite), Sweet Spot to 4.3 (SS90/40 4x10), VO2 Max to 4.0 (Mezenec).

To compare to what I progressed to get this same FTP in January: Endurance 3.0 (Malecoste), Sweet Spot 3.5 (Round Bald -1, this was technically 5 weeks before the detection so maybe my SS was down to 3.2 for SS90/32 4x8 by detection), Threshold 3.3 (Yukon and Beacon -2).

So by comparison it seems like AIFTP wants higher level PLs now.

For more context most recent block avg’d 390 TSS across the 4 weeks compared to previous avg of 259 TSS across the previous detection’s (for the same FTP) 4 weeks.

So it seems like AIFTP is demanding more for an increase compared to a few months ago.