📣 New AI FTP Detection Updates! 📣

I noticed that my AI ftp was 28 days at the start of my new block and now its starting to come down faster as i keep nailing 6 + work outs.

I like that, not wasting time on useless easy training

1 Like

Just wanted to shout out an appreciation post. I’m super impressed with AI FTP. I’ve had an incredibly inconsistent year for various reasons and I lost my power meter on the road to boot so a good half of my rides only have heartrate daya.

I just did an over under workout yesterday on the trainer where I could feel the 15-20 watt over and under RE that lactate heaviness fatigue feeling. It’s impressive how narrowly dialed I am.


TR Team: a note of appreciation for AI FTP feature. For a stop-starter like myself, 1 in 4 workouts was a ramp test in 2021.

With AI FTP, they are all productive workouts!


An aside: intervals dot icu’s latest workout graph update is :ok_hand:


Quick question about the remind feature/delaying ftp updates in general

I plan to turn down a few ftp bumps in favour of advancing PLs, but I’d still like to see them.

Does AI FTP work in a way that I’ll still see month on month improvements, even if I continually reject/“snooze” them?

Your FTP is not determined by if a month has past or not. It’s based on the training you’ve done and if that training stimulus had been enough to improve your FTP. So whether you accept/ignore/pass on AIFTP it doesn’t matter. The next time you’re able to use AIFTP it will analyze your work and determine your FTP.

1 Like

I get this, I may not have communicated my question well enough.

It’s more around how smart AI ftp is and whether it’s weighted to the current TR FTP.

I’m just curious whether declining ftp updates will materially change the output of ai ftp. Not for any training reason, just curiosity.

There is no reason why it would. Your FTP is what it is whether you accept it or not. The only difference is whether TR will adjust your future training (PLs). If you skip or ignore AIFTP then training continues as normal.

1 Like

Of course FTP is time dependent. Stop training for a few weeks and test.

:roll_eyes: excellent addition to the conversation. No one said anything about stopping training or it wasn’t time dependent.


I admire your restraint :smile:


Thanks, this is the assumption I made.

Are you asserting this on experience or assumption?


1 Like

AI FTP Detection looks at all the work you’ve done regardless of if you accept or decline any previous detections.

Ideally, while following a plan and bumping up your PLs, your FTP will still be going up!

1 Like

I don’t think you’re going to get the same results if you continually turn down AI results/manually adjust FTP vs going with the AI suggestions.

My experience: TR AI thought my FTP is 250

Every other test and model thinks my FTP is 220. Above that and I’m on borrowed time. I dial my FTP manually back to 220 to get better SS/TH zones.

The next AI ftp test I get is 229. 30 days apart my AI ftp goes from 250 to 229? I definitely didn’t get slower in that time, I just recently set a bunch of race PRs.

AI FTP doesn’t seem to work in absolute values, it looks like it’s working in gated plus/minus values. It thought giving me a +9 watts was a big jump forgetting that I was recently +30 watts

So, long story short if you ignore it for a while I think you’ll get different results than if you keep accepting. It’s designed for incremental changes.

AI FTP doesn’t seem to work in absolute values, it looks like it’s working in gated plus/minus values

The software engineer in me wondered about this and that was the motivation for my question.

It seems completely valid to use current FTP in the estimation of the AI FTP. I suppose only someone from TR can answer that question categorically and even then, that may be too sensitive to answer.

I had a similar experience where I was stuck in the 230s for ages, until I did 261 outside for 46 mins and manually bumped my FTP to 250.

All of a sudden I get a big jump to 243, 248 and now 258! Very anecdotal, or N=1 as TR like to say but I think it validates my question. Not that it matters, I’ll continue the course either way.

1 Like

I’m starting to wonder if AI FTP is biased to give increases for ego reasons. I had just finished a low-PL Threshold (2.5) workout with an “all-out” and AI FTP wanted to boost my FTP. I rejected that. Could be because of my low fractional utilization (i.e. terrible muscular endurance). Now it’s giving me stretch VO2 Max and keeping the threshold doable, so I guess that’s OK.

This is a good point here – while AI FTP Detection looks at all of your power data from your training, it also uses your historical FTP data.

The more accurate FTP entries an athlete has, the more accurate their AI FTP Detections will be. Feeding as much accurate data to the software as possible is almost always a good thing. :slight_smile:


Okay, then … but how to keep enjoying the higher PLs (equalling more time in zone for the threshold/VO2max workouts) while accepting the newly detected FTP. After accepting I am getting back down in my PLs, having to go through the short intervals again.

Just pick a different workout and teach the AI your PL

1 Like