Confidence in AI FTP Detection up?

Hi everyone,

For me I’ve finally settled on that I have confidence in that the TR AI FTP detection (AIFTPD) is as close to the mark as you can get to my real life AI FTPD based on my one person real world testing.

This is pure science.

The TLDR version:

3 X AI FTP detection in last 4 months.

completed Ramp Test and 4DP test as validation.

All 3 validation tests were within 3w of AI FTPD.

end result = I have confidence that for me - AIFTPD seems to be on the mark.


I like testing. I like the feeling of being in the hurt locker. I’m a touch insane, yea.

I also don’t like to blindly accept things as fact, especially when there are so many variables unavailable to TR and its input data. Things like sleep, nutrition, recovery etc

To that end, I decided to validate the AI with real world tests. So I did, and now I’m at a point where I’m comfortable accepting AI FTP moving forward. This attitude is based on me doing tests very close to the point of AIFTPD and then confirming with other tests.

I’ll still test if my gut says I should but my anxiety has gone. More specifically my stupid brain is now quiet about this as a concept.

So, has anyone had similar results? Or does anyone still question AIFTPD ?

Of course this is just one person’s experience but I’d be interested to see how everyone else’s experience has gone thus far.

Note: I edited the original post as when I wrote it I wasn’t wearing my seeing spectacles and was typing on a small screen so there were some grammatical errors that I fixed.


Yes I’ve tested and detected on the same day and got the same results


I signed up almost a year ago. The first AIFTP offering I was presented was in July & I was able to beat it by 5% with a 20-minute FTP test, so it was a significant difference. After my last two AIFTP offerings I tested again (20-minute test again), playing a game to see what I could actually do. On both those occasions that I put myself through the wringer, it was for one measly watt.

Yeah, we have to be a bit masochistic to play that game. :person_raising_hand: :laughing: But I think now that AIFTP is giving me results this close to my current potential, it’s maybe not worth losing the few days before a test as a taper/refresh, & I could instead be spending it in the hurt locker doing more hard workouts. :smiling_imp: Kinda joking, kinda not.


I feel you, my lizard brain thinks that way too. I think for me now, I’ll put my energy into worrying about things that matter. You know, like sock height and cap under helmet, glasses over or under helmet straps. Important things like that :grinning:


Priorities :+1:

1 Like

FWIW, in the last few months my Garmin’s auto ftp detection that comes up after a ride has been within 5 watts of AI FTP detection. I just had one come up this last Saturday from my Garmin after a hard group ride at 272 where a week or so ago AI detection had me at 276. I haven’t done a test but having both be close inspires confidence.


Yeah, I didn’t mention it in the post but I’ve had a similar experience recently with my Garmin as well. Almost like science works…pfft never

1 Like

I think it undershoots by maybe 3-5% for me, haven’t tested as such but have done long maximal efforts e.g. TTs and long climbs where I am consistently able to do higher numbers than AIFTP would imply (and get FTP updates from apps like Trainingpeaks and Intervals giving me higher FTP than TR does).

I’m ok with it as I do find it a little harder to put out power indoors (different PM so possible that’s the difference though I have dual recorded in the past and found them to be pretty close). So when I do an indoor TR workout in erg mode AIFTP is about right, and when I do outdoor TR workouts I typically just go by RPE and tend to overshoot the target power a little, or at least be at the top of the range.

Did you do the tests before or after you saw the AI FTP result?


I’ve done it both ways. I’ve seen the AI number first and got a higher number, or like today got the number bang on (Note I set a new max HR number so wasn’t like I was foxing) and I’ve seen the number and gone lower. I actually like to see the number and ride as if AI has disrespected me in some way and I’m out to prove it wrong. I make it fun and I’m not really that anxious about it. FTP is one small part of my tool box to make me as good as I can be at this particular point in time.

A lot of what I’ve written today is in gest.

I also turn the screen on my mobile around during the test and just ride till I can’t anymore and let the chips fall. I understand that your numbers on any particular day are just that - numbers on that day. If I feel like the test was compromised in any way I readjust.

1 Like

I have found similar outdoors. I only do my important sessions indoors and use the indoor number as the number for training. When I race I don’t pay attention to my data I just race on feel. All my outdoor riding is endurance or recovery where my focus isn’t really on the numbers but just being out and enjoying riding my bike.

Since 2017 I’ve had high confidence in WindWarrior™ FTP estimation… go out and do tempo around 145bpm, divide power by .85, and thats my rough ftp. Verified that against FTP estimation from 8-min protocol, 20-min protocol, and 30+ minute protocol. Given what I know about machine learning (ML) and training, I imagine TR AI FTP works something like that, except based on power and because ML it can use more efforts and maps you onto a curve from a bunch of other peeps. TR says it wants indoor workouts, partly because it knows the workout % FTP and approximate FTP for that power (from all those other peeps). TR also has stated that if your FTP was not correct, then AI FTP will not be accurate. Long story short:

The better your FTP estimate(s), the better TR AI FTP works.

yeah, hence me doing the testing to validate that. Now I feel like my FTP estimates are strong enough in the system to give me confidence moving forward that I’m getting the best training based on where I’m at at this point in time

1 Like

After a build phase indoor, AIFTP was 2 watts lower 238 v tested of 240. Lately it’s been within a few watts. It’s gotten better as AI tends to. Still hate to trust the machine though! Love the pain! Always that doubt that it’s wrong. Heck I like to prove it wrong…

As do I. But for now I think I’ll accept that our AI Overlords are smarter than I give them credit for. For now that is…

FTP for me is a training number. As far as I’m concerned it could be called something else to avoid the “what is FTP” debate, but regardless what you consider that number to represent, AIFTP does well in corresponding that to various training zones for me. If nothing else it does as good, or better, than my old coach prescribing “AT” or “tempo pace” back in my college running days.


which is why we have a debate, because people call FTP something it isn’t. Really isn’t hard, the concept of an upper threshold was established 100 years ago.

Last few month I feel the detected FTP is more accurate then in the past (benefit of more data I guess)
Last week I did a climbing tt and did 280W for 45 minutes.
This week I had an AI FTP detection from 280 to 282W, which I would define accurate.

My issues are:
Indoor I cannot push the same watt. My heat management is very poor, I live in very hot country and for many reason I cannot improve it right now. I will as soon as conditions allow me to do so.
I strongly believe that AI FTP put too much stress on the number and to let users see it (hopefully) grow over time, no focus on TTE. As I stated in another thread, I’m waiting for 284W to get a nice 4w/kg, then I’ll stop accepting FTP increase and focus only on TTE (for at least 28 days :no_mouth:)


to me it’s a just a number to guide my training and nothing more. As I said before, I pay no attention to it when racing because it matters even less. In terms of me putting out effort etc. I’m sure for a TT it’s important but for my local Cat 3/4 crits. Not so much


So close, go for it! :muscle:

Yeah I felt a bit of a party vibe a few months ago when I got to 4w/kg. Decided in that moment if I ever get to 5w/kg I’m getting a tattoo of a bike part.