What exacty does AI FTP tell me?

I recently did a 20 minute FTP and the Moore test both gave me roughly a 235w FTP. The day after I ran TR AI FTP and it gave me a 257w FTP. Considering nothing in my training suggest I can hold for 257W for a measurable amount of time, what exactly is my AI FTP? As far as I can tell I can’t use that number for anything useful

I also asked TR chat how they got the 257w number but they said that will take a week or so to get back to me.


I doubt it will cover your question exactly or deeply, but have you reviewed the support article for starters?

That doesnt tell me what this number is. It only tells me how to get this number but now practical way to apply it.

There are a lot of definitions of FTP out there (your “hour power”, your lactate threshold, yada, yada, yada) - but here’s what it comes down to: it’s a baseline from which you can define training zones to structure your workouts around. It’s not perfect, but when combined with TR’s progress levels and adaptive training, it works pretty well.

You’ll probably get a boatload of various explanations here about what FTP is and even more arguments about why those are flawed, and maybe even a nice little flame war about which method is the most accurate measuring method.

Personally, I recommend you ignore all that and just let the AI and adaptive training do their things. Don’t get hung up on the number, nor on weather this method or that one is the best measurement. Instead, worry about whether or not the workouts are feeling like they’re supposed to feel, and whether you’re getting faster, riding farther on the bike.


I’d be interested in what you’d consider “useful.” Holding your FTP for “a measurable amount of time” isn’t useful in and of itself. It’s a means for maybe determining your FTP, but your FTP is “useful” for setting training zones. You can argue whether that’s really your FTP or not, but regardless, it’s use is for you to plug into TR and start a plan and start progressing. If you find those workouts too hard, then you can start with either lower progression level workouts, or you can split the difference between your 20 minute test and your AI FTP.


But how are you supposed to know what its supposed to feel like? Doing something like 2x20 90% of my FTP should be a fairly easily workout for sustainability but that would basically me doing 2 max effort 20 minute test back to back. I might be able to do it but it isnt SS where im able to recover easily. Having an inflated number setting zones might be do able for a short period of time but if im essentially turning SS into Threshold am I accomplishing my goal?


I would go with the TR AI FTP for use on TR

You will soon know if it is in the ball park or not after your first few workouts at the new level

Answer the post ride surveys and stick with the plan so that the adaptive training can get to work too and it is likely to work out OK in my experience

Out of interest, is this the first time you have run the AI FTP and did the previous two FTP tests feature in the data it was analysing when coming up with its number?


Give a an honest response to the post workout survey (but don’t overthink it). If you’re marking hard workouts easy, or not able to finish workouts because of intensity or fatigue, adaptive training will do its magic and recommend adaptations. The whole point of AI and AT is that you don’t have to worry about it. You just do your workouts and it will adapt your plan as needed.

1 Like

Right now it gives me the workout Bolivar has a productive workout which is 52 minutes at 242W. I was just able to do max tte at 235 for 45 minutes. Now i need 7W more for 7 minutes minutes more?

You can manually enter an FTP value and ignore AI FTP.

My opinion - run when anyone tells you to use AI FTP if it gives an estimate that wrong. There is over 100 years of science of the upper threshold.

FTP is the border between stable and unstable physiology.

Literally that simple.

In my case, go 30W over and I blow up within 10 minutes or less. Go 30W under, and I can go for 2+ hours.

It’s really not that hard to understand. What feels like gaslighting is asking athletes to suspend belief in well established science and training (TR uses Coggan levels after all), and avoid manually overriding a clearly wrong value.

AI FTP requires more good data to do a better job, so give it a better seeding value.

Manually set FTP and let AI FTP catch up… not the other way around.


Bolivar is a Sweet Spot 7.6, which is a substantial workout. If you are remotely close to completing that, then it would explain why the AI FTP is suggesting what it’s suggesting. If it’s too hard, just tell the AI that it was too hard, and it will lower your progression levels to something you can do. I don’t think you need to lower your FTP. If anything you may need to raise it so you can do some workouts at lower PLs and give your progression levels somewhere to grow. If you try a Sweet Spot workout around 5.0, how does that go? It’s probably too easy for you.

I did an 8.6 workout before my AI FTP and it now has me at 7.6

So it lowered your progression levels when it raised your FTP, which is what it’s designed to do. However you are still at a pretty high level in PLs that it perhaps didn’t lower it enough. Maybe you had a really good day when you completed the 8.6 and a bad day when you failed the 7.6, but if you are completing workouts between 7 and 8 then the FTP is not that far off. Ideological views of FTP aren’t that helpful in a world of Progression Levels since the two intersect in TrainerRoad-land. It sounds like you just need to be honest with the surveys and let it drop you down to a PL that you can complete. I wouldn’t lower the FTP just so you can not fail a workout that is already near the top of the PL spectrum.

1 Like

To paraphrase, the logic you and others are suggesting is:

If FTP is set correctly but Adaptive Training needs PL headroom, then we need to suspend belief in having a correct FTP and surrender to TR’s algorithms.

Am I following along?


What exactly is AI FTP telling you?

Maybe you are bad a testing?
A value you can train at to increase fitness?
A random made up number to make you feel good about yourself and stay subscribed? :wink:

Sounds like you completed a real FTP test, particularly two longer efforts (i like the Moore version best) and you have a realistic number for FTP.

In your case, AI FTP gave you a made up number for fun, but you already have a real number ton work off of.

That’s what is called progression.

As far as AIftp giving you a bump, it might feel that you had something left in the tank at the end of the 20-minute test.

FTP is a replicable baseline to evaluate your fitness and a useful metric to structure your training. If you’re using TR software to structure your training, it makes sense to use TR software to set your baseline.


Interpret it how you want. If he is using TrainerRoad, which he is, he asked what he’s supposed to “use” the FTP it gave him for. And I said, use it with TrainerRoad’s workouts. Doesn’t seem like too much of a stretch. If you want to argue whether TrainerRoad is the best way to go about it, that’s another story. I personally don’t mind folks with your approach or a non-TrainerRoad approach commenting because it gives me other insights for myself to use, but trying to apply some non-TrainerRoad insights to TrainerRoad plans can be really confusing to someone who’s new at this and using TrainerRoad. You might as well say, don’t use TrainerRoad. Which is fair to say, but doesn’t answer the question.

It’s like teaching math in schools. You can say “common core” math sucks, you should do it the way I was taught in school, but they are trying to use “common core” (or have to), that doesn’t help and is in fact the wrong answer.


TR will not share how AI FTP works, nor tell you which workouts most influence it. As other note, FTP is a physiologically defined thing, but IME it’s better to consider AI FTP+PL as just the baseline on which your TR plan is built. I think it’s a useful measure of training progress, but I don’t think it actually measures FTP all that accurately.

I have also found AI FTP to be pretty bad and variable when based predominantly on outdoor rides. If you want it to be consistent and reflect your fitness over time, you’re best off doing at least 1-2 structured indoor workouts per week.

A TR support person did suggest performing Marion at an FTP that you think is accurate as a good way to tell AI FTP where you should be. Of course you can also just set your FTP manually for a cycle…

1 Like