On the 17th i got an ai ftp detection of 258 which i felt was a bit low. I had manually set it to 261 and was getting through training sessions pretty easily so was expecting more of a bump upwards. So today i did a ramp test and got 278, so thats a 20watt difference.
Im a little confused as thats a pretty big difference. I have been doing more outside rides recently but they are all with power meter and i update with rpe. How is it possible that there is such a big discrepancy between the two estimations
A 20w delta btwn a ramp test result and âactualâ FTP isnât remotely outlandish.
The question to ask yourself is ⌠would doing 3 à 15 or 2 à 20 or 1 à 40 at 275w result in a workout a) that you could complete; and b) that would be at an appropriate RPE?
Yes, that is also a thought that I have had, much as i love my shiney new number I will also have to live with it. But I still feel like im left in an unclear position. I had my ftp at 261 which I felt like was too low as I was finding the training sessions too easy, AI comes in and says it should be even lower which just seemed crazy which is why i decided to do a ramp test to get âanotherâ opinion and got a wildly different number.
Like you said though, I will try and do some workouts at 275 and see how it goes, the proof will be in the pudding. If I were to just trust my gut instinct I would bump my ftp up to around 268, still a great improvement but probably more realistic.
What troubles me though is that now I dont feel like i can trust the AI model at all, its not just a little wrong its way off wrong. Best case scenario im pretty sure its about 10 watts off which Im pretty sure makes a big difference whern your are trying to eek out improvements from every training session.
So Im really interested to know what is causing this. Is my data not good enough as im doing mostly outside rides? Am I not setting RPE correctly? If it is true that I should not be surprised by a 20 watt difference then I would say that that kind of leaves me in no mans land where really im back to just trusting my own estimation
The most apparent reason as why your FTP went down between your last 2 detections is because most of your riding looks to be commutes, unstructured evening rides + 3 interval workouts in total that were manually picked. In comparison to the other FTP Detections, where your FTP went up, you did a lot more structured interval work.
It looks like you vaguely followed a training plan from Start Date: January 9, 2026, to End Date: March 1, 2026 and after that, itâs been unstructured riding with a week off to Ski, so the FTP drop is expected.
If your goal is to build FTP, I would probably create a custom training plan [click], with at least 2-3 interval training days per week and you should start getting the right stimulus to the energy systems needed to increase FTP.
If your goal is to ride unstructured and here and there do structured interval workouts, then your FTP detection will vary on what youâre doing, but the system will still recommend the appropriate workouts based on your power data if you use AI Workouts.
Iâve returned to TR just recently. The Ai set my ftp at a very low value which I manually adjusted to what it was before. If I enable Ai FTP detection, it resets it back to what it set before.
My question is: is the ramp FTP test authoritative for the AI? In other words, if I do a ramp test, will the Ai now accept this reading for further predictions?
FWIW, Iâm not impressed by the Ai so far. I much preferred TR how it was before..
Basically what you are saying is that unless i have done enough structured workouts recently then AI will not be able to predict my ftp accurately. Thats fair enough and goes back to the old adage of âshit in, shit outâ. My recommendation to Trainerroad then would be that the AI should not even attempt to make an AI prediction if there is not good enough data or at least state this fact, otherwise it is very mis leading and can cause users like myself confusion.
To round out this discussion, I did in fact take @Helvellyn advice and did a 3x15 at my ramp test estimated ftp of 278 and as I think we both suspected I cracked and was not able to complete so clearly that estimate was too high
I did manage to get through to the last interval though and im pretty confident that at 268 I would have gotten through so I think this confirms that the AI of 258 was too low, ramp test too high and the truth was somewhere in the middle
I did the ramp test and AI did accept that as an authoritive estimate but as you can see form my last post the estimate for me was too high.
Over winter i did a lot more structured worksouts and was better at following a plan, AI was a lot more accurate at predicting my ftp and I was pretty happy with it. I thinks it important though that it know when it can make an accurate estimation and does just start âguessingâ.
The AI FTP Detection will detect your FTP based on your training data regardless of whether you are following structured workouts or not.
Now, that said, if youâre not following a training plan and just riding unstructured I would not expect your FTP Detection to be consistent month to month, it will change based on what youâre doing.
What structured workouts + training plan do is give you is consistency over time and FTP gains, which I am not seeing on your Calendar. If youâre just riding unstructured, your FTP will vary. I also donât see the benefit from a riders perspective to track FTP if they are just riding unstructured..
