I think you’re trolling now.
I haven’t seen you post any FTP from any other method that is more than 5w from what your current TR AIFTP of 302w is.
ZwiftPower - Login check zftp
Straight up false. I’m doing a custom plan with custom made workouts and still using AIFTPD. My FTP has been rising throughout, both the prediction and the detection every 28 days.
The system is designed to propose workouts that are appropriately hard and productive based on the work you have done. Whether the system is using an FTP 297 or 305 shouldn’t materially change the difficulty of those workouts (because it would be picking a higher level workout at 297 vs. a lower level workout at 305). At least that’s what the TR crew is saying, I have not had enough time with the new system to confirm or deny.
Again, all these systems have different ways of arriving at FTP. And you may have your own preference for defining/establishing FTP. If you have another number you like better that the TR number, I can’t fathom why you wouldn’t just use that number in TR as well. None of these systems are perfectly aligned with how everyone thinks or wants to train. TR has been pretty clear that we can manually set the FTP value and the workout selection should still work fine as long as you aren’t running into limitations in their workout library.
Coincidentally, my FTP is also set (manually) to 302 in TR right now (AIFTP would have me at 318w). And I’m a little half way through base and I’m bumping up against the limits of the sweet spot workout library (my last workout was a level 11). I’ll be increasing my FTP up this week to ~308-310 while TR will be suggesting ~320. Any of those numbers would result in similar workouts in TR, but I know ~310 is more representative of my physiological FTP (as I define it). And having my FTP set a little lower in TR creates a bias toward higher level workouts that tend to be longer intervals rather than shorter intervals at higher percentage of FTP. Meanwhile, intervals.icu will keep me at ~285 because it’s FTP estimation is based on shorter durations and I’ve been doing nothing but long steady stuff this year. I could be “angry” at intervals being dumb and not seeing me do 3 hours at 270w last week (clearly my ftp is higher than 285 based on that), but I’d rather just adjust my intervals.icu FTP to the number I like to use. Again, all these systems are working off different approaches and they are all “right” if you bend to their definition. But TR allows you to set it to the number you like and will still give you appropriate workouts, so you don’t have to bend.
I think you might misunderstand the point of TR FTP prediction. (Or maybe I misunderstand it! We’ll find out 
.) It’s not just literally trying to predict your FTP. It’s trying to tell you, given the scheduled workouts between now and prediction date, what it expects your FTP will be. As far as I can tell, the goal is largely to inform you about your upcoming schedule, both to motivate following the schedule and to allow you to experiment with the impact of tweaking your schedule.
So it’s just giving you the information that, if you don’t train at all in the next nine days, it’d expect your FTP to end up at 297W. I’d actually guess that this means it thinks your current FTP is > 302W. (Seems reasonable to lose ~10W in a > week completely off? It would also come back really quickly!)
I’m pretty sure if i kicked up my feet for the next 8-9 days i could push to easily hold 300w for 90 minutes because i wouldn’t have any training or race fatigue.
If it’s important to you do that and report back 
Hmm, maybe your body is very different from mine, but acute training and race fatigue don’t usually impact my threshold and below power that much. I usually feel acute fatigue more at VO2 Max and above. I’d actually guess that after 9 days off your sprints would be stronger but anything aerobic would take an L… that’s usually the case for me! I think this is also the case for most people (?? low confidence??) so is probably what the AI would predict… unless you have time off like this in your training history already?
For science.
You managing to find one estimate that gives you an FTP on zwiftpower that is all of 9w higher than your current AIFTP detection actually proves the point I’m trying to make…
You need to chill out on this FTP obsession.
Chill out on the app to make you faster by improving ftp?
It seems like you are fixated on a FTP as the “be all end all” measuring stick for your fitness. Lots of people get caught in that trap and TR doesn’t do themselves any favors by making it such a big part of their marketing and so prevalent in the app. And FTP can be an OK measuring stick for performance as long as it’s defined in a measurable and repeatable way. And again, this situation just sounds like TR’s definition of FTP doesn’t match yours exactly. In the context of TR (and most training systems), the primary purpose of FTP is to serve as a basis to drive productive workouts. If you feel your workouts are wrong based on the set/proposed FTP, then either manually adjust it or wait for the system to adjust based on your workout responses. If you keep saying workouts are easy, they will quickly getting harder. As you do harder work, the system will increase your FTP.
I’d suggest trying to be less focused on FTP and more focused on things that are aligned with actual performance. I’m only a month into my season and still in base and will be adjusting my FTP tomorrow, but after that I don’t expect more than ~10 additional watts on my FTP for the balance of the season. That doesn’t mean I won’t be dramatically improving my fitness over the next several months, but those improvements will not result in a significantly higher FTP. FTP is just not a good indicator for race performance for me, all my target races are long. Doing a 20 minute test after riding hard for 3-5 hours is a much better indicator of where my fitness is at. For someone who focuses on shorter races, their key performance metric might be more focused on vo2 max or anaerobic repeatability. Maybe figure out some metrics that are good measuring sticks for your cycling goals and focus on those rather than obsessing over whether your FTP is “off” by 10 watts.
It’s almost like it’s predicting your potential ftp by using the work you have done, absolutely shocking!
You do understand that if you fill in workouts for the upcoming week or so it will update based on those right?
eta:
To further clarify. aiFTP prediction is attempting to predict your ftp based on the planned training between now and that prediction. How you rate workouts, power and hr all influences this prediction
WKO5 is fundamentally different than this. It is modelling your current FTP based on historic data and requires maximal performance across different durations to accurately model your physiology. WKO5 is not predicting anything.
You are not getting any real difference in workout quality with an ftp of 302 vs 307….. those are essentially identical.
At this point it’s pretty clear the disagreement isn’t about data or physiology. It’s about how tightly some people here identify with TR’s AI output.
Every time I present actual performance files, the conversation shifts away from the numbers and toward defending the prediction as if it can’t be wrong. That’s fine, but it means we’re no longer talking about training or modeling… we’re talking about belief in the tool.
I’m not trying to convert anyone. I’m just pointing out that the AI’s behavior doesn’t match real‑world performance, and the reflex to treat its output as unquestionable is exactly why the discussion keeps drifting away from the evidence.
Your current TR ftp and your current wko5 FTP are within 5w of each other, that’s much of a muchness, essentially the same.
Obviously aiFTP can be wrong (see the many many threads with people questioning it) but it will definitely be wrong when you don’t give it anything. (Again the prediction vs modelling current FTP problem).
No one is saying it’s unquestionable.
I will ask again but doubt I will get an answer. Have you filled out your planned training for the upcoming week or so between now and your next ftp detection?
To answer your question: No, I haven’t filled out the calendar. And today’s ride proves why that shouldn’t matter.
I just completed a 2-hour Xert workout (Innsbruck KOM After Party)
-
The Work: 2 hours, 216W NP, Average Power 209W.
-
The Physiology: Average Heart Rate 126 bpm (Zone 1/2). Max HR 139.
The Result: immediately after syncing this single ride, the AI prediction bumped from 297W up to 300W.
This confirms exactly what I’ve been saying:
-
It’s Not a Plan-Based Model: I didn’t add future workouts. I just did a ride today. The AI didn’t find those 3 watts because my future changed; it found them because the decay timer was interrupted by activity.
-
It’s Still Low: Even with the bump to 300W, it’s ignoring that I held 305W for 40 minutes just a few days ago.
-
10W Matters: Regarding your much of a muchness comment, a 3.3% difference (297W vs 307W) is the difference between a Threshold interval and a Sweet Spot interval. Precision matters if we are calling this Smart training.
The system is simply reacting to volume, not analyzing the physiological cost (or lack thereof) of the work being done. A 126bpm ride shouldn’t be the thing that saves my FTP; the 300W+ race efforts should have already established it.
I don’t know how you got to that conclusion.
I’m here to answer any questions you have. You’ve got a lot of misinformation spread across the forum. Please ask me what ever questions you have and I’ll try to help you out.
9 Likes
I’m hoping @Nate_Pearson or another TR staff can answer this. There has been a lot of discussion/debate about what AIFTP actually is. I’ve listened to all the podcasts and I’ve read way too many posts.
I 100% understand the ‘functional’ part of the AIFTP - it’s the ‘threshold’ part I really don’t understand. What is the meaning of threshold for AIFTP. My actual question: does TR have this modelled for how long I should be able to hold this ‘threshold ’‘power for?
I’ve seen Nate give folks their 20m and 1h modelled power, but I have never seen what AIFP power actually is in time. I could have also missed this somewhere completely!
IIRC Nate said they have a set of threshold workouts around level 3.0 (did not say how many or which specifically) and the AI model finds the FTP value that these workouts should feel hard for you. And from there you progress against your last 6 week PB NP curve.
So AI FTP is no “threshold” duration estimate you are hinting at.
The 20 / 60 minutes probably come from a different model they did not talk about. Not sure if they just asymptotically model the PDC or if some AI logic is involved. Obviously if you only do RPE Hard workouts they would need to model something Very hard to All out but not sure if that is formula based or AI model.