Concordo 100% com o relato do “TKilian”. Comecei a usa o TR este ano (2026). Gostei muito da plataforma no geral, mas esta previsão de FTP nunca consegui confiar nela. Também tenho dúvidas como o TR define os treinos, se a prioridade é a busca de um FTP maior ou como no meu caso a preparação para uma prova específica no calendário. Na minha visão isso é um conflito. O TR deveria evoluir sua configuração e fazer com que o usuário pudesse definir qual a sua prioridade: evoluir o condicionamento físico, evoluir o FTP ou preparar o “motor” para uma prova específica do calendário.
I 100% agree with “TKilian’s” account. I started using TR this year (2026). I really liked the platform overall, but I’ve never been able to trust that FTP prediction. I also have questions about how TR structures workouts—whether the priority is chasing a higher FTP or, as in my case, preparing for a specific event on the calendar. To me, that represents a conflict. TR should evolve its settings to allow users to define their priority: improving general fitness, increasing FTP, or tuning the “engine” for a specific race on the calendar.
What if I subscribed on late december 2025, for 200€ a year, because they claimed in January 2026 there is fhis new AI coming or even before new year? And I am now locked in with a system that predicts my FTP on today will be 120w, after logging 50 days of workouts, daily. And then I do a Ramp test it shows 180w. And all wo vere easy.
Have you been filling out the post workout survey?
The system has been really good about adapting workouts if you’re finding them too easy or too hard but it could take a few workouts until it gets it about right in how it should feel.
Are you talking about FTP Prediction or FTP Detection?
If you are talking about Prediction, I would say turn it off if it isn’t working for you, it won’t impact your training. It is also a feature that no one else is doing so you aren’t missing out by using TR vs some other platform. What does your training plan look like that it’s prediction such a huge drop?
If you’re talking about Detection, I would say that being off by 60W definitely sounds like a bug and I’d contact support.
I paid 200€ for prediction to now turn it off. I was a user for a long time. Unsubscribed in november 2025. Subscribed again as I said. And now to turn it off? Would I get my 200€ back?
As I said, I did all my training on TR. No plan. Some walks, but no hils. Nothing with high HR, very low. All WO, easy.
It predicted in advance and corrected prediction until the last day. Which I would expected value would converge to as close as it could. On the date of predicted result showing 120. On test 180.
It does not bother me to see it on screen. I am LOL. It is behind the screen something must be very wrong, don’t you think?
I think I might know what is a big problem in system.
I don’t think the prediction was advertised at all to me, but it may have been to you. I pay my subscription for access to the workout library, plans, calendar, and workout player. The prediction is an interesting item, but I wouldn’t consider it in my purchase.
If you don’t have a plan, do you mean you have no workouts scheduled? TR won’t be able to predict your FTP accurately if they don’ t know your future workouts or think you have no future workouts (because you aren’t following a plan).
Did your prediction match your detection at this point? by test do you mean a 20min test? ramp test?
It sucks that people are having problems with the prediction.
As of now I still think it’s a valuable tool and like the ability to turn it on/off.
I hope that TR figures out that outside workout bug that is affecting it at this time.
We’re doing what we can to get this resolved as quickly as possible.
AI FTP Prediction is just one feature, and although it was released to the public, it’s still in beta. We’ve learned from going this route.
I hope the other features TR has to offer justify the price of a subscription. Let me know if you’re not satisfied with what you’ve got going on, and I’ll do my best to work with you.
Yes, it has been advertised. You might have missed it. Nothing wrong. I am sorry I won’t take time it is not that important to me if you have missed and because of that you hink it was not advertised.
In the morning on 5.7.2026 system announced my new AI-FTP is 118w or 120w. Based on 2 months of continuous and consistent workouts on trainerroad platform. Easy workouts. No TR plan. Just pure data from it’s own workouts measurements.
On the same day i make a FTP Ramp test on TR platform. In what range would a user expect a test result be? Something between 115 and 125, 130. It knew all of my watts and HR data for 40+ workouts. Result of a test was 178w.
Thats like shutting off the lights because they don’t work . Trs ftp detection before if not perfect was at least consistent . It actually was relevant . This is not. I’ll take @Nate_Pearson on his word that they are working hard to fix it ,but in the meantime shutting it off does not fix the problem . There are many ways to skin a cat it’s just that a stick of dynamite not exactly what we are looking for
I wonder if this is the root of the issue? 2 months with no hard workouts would be difficult to detect your FTP. Out of curiosity can you post those 2 months or make your calendar public?
Have you been using the same power meter for all your workouts?
yes yes for sure it is bad, it’s a smokescreen. Follow solid conventional training principles and stop falling for AI promises that are based on bad algorithms. I too myself got too many stupid AI numbers back from TR and realised sometimes AI is marketing > reality
Remember that AIFTP prediction and detection are not independent. The detection is nothing more than the prediction for today, so if the prediction is buggy (or in beta), so is the detection. That’s why TR is talking about FTP drift in affected users. Prediction affects (is) detection.
it’s almost like the Heisenburg uncertainty principle. You either have great accuracy and very few tests (1hr at your supposed FTP that you never want to do) or lots of tests with low accuracy (a day to day AI guess).
Or, the more you test FTP, the less time you have to train and shape your fitness in ways that aren’t related to your threshold.
