Why do an FTP test when TR has AiFTP that knows it better than you can test it? That just wastes a training day and confuses the Ai with data it does not need.
FTFP and get faster fast.
Yes, thank you. Now it works again.
I will simply reject AiFTP suggestions in the future. Then the problem should no longer occur. I simply use my FTP from ramp tests.
The powermeter for outside works correct. The interval I was supposed to ride indoors would have caused me problems outdoors, too.
TrainerRoad includes these RAMP Tests as part of my training plan. The AI‑FTP never works for me so far.
In fact, I trained for a month. My estimated FTP was reduced more and more during training.
The FTP test then testet 159. One day after the AI predicts 135.
The AI already intended to reduce it to 152 in the next training. Until it recognized 169 a few days later.
Right now, the AI thinks, this is the correct FTP:
So… I trust the FTP test, and the intervals with this FTP also feel correct.
I do think AI FTP Detection will work for you if you keep consistency with your power sources. Again, what I’m seeing suggests a pretty significant discrepancy between your indoor and outdoor rides.
All AI‑FTP changes in my previous post, except for the RAMP test, were performed outdoors using the same power meter.
Hopefully someone from TrainerRoad will eventually work on the AI. Even if the power meters on all three bikes were faulty, even if they feel correct to me, both Garmin and Strava reliably determine the FTP accurately. Only TrainerRoad gets it completely wrong.
But I’m not a software developer at TrainerRoad. Someone will eventually figure it out and fix it.
Please let me know if you want to work on this together. I’m doing my best to communicate what I’m seeing with your account and training data.
My suspicions appear to be correct, given that I was able to get things working pretty quickly by removing two activities with bad data, and without addressing that issue further, you’ll likely end up back where you were after more outside riding.
I’m here to help, but we have to work together to achieve something.
Don’t hesitate to reach out. 
I don’t know how to work on it.
I do 95% of my workouts outdoors. I use two different bikes outdoors with two different power meters. AI FTP is always wrong. It’s not a problem for me, because I ignored the suggestions so far. You also gave me the solution in case I mistakenly accept an FTP again. I didn’t know that I could retroactively specify for individual trainings that they should not be considered for AI.
For me, that’s a good workaround.
If you want to work on the AI, take a look at the following: My FTP is 159. Determined by a ramp test. Then I did a few training sessions. And suddenly there’s this one ride that the AI uses to estimate my FTP at 169. Your theory is: It’s because of the power meter. Now you have taken this ride out of the calculation and I can train again with my entered FTP of 159.
However, I will ignore that; AI again believes it can calculate an FTP: 132
So, a ride raises my FTP from 159 to 169. Removing this ride from the calculation reduces my FTP to 132. And I also think to know why AI comes up with this idea. I had discontinued a training based on the FTP of 169.
But you will have to admit that this AI detection is at least very special. My cycling buddy says: If an FTP was 159 two weeks ago, the probability is high that it will still be 159 two weeks later. And I think it is therefore significantly more accurate than the AI at the moment.
If I get the chance, I’ll compare the power meter with a watt‑measurement pedal and give you an update. At the moment, I actually don’t have any here.
I think that maybe some of my messaging is getting a bit lost here.
Stepping back a bit, I think it’s first really important to address your power sources before we land on what your FTP is. You’ve been saying that it’s 159, but which ride / power meter is that based on? Your FTP might be 159, but it could also be 169 or 132. I can’t say with any confidence until you determine how your power meters are all reading collectively.
FTP is all based on power, and if each of your power meters is reading differently, then there will be no consistency in what your power-based performance looks like, and we can’t work with inconsistent data.
What I’d recommend doing is finding a way to compare all of the power meters that you use during training (it sounds like there may be three (two on bikes for outside rides and one smart trainer?)). Make sure they’re all calibrated to the manufacturers’ specifications before testing, of course.
Once you have that data, you’ll know if you’re getting consistent data or not. Right now, all we have is a handful of rides that report significantly different fitness based on the power data, and there is no way to know which is “accurate” from our end. If you were only using one power meter, accuracy wouldn’t matter as much, it’s just about getting consistent data.
The only thing that’s clear to me from my end is that the data you’ve provided so far has been inconsistent enough to stop our model from being able to confidently prescribe your training.