I felt the same… nearly 49… finally doing structured training… busting a gut to hit the watts… prediction is high enough to be motivational but not excessive… gradually fades into nothing.
I understand bugs… but it does feel a bit random
It’s definitely not only for those with outside workouts. I’ve had the same issue since AI FTP prediction started, and no outside rides through March. Up to now, every block starts with an approximately 10W gain, which inevitably drops to 0-2W. I’ve given up on the prediction completely. I had turned it off, but out of curiosity turned it back on.
This block started with the usual 10W gain, and shockingly it has not gone down, with a mix of inside and outside workouts. I have 4 days, including one hard workout, before detection day. If it sticks, the FTP will be what was predicted after the second block, back in Feb. I’m not holding my breath
-Tim
Ramp test and show the ai who is the boss! It immediately gave me a bump.
Joe
It’s interesting (apologies, I may have said this previously … i lose track of which threads I have commented on!)… the bug which meant people were given the “second best workout” affected the FTP predictions. I almost feel like the AI generated workouts weren’t testing me enough - hence my FTP prediction dropped and dropped. I did a Zwift crit race and my FTP prediction rocketed! Doing a ramp test would have the same effect i suppose.
I’d be happy to look into your account to see what’s going on, but if you’re doing your workouts outside, then you’re almost certainly running into the bug that’s being discussed here.
Can I share what I find with your account in this thread, or would you rather I DM you with the details?
I’d be happy to look at your account as well. Let me know if I can discuss it here. 
for the first 2 months the prediction was accurate when I was doing inside rides, however for the folowing 2 when I changed most of the rides (on the day) to outside rides and the prediction kept reducing. The detection was always the same figure
Based on this, though, 
that sounds like a perfect description of the bug. 
The same goes for you. I’d be happy to take a closer look. Let me know where you’d like to discuss what I find.
Eddie, can you tell us more about the bug? I do nearly all of my training outside, and would like to know what is going on.
Sure. You are welcome to look into my account.
I also opened a ticket with Support on the issue, this morning.
I’ve got 2 days to my FTP detection and and the prediction has steadily been dropping until after todays outside ride it droped to 5 watts below the last prediction. There is no point in sharing my calendar because although I’ve got a training plan loaded now I’m outside I’m just doing my own thing which one long hard ride every week followed by rest days and easy days. My CTL is rising and tbh I’ve never felt stronger. If the actual detection shows a modest increase I’ll be perfectly happy. If it reduces down to the level of the prediction I’m just going to give up on it and use another platform until it’s sorted out.
Eddie, no problem in showing or discussing anything.
Intially it worked for me (inside), good steady predictions that matched the detection.
Over the last couple of months (outside) the prediction started dropping regardless of what I did (exactly target or exceed target) and the detection matched the prediction.
The last few weeks I’ve gone rogue (in frustration 
) and not stuck to anything TR has recomended.
It’s a shame as I’ve been subscribed and used TR since 2014 (with great sucess previously 2022 
)
Can anybody please explain what the bug is misinterpreting on outdoor workouts?
Is there a workaround to overcome the bug? What this actually means for the suggested workouts on the plan?
We share your pain 
There are about a half a dozen threads describing the same issue. Many, many users are frustrated. TR is addressing the bug but no workaround (for example, manually bump your FTP so the workouts are harder, always rate everything “EASY”) is suggested.
I have been through several training blocks with the above described result: Initial prediction seems overly optimistic. Follow plan to-the-letter but with interval sessions outside and recovery inside. Gradual drop in predicted FTP to +/- 1W of starting FTP. Detected FTP matches the final predicted FTP.
If I understood it well, the FTP detection is not affected by the bug and giving correct numbers, even when doing outside workouts?
That’s not my experience, the detection has always matched the prediction.
UKCarl: Do you mean to say that over the course of the plan as your FTP prediction has dropped, when you finally got to the “Detection day” that the Detected FTP matched your declined-Predicted FTP? i.e. essentially no increase in FTP over the course of the plan, according to AI FTP Detection?
Exactly that. At the start of the 4 weeks I was showing approx +30w. It slowly delcined with every workout to end up being maybe +2w. The detection has then always matched the prediction. This has happened the last few consecutive blocks.
@bobgade The bug affects athletes who do their TrainerRoad workouts outside. If you have your workouts scheduled to be done outside, you’re pretty much certainly going to be running into this issue.
Essentially, we’re looking at those planned outside workouts incorrectly, and we give you an FTP Prediction that is higher than it should be. Then, after each workout is completed, we drop your FTP Prediction slightly until you reach the day of your scheduled AI FTP Detection, where the prediction and detection converge.
It gives the impression that you had all this fitness to pick up ahead of you, and then you’re missing out on it bit by bit each day. The issue is, though, that the initial prediction was too high, so it’s not that you’re necessarily doing anything wrong in your workouts, but instead, we’re recognizing that the prediction was too high after each workout.
In @UKCarl’s case, it sounds like things all worked as expected when workouts were getting done inside, and then went wrong when they were done outside.
@nopacer Let me know if this helps. As of now, there isn’t a workaround for this issue, but we’ve got a fix that we’re testing now, and it seems to be working really well. I’m hoping we can get that out very soon. 
Thanks, Eddie. Is there a fix on the way?
Yes. I don’t have an exact ETA, but it’s in testing now.