I know you guys are working hard on it. Appreciate the reply.
Another data point from my account ( TrainerRoad ).
My current FTP is 354 W. Before today’s outside workout, the prediction for next Tuesday was 360 W. After completing the workout it dropped to 348 W. I know that immediate drop is the bug being tracked here, so I’m mostly wondering how I should handle the upcoming detection.
I’m at the end of a four-week high-volume build block. The last few weeks have included multiple 20+ hour / 1,000+ TSS weeks.
A few questions:
- Should I wait until after next week’s recovery week before running AI FTP Detection? The week includes a priority gravel race on Saturday, and my Speciality block for the Swedish Gravel Championships starts afterwards.
- Am I enough of an outlier in terms of volume that AI FTP Prediction may struggle to distinguish accumulated fatigue from reduced fitness?
- Is a five-week Build block with four loading weeks followed by one recovery week intentional? I was surprised Plan Builder gave me four hard weeks in a row.
For context, today’s workout was Sitkin outside. I completed all four 11-minute intervals within the 331–351 W range sent to my Garmin and rated it Hard. It felt fairly horrible, but mostly because it was around 27–30°C, I struggle badly in heat, and it was the final threshold workout of the block. In normal conditions I think I could easily have completed another couple of intervals.
I’m currently leaning towards keeping FTP at 354 W and waiting until after the recovery week before running AI FTP detection, but I’d appreciate it if someone could check whether that makes sense for this kind of high-volume block.
It’s hard for me to keep up with the forum- but it’s good to know I’m not the only one still having an issue with this. My FTP has done the same thing for the past couple of months- initially predicts a big gain coming in ~ whatever time block, lets say 6 weeks… Then it gradually lessens as time goes by, today it’s telling me it will drop 3 points tomorrow. I do every workout outside. And I complete everyone as prescribed- always aiming for the high end of the spectrum. And I mark the workouts accordingly. I have never failed a workout- but now it says my FTP will drop 3 points…??? Ironically my TR workout is “FTP Build” and has been that for a few months. I’m going to delete or not allow the drop if I have that option.
I’ve had the same nonsense over the last week, today was prediction update day. Chinned off the 1 watt drop and manually input the higher prediction from last week. Rode a 2 hour endurance workout, marked it up as easy. As it was pretty much zone 2 all the way round. Just not sure what to do going forward. The option to opt out of AI prediction seems to have disappeared? 
Find your next prediction on your calendar. Click on it. Click the edit pencil.
Brilliant, cheers its has worked. But had to accept lower prediction first, then go to next prediction and hit the pencil. Then go back and delete today’s prediction, as it had set all workouts to the lower FTP. Even though the manual entry was still there, in the profile. 
Is this still unsolved?
I’m going to train entirely outdoors in the next weeks and reading the above is quite disappointing. This bug has been around for a while now.
Especially considering a) you pay for TR on a monthly basis and b) you have the opportunity cost of using a platform that works the way advertised.
I am a big fan of tr, but wow, it seems since the AI somehow some control was lost.
I know I’m really weak- and have a low FTP. But the point is, THIS is what’s been going on for the past few months. All of my workouts are outside. I don’t have a trainer, live in the south, and ride consistently. This is the first time though it has DROPPED my FTP. Current is 271, wants to drop it to 268- but hey- good news is that in 28 days it’ll be 301!! No, it won’t. That number will slowly slide down ever couple of days REGARDLESS of how well I execute the workouts. And I know about ‘grading’ or marking the workouts. I mark them appropriately.
In your case, I don’t think it matters too much. I’d just run it when it’s available.
Good luck with your events!
How do your workouts feel? Are they challenging enough?
You can always manually set your FTP to whatever feels best for now.
I’d differ a bit to say, fluctuation up and down would be totally understandable. What I’m seeing is that the prediction ticks down seemingly no matter what and then seems to have a bigger drop before the detection.
Interesting…
FWIW, I’m at the beginning of a new block, and if I set my planned workouts to outside, I get a 12 watt increase in 13 workouts. If I set my planned workouts to indoors, I get a 10 watt increase in 13 workouts.
Not a huge difference, but enough of one that it’s clear the AI is predicting higher following outside workouts, which seems to fly in the face of the notion that indoor workouts should be more effective.It is somewhat more annoying when I take into account that a 2-5 watt increase in a few weeks is a more typical result, so the 10-15 watt estimated gain that gets reduced with every finished workout serves as a disincentive rather than a kudo.
Indoor workouts are more effective in terms of time-gain efficiency. If you are completing the same prescription of workouts outdoors, perhaps the AI is accounting for an expectation of extra time on the bike?
Not all outside workouts are exactly the same 1:1 as their inside version at the moment, so there could be subtle differences that are pushing that prediction up a watt or two. Execution is usually slightly different too, so how things actually shake out could even things out.
@eddie to borrow an analogy from another thread: this is akin to saying “the automatic transmission isn’t working on your car so just select gears manually” 
I’m aware that it’s not ideal. It might be the best path forward for some athletes for the time being, though. 
If there were a better option, I’d be the first to suggest it! 
I give up on AI detection it’s complete bullshit at the moment, for someone that only trains outside.
I really miss the previous algorithm that was always spot on.
On my last detection i said it’s enough, it wanted to drop 13 point from my ftp. I decide to put the value from the beginning of the prediction and see how i would feel the training adjusted to this value.
First VO2Max train done and it feels good to know that the train is actually making me feel something.
I have spend 3 months training, doing Vo2Max that feel that i never leave the couch…lol
Experienced the same. At first I bought into the idea that the old plans were too hard despite the gains they produced. All I needed to do was lower the intensity and TSS the AI was recommending and I’d see huge gains. Turns out, no, you have to suffer, a lot, to see improved FTP, VO2 or endurance.
yeah, gets lost in the discussion, but I would say the previous FTP detection was really good, when taken against traditional tests and other systems, and seemed totally agnostic about whether or what plan you were following, which is the gold standard.
I’m sorry about the trouble. 
Let me know if you’d like me to take a look and try to get you back on track. 