I hadn’t really seen too many issues with this but I am somewhat curious. My current plan had predicted a somewhat large FTP bump, which in itself I’m not sure is 100% accurate, but at the start of this week I was at a 339 FTP target. This week I’ve done all my workouts outside, 1 V02, 1 sweet spot so far, both hitting intervals right at or above targets, and the prediction dropped to 334.
Honestly I’m not too worried about it as it’s probably still in the ballpark, but I did think it’s interesting how hitting workout targets and rating workouts with “hard” (as I understand it this is the “goal”) would drop the prediction
Yes, I submitted a couple. It was predicted to be 319 with no workouts left until the day. Just got my prediction today and it dropped to 316 for no particular reason. Guess I just stick with it.
The next prediction in a months time is again a 10% increase to 348. Honestly don’t think this feature works very well.
It’s seems to have been fixed for me I also showed a increase of 1 after a 2.5 moderate ride with my pal this was prior to a week in Mallorca riding 4 days & then a week off on my return
I just encountered an issue that’s not quite the same (so I wasn’t sure about using the form), but seems related. I have a plan with 100% outdoor workouts.
My 278 FTP from before starting the plan a few weeks ago felt low, so I detected and got a big increase to 297: seems a touch high, but not unreasonable.
28 day prediction before FTP update was 325. Seems ridiculously optimistic already, and I didn’t take it seriously given the known issues. After FTP update, my 28 day prediction jumped to 341! Completely crazy!
Surely updating current FTP should have minimal effect on predicted FTP?
Happy to see that TR sees it as an issue. I used nice words, but was extreme unhappy. I can say, that I missed a 100% perfect prep for my 2026 A-race. That said, I know there is a bug for outdoor rides and I mainly ride outdoors. I can say I stick to the plan and when not, it is because of traffic lights, cornering (so not pedaling) or when my workout finished and I’m not home yet. When now and then doing a club ride, I’m sure I hit the target IF - TSS. And still my ftp prediction drops. When getting a new (much lower) ftp, I accept it to not disturb the model (advise from Eddie).
I found it strange that Jonathan is getting his ftp up and all the paid member their ftp is getting down. He also noted in one of the podcasts he is using a “secret sauce” that we as paid members don’t have.
What I do want to know is, when is the problem solved? I pay a lot of money for a tool that’s not working 100% (and I already do this from Feb on). Was already saying goodbye and looking for other platforms, but if I know when the problem solved, I will stay…
I know, that I use the “I” very much, this is because this is my feeling, my words, my thoughts etc.
I had an AI FTP detection for today. As of last night/this morning it was supposed to be 313/314 (can’t recall). Today I crushed a challenging 5.5 hr race. Before even rating the workout, it dropped the prediction down to 309.
My prediction in 28 days is a whopping 334 watts. Will put me at about 4.75 watts/kg lol.
My FTP prediction dropped from 270 (+4) to 264 (-1) after I uploaded my unbound 100 ride via Strava. I’m doing the Gravel - increasing FTP plan with 4 days of vacation during the next month.
As an experiment, I deleted the Unbound ride (345 TSS day) and then the AI FTP Prediction changed to 272W FTP Prediction. I had a red day following Unbound where I was already on a rest day. The rest of the plan is unchanged.
I also tried marking the Unbound race event as completed with as close to the TSS as the actual race, and the prediction went back to 272W
AI FTP has been weird for the past month. With the nice weather I’ve been doing more workouts outside and the prediction always goes down after. I’ve been trusting the process but I’ve been stagnate for a month now. Yesterday I did my scheduled VO2 max workout (predicted to be Hard) and I rated it Easy, which it truly was, and AI FTP still went down.
There is a bug with outside workouts that matches your described behavior perfectly. We’re currently working on a fix, and it’s the company’s top priority right now.
At the top of this thread there is a form that we’re asking athletes who are running inot this bug to fill out. If you’re able to take the time, please do that. Every bit of info we have helps!
FTP prediction dropped by 2W after completing a VO2 workout ( rated moderate). Submitted form above and predicted FTP went up 1W. At least our AI overlords have a sense of humor
Certainly won’t update here on every ride, but did notice this was the first outside ride that didn’t drop my AIFTP…not that I’m that fussed. I decided to do my 1.5hr 80TSS (James) endurance ride outside. Nailed the time and a TSS of 83. Matched the workout as Easy and prediction stayed the same pre/post ride. Even stuck a little note saying 338-338 and might track that way going forward.
@eddie please dont put time in all kind of forms. please solve the bug!! I know, its not you, but as you read we are all getting frustraded. And the more frustrating is that @Jonathan is still telling it all soooo good he is getting xx watts and on strava we see he is also riding outdoors, he is doing other things beside the plan (reno devo etc), so why is he not having the bug and we as paying member are already using (and paying) for months for a not correct working platform? This morning again, frustraded. After my workout (which I have to change because of rain and thunder) my ftp dropped 2w. oke, no problem, this moring I opened TR and again drop a couple of watts (only by going to bed and have a sleep). My whole plan from detection to detection, the drop in watts is 12w and brings me almost at the beginning of the cyclus and if you take the whole time (from start AI - thingk Feb), my ftp drop is around 20w. maybe an idea to make a podcast that explains the problems, says whats been doing to fix it, and show us a future plan, when can we expect this is solved.
Think a refund or discount is maybe a good idea!!
Yesterday I submitted a report after noticing something odd.
After successfully completing an outdoor workout, my FTP prediction dropped from +2.5% to -2.5%.
Today, while reviewing my calendar, I noticed that I had pinned the workout planned for Thursday. I unpinned it, and the AI immediately replaced it with a very similar workout. The interesting part is that this also fixed the prediction, which went back to +2.5%.
The new workout (Wheel-3) is probably 95% identical to the previous one (Three Sisters-1). Yet Three Sisters was apparently not recommended and was associated with a lower FTP prediction, while Wheel-3 kept it positive.
At this point, I’m tempted to do Three Sisters anyway, since it doesn’t look significantly harder, just to see what happens .
Not trying to be rude or a jerk here. But why not just turn off FTP prediction. It is still in “beta.” My opinion is they’ll at they should turn it off for everyone and then allow people to opt in knowing it is a beta feature and has bugs (@eddie@Jonathan ) FTP detection is working as planned.
also your FTP didn’t go down 2 watts. The prediction did. IMO people are way to focused on day to day movement of an imperfect predictor. The new system definitely picks better workouts and the FTP detection has been spot on (at least for me). I turned off FTP prediction about a month after release. Now I just let the system do its thing and see how I’m doing once a month with the detection.
Turning off the ftp detection is an option, but if the system in the background runs the same predictions (but only you are not seeing it), it is still not good.
.
