So 3 days from prediction and it was predicting 265 (about 3.6wkg) )if I completed a VO2 today and 262 (about 3.5wkg) if I sub it with an endurance with bursts today - which I will be doing. Building a lot of trust in the platform and AI model. Workouts have been perfect but life stress and nutrition have been a bit of a drag.
Peak prediction was 271/272 but I couldn’t stay on top of it in fact - combination of low energy/fatigue and I think that led to some knee niggles (as probably did doing a sportive at the end of a rest week!). I am going to try the new Energy Availability calories model because I was feeling absolutely whacked coming into Easter.
Just coming off a rest week now, where again work/life stress was high and I did zero cycling, just my daily 5km round trip walk to the office. Knee is a lot better thankfully. Entering two week taper for my “A event” (group trip of 3-4 days back to back big rides in the Tramuntana mountains in Mallorca) so a handful of high intensity rides of decreasing duration are coming up.
Let’s see what Wednesday’s FTP detection has in store
Hoping I can continue gains in the summer if I improve the fuelling. NBD if I can hit 4wkg!
Ive had another detection today. The initial estimate was 269w but after missing a workout it fell to 264w today (a 3w increase). The new Predicted AIFTPD was 274w but after squeezing in a VO2max session after my ride today its went to 283w but that session was in lieu of Tuesday’s so I’d expect it to fall back to 274w when I don’t do Tuesday’s and fall further when I do a TT instead of my Threshold session on Thursday. That’s what happened when the TTs restarted two weeks ago.
I don’t think AI FTP detection takes training history into account, as the kind of improvements it is suggesting for me are more akin to noob gains. I have only been back with TR for a month (after a year hiatus doing my own training). The kind of FTP improvements it is suggesting would make me a world tour pro in a few months…
I joined with an FTP around 370w in March (taken from an hour test). It suggested 392w as my initial FTP using the AI detection, which seemed high, but the training sessions felt great. After going through the plan for a month, it put me at 409w, and is suggesting this increase in 4 weeks time. I understand it is just an arbitrary number, and if the sessions feel good, then the number is right. However, in this case, it shouldn’t be referred to as FTP, because these gains seem far-fetched.
Do you train exclusively outdoors? There is known bug for outdoor riders, AI FTP prediction is inflated, but the actual detection should be accurate. Maybe system needs more time to access your FTP, its like the weather forecast with time its getting more and more accurate. If training sessions feel good i think that you are on the right path.
december 15, restarted TR after using a coach for the past year (ill likely resume with him but he has some life things going on). 217 which felt insanely easy as my previous setting on training peaks was 260.
jan 19 aiftp: 232
feb 19: 238, predicting 250 3/19.
3/19: 260 (which is where i was last fall)
new prediction for 4/16: 261, which seems like a vastly different projection from the previous 2 increases.
4/19: 201 detection. which seems weird because I have had (3) 3 hr races in this month with NP of 210 each. Also I have not gotten a new prediction for the next month just “needs more data”. so I am VERY confused.
March 6th - 351 AI FTP; 363 AI FTP Prediction
April 19 - 357 AI FTP; 366 AI FTP Prediction
I ended up delaying by AI FTP detection because I had a week off in the middle of a training block for a ski trip. I decided to wait until after completing the next training block. The addition of weekly Wednesday night group rides a few weeks ago, replacing one hard trainer workout, doesn’t seem to have dramatically thrown anything off.
It all seems to be working just fine for me. I agree that AI FTP Prediction starts off a little too optimistic, but that really doesn’t bother me. Sometimes I just turn AI FTP Prediction off if I find myself fixating on it too much.
I believe that there is a bug that affects athletes with FTPs greater than ~400 watts. It looks like that’s probably what you’re seeing here. If that 409 figure is working well for your training, feel free to stick with that for now. Let me know how things are going..
Hi @eddie this happens to me every month. Can you tell me more about the bug with FTPs >400? This is where I am and my prediction is crazy high at the start of a window, only to drop consistently over the weeks until prediction and detection align at only a clip dm of watts increase. Which is ok and the kind of progress I would expect. Would just like to understand what the bug is doing and when it might be fixed.
my prediction is crazy high at the start of a window, only to drop consistently over the weeks until prediction and detection align at only a clip dm of watts increase.
This is actually behavior from another bug that affects athletes who do their workouts outside.
It doesn’t look like you’re being hit by the 400+ watt club one too badly, and as long as you arrive at a FTP Detection that’s reasonable, I’d say that’s the best-case scenario here at the moment.
I know it’s annoying to constantly see your prediction drop, but we’ve got our best people on fixint this at the moment.
The figure definitely works well for my training, but it isn’t reflective of my FTP, as I feel that I could probably only hold ~380w for close to an hour, which is significantly less than the current 409w it suggests. I know it would be a hard sell to significantly drop everyones FTPs, but I think it is necessary to make the value more in line with everyones true FTP. The alternative is to call the AI FTP something else.
Hi Eddie, Jonathan mentioned in a couple podcasts ago, that a small amount of athletes get a more aggressive AIFTP. He didn’t expand on it though. He seemed to suggest, unless I’m read too much into it, that some athletes benefit from using a higher AIFTP than a lower one. Do you have any insight into this? If I use myself as an example, I’m not sure I could have held my last AIFTP for more than 12-15m. and that would be everything I had rested.
I’m not completely sure what you’re referring to, but I do think that some athletes may have been given higher FTPs than they might have been training with pre-TR AI.
I don’t think there’s anything that would trigger us to give certain athletes FTPs in a different way than anyone else, but it could feel different to them depending on how they were training previously.
“Some athletes, and this is going to be controversial, but some athletes should train with a higher power number than their actual maximal lactate steady state, or a lower number than their actual maximal lactate steady state, because maybe they have some other bottleneck in their system that is stopping them from being able to ride like one would expect at maximal lactate steady state. And as a result, if you’re just banking all of your training zones and all of your prescriptions just based off of your FTP, that’s why you might have a bad training experience. This is why we don’t like, your FTP is almost irrelevant to the workouts that you get in TrainerRoad because it’s just looking at your abilities in each zone and then treating it independently.
But this is a big misnomer with FTP and threshold, that it has to be perfectly aligned to insert whatever definition you have. To what you get at an hour riding maximal lactate steady state or, or 5% minus your 20 minute average or 10% minus your two by eight or one, whatever your like 75% of your one minute in a ramp, whatever you say it is[…]”
From Ask a Cycling Coach Podcast - Presented by TrainerRoad: How to Build Cycling Workouts (Most People Get This Wrong) | Ask a Cycling Coach 578, Mar 19, 2026
I may have taken my interpretation too far from his intended meaning, but I’m still baffled that a small number of users, including myself, have been consistently getting an AIFTP way higher than what they could possibly hold for close to 60min. You mention above that most of your athletes get FTP that are slightly less than their predicted 60 min. power - why are there a few folks so far away from their predicted 60 min. power for their AIFTP?
I’ll have to ask Jon about that and get back to you.
When I said that most athletes have FTPs that are slightly less than their 60-minute power, I probably should have eliminated the “slightly,” as this isn’t accurate for everyone. I’ve edited that statement.
What I was trying to get across there was that in many, many cases, athlete’s FTPs are lower than their “hour power.”
I often see athletes training well with an FTP somewhere close to their 30-40 minute range, but there are, of course, always outliers.
