I was also wondering how this was implemented and whether 13 days and 23 hours is enough to get another detection or not. The app should tell the user exactly when it will be available again. Two entries appeared in my FTP history with the same FTP - one on Monday, March 28, when I had it detected and another one on the next day when I had the first workout with that FTP. It’s not the first time I’ve seen a double entry like that. I deleted the last one in the hope of being able to detect FTP yesterday but even today I can’t (at least for now).
1 Like
This is a little PSA and I have not read through this whole thread. Maybe this has been covered.
I’m running TR Windows app on a dedicated trainer mini-PC. When I do FTP detection, it will throw me a replacement workout based on the progressions levels prior to doing the FTP Detection. I have to follow this procedure. Click the FTP detection, accept the value and then restart the app so it will adjust my progression levels. Then it will pop up Adaptations Pending, accept the adaptations and finally will be ready to do the replacement/adapted workout.
Ron
So I am two weeks out from my next scheduled ramp test but I was curious and decided to arbitrary schedule a ramp test yesterday just to use the feature. Saw what my calculated FTP, didn’t accept it, and delete the ramp test.
Today I decided to do it again and saw the pop up saying I can’t use the tool for 14 days.
Bummer. I have not scanned this thread deeply in a bit and I am sure the TR staff put out documentation saying this was a new thing when this was implemented. But with that being said I wish there was some sort of confirmation message in the UI saying something along the lines of ‘Are you sure you want to run this now? If so you won’t be able to do it again for 14 days’. Now I won’t be able to run the tool for when my actual ramp test comes up. I guess it serves me right for trying to open presents before Christmas morning…
2 Likes
Hi trainer road team
So I’m kinda not sure if AI FTP detection is working properly with me. So I got my first AI FTP from 349 to 356 which I was stoked about. Then a few weeks later I did it again and it was suggested to be down to 350 which I didn’t except. I checked it a lot. Which I’m sure wasn’t the right way to do it. The workout were feeling good but nothing to difficult i do most of my workouts outside so most of my workouts are consistently off my bike power meter. Which is regular calibrated.
I wanted to check how accurate the AI FTP. So I went to do a all out effort 20min ftp test outside and managed to hold 394w. Which puts my ftp at around the 374ftp mark. This is significantly higher then 350 the AI FTP detection is wanting to put me at.
So my question is did I kinda stuff the AI FTP detection because I checked it to often. I’ve since Hurd in the pod cast to not do it all the time and just finish a block then start another block to then use it. I wonder if this is why my numbers are so far off.
I understand ftp it’s just a number to get your training in line with the right about of stress.
Any input would be appreciated
Check with support: support@trainerroad.com My 2p though is that FTP indoors can be constrained by lack of cooling and freewheeling compared to outside (get it right, cooling, etc they are actually the same). So perhaps its using mainly constrained indoors workouts to detect your FTP and that will be correct for you to train to if you usually train indoors in a constrained environment.
Can confirm and 15 days after last running it (but 13 days after that mysteriously 2nd entry) i still can’t use ftp detection
I have contacted support and they are aware of this duplicate entry issue and are working on it. Unfortunately it resets the FTP detection timeout whether you delete that entry from your FTP history or not - so mine was unlocked yesterday around the time I’d started that first workout 14 days before.
1 Like
Last week had big cycling vacation. No structure, just typical 8-13h/week → 24h/week, power-based 600TSS → 1400hrTSS, 11km ascending. Rode with rented bike, no power meter, only HR monitor. It means no AI FTP detection. In fact, AI even proposes to lower FTP by 2%. While it is possible due overreaching, I don’t think it is true – did HR capped ride yesterday (~72% of maxHR), usually it is ~170W, this time it was ~185W. Plan for next 4 weeks are Z2 rides (weekly duration 11h → 14h).
So, considering above, what is the most sensible action for now?
- accept FTP decrease because slightly lower/higher Z2 does not matter much, or
- keep going with current FTP with hope that over next block AI will detect new FTP by lower HR eventually, or
- bite the bullet and take ramp test to get it right for now (I don’t like testing much, though)?
I think there’s a tiny bit more conservativeness built into the AI-FTPD at the moment. Just a hunch based off posts seen here and my own FTPD. I don’t think it’s a bad thing necessarily, though hard to get your head around depending what you’re up to.
How high are your PL’s?
- Have you contacted TR support to see what they say?
That’s the issue, because all rides were without powermeter, all PLs dropped due inactivity. During trip I probably got around 5h SS + 1h Threshold due ascending, rest was mostly Z1 + Z2 mixture, estimated by time in HR zones. But all this is wildly inaccurate, it was first time for me to hit the mountains 
Yeah, will do that 
Makes sense. Obviously WL V2 will fix a lot of this.
This is all completely guess work, but I still think the system is just dialling you into a lower (potentially more accurate) FTP that would lift your PL’s. I know they are lower than they should be, but you’ve got room to go up, especially Threshold. I’d accept it, and just give the appropriate response to any workouts you do. As you’ve already stated, you’ll mostly be doing Z2 for the next month and can ride based on feel, HR, and more loose Power limits.
Remember, AI-FTP is still in beta so will be getting continuous changes and improvements as they get more data from successful/unsuccessful workouts post FTP update etc. It may be your previous result was too high based on their system data. If it was dialled you will soon see AI lift your workouts and that will help improve the model too.
1 Like
Yep, with HR cap, will get intended effect for current block one way or another. I’m more afraid with unknown FTP power:HR will generate inflated TSS that may confuse AI FTP detection later. On other hand, it actually might help AI to correct it as well 
1 Like
I don’t think AI takes too much notice of TSS at this points. I might be wrong.
As you say, your data can help improve the system. It might also be that your TSS has been inflated to now. While you can complete the workouts, you might be riding a higher percentage of FTP than you thought.
If Support does come back to you, I’d be interested to know if they confirm my suspicion of improved FTP accuracy with more data etc.
Good luck with your training. I’ll follow along.
1 Like
The FTP detector does take into account non- power meter riding. They’ve said it takes into account everything.
In your specific case I would just wait a couple of weeks and check again - my experience is that it doesnt deal well with acute (<2 week) changes but if you do a week or two more rides it will do it no problem.
1 Like
any comment on how AI FTP detection works after two weeks of sickness? Not sure I want to bury myself with a ramp test but wondering how accurate AI will be.
Use xert. It has HR derived power metrics.
Got response, in nutshell: to be considered in current production AI FTP detection state, ride needs to be matched to existing TR workout. Did not ask for it explicitly but i guess custom created workouts should be ok as well because TR will assign PL to those.
And kudos for TR customer support, they are always very fast 
Anyway, for my own special case: will continue with existing FTP because this number is not important for upcoming specific block and hope AI FTP detection will pick up wherever it is before next block that involves power based higher intensities.
-
Be careful with that guess. PL’s do get assigned to Custome workouts, but there is a known bug that those levels are not always accurate. Its possible for them to have incorrect values that then skew your PL.
-
The pending Levels 2.0 theu keep mentioning is supposed to address the problem.
1 Like
Interesting to compare the estimated FTP results from different platforms, each with their own differing methodologies.
A couple of weeks ago I was offered these estimations:
Xert = 346, Intervals.icu = 339, TR = 322
After a recent race with 5 & 10 mins power PB’s (by 19 & 39 W, respectively), the updated recommendations were:
Xert = 365, Intervals.icu = 361, TR = 317
I suspect that it was probably mid-high 320’s on both occasions, just expressed to different degrees (i.e. not fully pushed prior to the race). I have no appetite or plans for a real FTP test anytime soon, so no interest to validate. Quite a different set of results.
1 Like