That’s not correct…you seem to be missing the fact that you are now doing those “easier” workouts at a higher power number. They are only easier relative to a specific intensity setting, not easier from one FTP to another.
Look at the absolute power numbers you are putting out and work you are doing rather than percentage of FTP.
TR is a new way of training - with a huge workout library - compared to traditional intervals based on 6x3min, 5x4min, 5x5min, 5x8min, 4x10min, 2x20, 3x20, 3x30, etc. You progress by being able to do those intervals at higher power numbers. Traditionally folks would do them at a constant percentage of FTP, after a while, reassess FTP and then keep doing them. TR is just implementing a more varied and sophisticated form of that.
You can throw out the GC prediction and I’d still have said 112% for 17.5 min in the middle of a ride is getting into unbelievable territory. And there aren’t many workouts in the library that are gonna stress your system if he already has that capability. Only thing to do is ride a bunch of volume then.
There might be something in the TR AI FTP suggestion, in making the work repeatable and therefore consistent. Just seems a big gap vs CP.
If I use the higher CP estimate I know I’d be at very low PLs. Just not sure how that’d work in the long run.
Either way, I’ve always found CP remarkably good at predicting maximum achievable power for a given duration. Maybe AI FTP can’t do that but it sets good levels for training consistently.
I don’t know much about Golden Cheetah or what interval it bases its estimation on. I have had intervals on its maximum calculate an e’ft for me of 300w + which I know is nonsense. I wouldn’t trust anything based on one offs or short intervals I would rather use one consistent package I am paying for than polute it with other freebies and confuse it further and perhaps break that package; trust the whole package. But as I say that’s just me do what you want to do.
What PL are you at currently? What FTP input are you at currently?
You may very well have low PL at your current FTP input - indeed AT will push them down if you raise it. That could bottom them out. But you are assuming those PLs are accurate. Based on those numbers you provided you should be level 10 in almost anything I can think of at an input of 364. For an exercise, just go and look at what the actual numbers would be at a given workout level in a given power zone and decide based on that what you could do.
AT and FTP prediction are models. They are only as good as the data they get: garbage in = garbage out. I see it happen all the time in modeling, you always have to use a sanity check and should never blindly trust any model.
Also, the reason CP modeling works so well (which does not originate in GC, but is the basis of many studies) is the same reason AT works well, there are underlying linear relationships in human physiology that relates work capacity with work duration. Those are really easy to find, assuming you give each good data. AT is not yet there with unstructured workouts and the FTP prediction is in beta stage…
Anyways, report back with what you end up doing. You have an interesting edge case for sure.
I think if TrainerRoad could get away from calling it FTP at all they would.
If you think about it as just a setting it calculates to set good levels for training then it makes more sense.
You yourself admit you put out your highest numbers outside.
I’d use the AI ftp for training and the golden cheetah estimate to tell your mates.
Out of interest though, have you done any actual TrainerRoad ramp tests? If you actually test at the higher number then that would point to the AI prediction beings low…
I often see quite a large difference in FTP winter to summer. I had been training off 355 and going through the AT process. I can see it’s pretty clever but it does take some time to hone in on the PLs as I was repeatedly switching out workouts to do stretch or breakthroughs. I was around 7 for Tempo and Sweetspot but hadn’t topped those out yet.
At the higher CP estimate I wouldn’t feel confident at all about the 2 hour SS workouts at 7+. But eyeballing the c.4.0 workouts they would look ok (easy to say when sat here typing).
In the past I’ve been cautious about training from the highest possible number, but maybe I’ll give it a go and see if AT can get me through. Hopefully without breaking me
I’ve done most of my tests over the years outside. I’ve done lots of 20min tests and got similar results to the other day.
I really dislike the ramp test. I’ve done a few of them (8 by the looks of it), as I just use them to restart training after the off season/injury if I’m on the turbo. I’ve never done a ramp test on form as by then I’m outside and/or racing
I’d suggest decoupling the FTP input in TR from CP or FTP elsewhere honestly. The only reason I would suggest the CP estimate is that the FTP prediction seems way too low to be useful. And I give AT a lot of credit for handling big margins of error.
If you can look at a level 4 and not be scared, I wouldn’t worry at all and try it. Worst case is you lower your FTP input some and keep the same progression level.
The other part of the test is look at level 8-10 workouts in each zone with an FTP of 364 and tell me how scared you’d be. My guess, is not very for anything up through V02 work at least.
The last part of this is a key piece of evidence - you’ve been using 355 and got a pretty big jump because you’ve been doing stretch and breakthroughs…my guess is you’ve been progressing fine in the workouts and rating them as moderate/hard. This is consistent with my hunch that it predicts the delta from a current setting more than an absolute fitness value - may or may not be intended, but I’d bet a dollar that’s what’s happening.
Any chance you got bad power data from the outside ride? Poor indoor set up compared to outside? Need to rule out those possibilities too…
I’d say those level 8 workouts at 364w look challenging and would require focus but certainly not impossible.
I don’t think it’s bad data for the recent outside ride. They were 2022 bests but not all time bests and it’s consistent with how I’ve been performing against mates. I also spent the winter creating a spreadsheet comparing all my power meters against each other on every turbo ride with multiple recordings. In the end there’s only a couple of watts between them. At least now I know!
Hi,
yesterday I tried the new feature before my scheduled ramp test and it gave me a reasonable ftp estimate. I did not accept the estimate. Due to illness I added two more rest days and the ramp test was postponed two days within my training plan. The test is now scheduled for tomorrow and I am now trying to use the ftp detection as it is not advisable for me to do a full out test tomorrow. But the ftp detection keeps responding that my data is processed and I should get back in a minute. I’ve tried the whole day with several restarts and tried on MacOS app and iOS app. No success. Is there a reason why?
-Jan
I’m having same issue (over 6 hours and still waiting) when my RT rolled over to today because I was traveling and without bike yesterday. @ambermalika earlier in this thread asked people to reply to her if you’re having this issue. I also sent ticket to customer support this morning but no response as yet.
impressive feature. I got 347w prediction and tested at 356w on the ramp. I generally over test slightly so the estimate is probably close to reality and I’d happily rely on it in future if not feeling up for an all out test.
As ever, expectations of some TR users are so high. For this to work you’ve got to consider how AT works and use it appropriately - i.e. be honest about how hard you found a workout.
It’s similar to people who moan that ramp tests aren’t perfect. Of course physiology varies, all you need to to fix the problem is know how much it typically over/under estimates for you and adjust your FTP accordingly if necessary.
