I did a Zwift ramp test to get an idea of where I am after some extended time off the bike. I manually put in the result to start my training plan. I noticed AI FTP wanted to give me an FTP much lower as a result of the long period away. I want my workouts to be appropriate so I manually input the ramp test result but now AI wants to predict my ftp as the same Lower number even though I’ve done the ramp test. I tried accepting the prediction then overriding it but it wants to return back to the same number. At what point can I use the prediction? I only did the ramp test to get a gauge of where I was at. Will it eventually adjust?
As I understand it, manually adjusting your ftp will not make the suggested workouts harder or easier. TR determines an amount of work/intensity based on what it thinks will be productive (ignoring ftp) and and then tries to match that to a workout in the library based on the set ftp. So, if you increase your ftp manually, TR will just choose a workout with a lower difficulty level to make the work roughly the same (although the interval duration/intensity mix will often change). If you want to quickly get to harder workouts, I believe the only path is to manually pick harder alternates. Otherwise, all you can do is keep marking stuff “easy” and wait for the system to adjust.
I just want TR to acknowledge my actual current FTP as when it was much lower all the workouts were too easy. But when I override it it seems to throw the prediction off. I don’t care about FTP past getting me started and going from there but it does matter a little because my SS was feeling like endurance at best in the beginning. When I tested I was 90 watts higher than what it had predicted which is understandable since I had such a long lay off. Someone from TR has responded so I’ll keep the inquiry there.
I think what I’m struggling with is I want to use AI detection but with my actually ftp I tested. Because it doesn’t see a training history(and I did my ramp test on Zwift) it’s assuming an FTP that’s too low.
I didn’t turn it off I just haven’t accepted the new prediction because it will make my workouts too easy. My question is will it pick up my manual ftp and draw predictions off of that eventually? It doesn’t seem to update even having done a few rides.
The ramp test ride was imported into TR right? I was prompted to do a ramp test after some time off, and the AI was able to work with that immediately, but that was in TR, not zwift, so not sure if that’s different.
But the AI doesn’t use FTP to calibrate workouts, it prescribes it based on watts and training history. So if you’re set on starting with the ramp test ftp, just do that. What do you need the AI for? Otherwise just take what it prescribes. Coming back slightly easy (if that’s even the case. Zwift ramp test has run hot for me) isn’t necessarily a bad thing after time off.
Regardless, the AI should dial in after a ride or two, so I wouldn’t sweat it regardless of what you decide to do.
If you can do your ramp test on TR I think that is your fastest option. I’ve seen a couple of times that the zwift one doesn’t populate but doing the TR native one does.
It’s worth noting that the set FTP does have an impact on the workouts. When I manually adjusted down, my sweet spot workouts dropped by several watts for each interval. My O/U workouts had lower unders. When I adjusted back to the FTP detection, the watts went back up.
FTP and workout power targets are not decoupled.
Edit: other things change, too. Like workout durations and such. I was immediately prescribed higher PL workouts. But just giving my own account of what I have seen in my training with TR (Which has been awesome).
It seems like the issue is I started with the Zwift ramp test. I should have started with TR but I wanted the visuals of Zwift while I suffered.
I’ll see what happens after today’s over under but I’m not expecting the prediction to anything unless I guess the much lower ftp. I understand AI doesn’t look at FTP but to some degree it does because if I accept that FTP it wants to give me my workouts will be extremely easy. Then a few weeks before it jumps. I don’t want to waste my time with that. So I guess that’s a lesson learned when doing the testing.
I think what I’m struggling with is I want to use AI detection but with my actually ftp I tested.
I see. This is not possible, because AI FTP Detection is it’s own thing.
The updated AI FTP Detection you’re getting puts you around a level 3 threshold level. This will be consistent between detections, and we think this is a good insert point into your overall zone progressions to get faster.
The old AI FTP Detection was predicting the result of a ramp test, then we would make adjustments to your workout levels to try to make things incremental. That means you could end up with an FTP with a threshold level of around 1-8. Although you could still get good workouts, this wasn’t a consistent way to assess FTP from one detection to the next, or to compare it with others.
And then AI FTP Prediction is based on your recent workout performance, recovery, fatigue trends, and how your current training is expected to affect your fitness over time.
I didn’t turn it off I just haven’t accepted the new prediction because it will make my workouts too easy. My question is will it pick up my manual ftp and draw predictions off of that eventually?
It won’t make workouts too easy since it’s prescribing workouts based on watts and training history. You can manually input your FTP or use AI FTP Detection however, AI FTP Detection places you at a more optimal level so you have wider of room for fitness growth. I would actually try the lower value and see how the workouts feel. You’d be surprised
I may be wrong, but I thought by doing a TR FTP test in TR (ramp, 20min, or 8min), the AI would use that and continue adapting from there. Is that not the case?
What does this mean? I’d wager that predicting ramp test outcomes is absolutely comparable to other athletes. Didn’t the podcast crew do group ramp tests together on stream?
[edit - nevermind! I’m wrong here] ~No, the AI will not use the result of some other FTP test.~ The new AI FTP redefined FTP - it’s not going to match the old FTP definition in most cases.
The updated AI FTP Detection you’re getting puts you around a level 3 threshold level. This will be consistent between detections, and we think this is a good insert point into your overall zone progressions to get faster.
The old AI FTP Detection was predicting the result of a ramp test, then we would make adjustments to your workout levels to try to make things incremental. That means you could end up with an FTP with a threshold level of around 1-8. Although you could still get good workouts, this wasn’t a consistent way to assess FTP from one detection to the next, or to compare it with others.
I agree with this. Of course ramp tests are comparable between athletes. I guess the question is “if athlete A has a higher ramp test FTP than athlete B, what does that tell us about the power curves of A vs B?” I think it’s likely that A has higher 5 minute power than B, but we might not be able to confidently make claims about 40 minute power.
It might be less “consistent” because it’s (a) a single point in time, so it’s more susceptible to fluctuations in sleep, nutrition, stress, fatigue, motivation, etc (b) ramp tests have a big anaerobic component, so (I think!) freshness will impact the result more.
I just don’t use/need the “AI FTP Prediction” – I don’t have any workouts scheduled, so it just has me spiralling into the dirt. This is further compounded that the current FTP detection is useless for me because I cross-train (swimming, rowing ergometer) on the days I don’t ride, so there aren’t enough data points for it to function.
Obviously, the “don’t pay attention just pick a workout” is fine, but it’s obnoxious to have it screaming at me not to pick a workout because of high chance of failure – for a workout I can complete with non-heroic effort!
It just needs to be easier to turn some of these things off vs. force them down our throats because some simulations or statistics “prove” they’re better for most people.
