Forecasting future FTP using AT and AI FTP

Now we have Adaptive Training and AI FTP Detection, is it be possible to theorize what my FTP will be in 4 weeks time if i completed all my training with 100% accuracy?

The above is just an example scenario.

Please excuse me if this has been discussed already, and feel free to point me in that direction.

The TR team are working on it (AI FTP Prediction).

Yup, future FTP prediction is already in the works, per Amber and Nate.

2 Likes

wow… that’ll be awesome and a serious motivator.

Imagine it saying “do this 6 week block 100% and you’ll bump ur FTP by x%”

Yeah, and I shudder to see the times when people complain that they can’t reach that target… and what they choose to blame :open_mouth:

19 Likes

My fear is that it will predict a flat FTP when I feel/hope I have lots of room for improvement. 54 years and 2.5 W/kg, Up from 2.0 since joint last summer, but mostly due to weight loss. I’m hoping 3.0 is attainable and maybe a 300 FTP someday.

There will certainly be some sophistication behind anything that does it. But in reality, the prediction will likely turn out to be a 2-3% increase in intensity setting for a successful block for most. That’s pretty much nominal for folks other than those in rapid states of fitness increase - be it new to structured training or coming off a break.

Guessing a flat ftp will be a big one followed by pointing out their PL levels increased so there was improvement followed by “so why doesn’t in predict PL levels” followed by a few months pass followed by “It predicts PL now”.

The moderators who keep closing all the different threads poping up to complain about not realizing there is already a big thread about it :-p

:rofl:

1 Like

Those are the ones who are more important to predict. If your FTP changes by only a few percent your workouts won’t really change by much but if they change by a large amount it can have a great influence on how you train. Plus a prediction of a large growth can help people who are new better apreciate TR so become long term subscribers and help them figure out the best training load for improvement. (ie. compare the predicted FTP gains for low volume vs mid volume)

Exactly. I’d be super cautious introducing this feature. Or made it available to those who’ll tick a box under a
Statement from TR team that the predicted FTP is pure speculation. Otherwise I see a barrage of complaints heading their way…

1 Like

I’m confused by what you’re trying to get here. You do get those changes with a re-assessment every 4-6 weeks. Knowing if you’ll get that much now, which is just really a WAG has no real impact
on the training you should do.

And a 2-3% difference in intensity is a really big deal. It makes a huge difference in workouts - without AT I’d often start failing workouts coming off a change like that.

Guessing what fitness changes will be is irrelevant. What we really want is actually sustained 2-3% growth. Compound that for yourself over a year or two and you’d be extremely happy.

I do not understand, from the podcast I had an impression that FTP estimation feature is already available, but clearly it is not, because today is my ramp test day and it looks like I have to do it because there is no estimation option available.

On the other hand, I think Garmin, has already been doing FTP estimation for a few years. Before I started using TrainerRoad, I did my own intervals with Garmin and from time to time it suggested me a new FTP settings.

This is what is available right now:

1 Like

Oh, is this a different topic? “Future FTP predictions”? I do not think I heard about it even on podcast. Sorry, if I was mistaken.

Right, there are two different things, that I touched on in another topic today.

  • FTP Detection = FTP estimate at the day the tool is used ( now ).
  • FTP Prediction = FTP estimate at some date in the future ( weeks or months ahead ).

We have Detection right now, while Prediction is something still in the oven and yet to be offered.

And yes, this topic is about the future forecasting, prediction of FTP.

1 Like

I see, I thought FTP Detection is the Ramp Test. When you use AI it is “FTP Prediction”. Then the difference is called “Prediction Error”. I am sorry I am a Machine Learning Engineer this is the terminology Engineers use. :slight_smile:

You can argue about that term use with TR then :stuck_out_tongue:

I am just using the terms they have selected and how they choose to use them (per their published articles and forum info as already referenced).

There definitely appears to be a lot of sophistication behind it. Nate talked about it in the other thread discussing Analytics.

It sounds, to me, like it will be using WL V2 and will factor in overall TSS/volume (as scheduled) and will be able to utilise outdoor rides etc etc. I don’t think it’ll be cookie cutter based off of x% per block, but that’s me reading between the lines.

Either way it sounds glorious to me. I can see the issues mentioned about failing, but to me it’s hugely motivating.

I think this may actually be demoralizing. Suppose I look into my potential future FTP and realize I have no chance of making it to the podium. Do I need to even start? Maybe relaxed riding in the park will suffice to support my health… What would motivate me to still progress if I know, it will be in vain?