Nate Pearson says that the new Trainer Road AI progressively trains you without regard for your FTP (AACCP # 573 at 17:30). So when this month’s new FTP became available for me, I passed on it. I see all kinds of posts on the forum worrying about why their FTP moved x or y amount after x or y workout. Who cares? If you’re using the AI to train, FTP is now an outdated and useless metric. Just do the workout the AI prescribes and watch yourself get faster. You’ll have a rough sense of what your FTP is when you get something like over/unders anyway, and I bet that workout will be just about as hard as you can tolerate, which means the training is working. Right?
Edit: a few people have commented that the AI still requires an updated FTP because the workout library selection is contingent upon your FTP. So, I guess I must admit that FTP still matters, and I’ll be asking TR to update my FTP.
Partly agree. But would still accept the change on the basis that keeps your FTP in a spot where the AI has a good selection of workouts for you (AI may care about the watts it’s prescribing not what your FTP is, but the workout library still runs off % of FTP).
And then depending on the context I might still care about the trajectory of my AI FTP, at least in an apples to apples context. I can’t really compare it to previous years. But if I’m following a structured, progressive training approach and I’m nailing my key sessions, getting enough rest, not skipping workouts, etc but my AI FTP is dropping from one block to the next then yes that’s going to bother me and I’ll be trying to figure out why.
I agree that it’s unnecessary to peg workouts to FTP, with or without AI; but if you want to use the PMC, quadrant analysis, IF for pacing, track fractional utilization, model substrate usage, etc., then you need a somewhat accurate FTP estimate.
No, because TR AI needs an FTP to access its huge workout database (which still is based on FTP). I expect that this will change in the future with dynamically generated workouts, but right now an FTP is necessary.
I am following a training plan but don’t see the value in knowing FTP month-to-month given that the model does not know, and doesn’t not need to know, your FTP to work.
It very much does still need to know your FTP. Since all of the workouts in the TR library are based off of a % of FTP having a wrong FTP limits the workouts the model can use and if you have a wildly off FTP there will be no workouts that meet the requirements.
How can it both be the case that the model does not need to know your FTP or even use your FTP AND that it needs to know your FTP? I’m having a difficult time understanding because your position seems 100% contrary to what the TR team is saying. Maybe your position is that the TR folks were just being imprecise, and that to get the right workout the model needs to know your FTP.
I care in the sense that I use FTP as a measure of progress. I understand it’s only one part of a big equation, but it’s the most visible metric. During a race, there’s a lot of other variables that come into place (in a group or solo, temperature, terrain, etc.), but knowing how much I have changed from last race may be helpful in that planning.
My biggest gripe with this change is that I have now lost a consistent measure of my progress from ‘pre-update.’ I wish I knew what my FTP was now, based on the old formula, even just for comparisons sake.
Either way, it’s not the end of the world for me and it should push me to do more 20 minute power tests and use those as a more consistent metric. Granted, all is good and well until the formula changes again, and then I’ve again lost that tracking mechanism..
It’s also a bit weird that with everything AI can apparently do with our training, that it can’t spit out a reasonably accurate “FTP” measure, based on however TR wants to define it.. I’ve always considered FTP as being a theoretical one hour power in conditions comparative to what your power is being evaluated from. I.e. if you do 20 rides at 50 degrees on your trainer in your garage, that would be the environment. I understand there’s a million interpretations, and this thread doesn’t need to go into all of those.
At this point, I don’t think I’ve seen TR’s version of what they are calling “FTP” defined. I know they’ve said it’s not FTP.. so what does it reflect?
The model assigns workouts by what it thinks you need (so if it thinks you need to do 3x10m at 200w it will find a workout like that and assign it to you).
But since all of the workouts in the TR library are based off of FTP you can run into situations where that workout does not exist.
For instance if your ftp is set to 200w there are definitely workouts that hit those numbers and will be perfect for what the ai wants you to do.
But if you had a 300w FTP set there might be zero workouts that hit the 3x10m at 200w that the ai wants.
Thank you! This makes sense. It would be a dream for my FTP to exceed the library’s current selection in the way you describe, but I don’t see that happening. Still, I’ll accept the new FTP…
Yes, the model doesn’t, but the entire workout library does. The way I understand it works is that TR AI plugs in your FTP into the relevant subset of the workout library, say sweet spot. TR AI judges these workouts based on absolute power and selects the one that fits best. It is this “scoring & selection” algorithm that doesn’t know about FTP.
