Thank you, Caro. I believe the new AI found me some really great workouts last few weeks making me stronger so I applied the b2b2b threshold workouts and completed them with no failure. I do love the new AI training, so much less failure than before. Just that the mechanism behind the fluctuating prediction is causing some anxiety in me.
For the weeks around the holidays, I admit those were crazy days filled with overtime, over-eating and vacations. I worried about not having enough training and tweaked the trainings and went into a loop hole of fatigue. It was also that I haven’t done any sweet spot workouts for a long time that yielded lots of the Z4 failed workouts. So glad the new AI saved me from it.
What do you think of the new AI prediction of 313 watts? That seems too high to be true IMO…how could I gain 7% in just 4 weeks.
Part of what I think is happening is we are all so used to using FTP as a measure but the FTP we see in TR now is not a direct measure - it is the setting that the system thinks will give the most effective training.
So while yes, it is unlikely that actual fitness would change that much in that time, the method to give the most effective training could.
Sure did. It was lower than previous SS workouts that I rated as “max effort,” yet the last 2 intervals were certainly out of the 85% of max HR range typically defined as Sweet Spot. But I certainly could have done two more intervals. Could I have done them at the Sweet Spot HR threshold? No, but that’s not what the survey asks. Or am I being too literal with the survey questions?
That’s weird. The workout I have been referencing I did the same thing: I changed from Hard down to Moderate to keep that predicted FTP increase. It was between a Moderate and a Hard, so I eventually changed it down to Moderate, as only the last interval felt truly challenging.
Rating Hard: +1w FTP
Rating Moderate: +3w FTP
All off that one workout. So I must be on the threshold of some unknown quantifier where the HR was less relevant than my RPE. And that’s why I want to understand what the AI is doing because that prediction is more vibes-based than data-based.
“ while yes, it is unlikely that actual fitness would change that much in that time, the method to give the most effective training could.“
Your fitness or FTP isn’t changing based on how your rate a work. But the TR metric the system uses to determine optimal future workouts does change. @Nate_Pearson and/or @Jonathan should bring this up on the next podcast. It would clear up so many of the questions/issues that are coming up on the forum. People are trying fight or game the system because they don’t understand the what Nate means when he says, “it’s just about the watts.” For example I had to swap out my 90 minute threshold workout on Thursday for a 60 minute threshold. The 90 minute had a WL of 4.2 while the alternate had a 3.0. Lots of people who say that the alternate workout selection wasn’t working but it was dropping the level. But that’s looking at the wrong thing. The 90 minute workout had 30 minutes of over/under time and then added zone 2. The 60 minute workout had 30 minutes! So it found me just the right workout because it was looking at watts! It didn’t care what the WL levels was.
If it were “just about the watts,” then it shouldn’t change by 2w based on how I rate one workout. It would look only at my HR and Power and make a decision. But it’s clearly not doing that.
That 2w increase gave me a significantly different workout than the one I had previously seen it had scheduled for me. I’m looking forward to crushing it tomorrow. Both 2 hrs of threshold but different ways of doing it.
And if it’s too hard and I’m wrong, it will or should be able to tell that it made a mistake. But I don’t think it did. And again, that difference is all because I rated the workout differently, not based on HR and power. Which is fine, btw, but it makes it hard to understand.
No idea, I do know that the AIFTP can swing 2% in response to a hard versus very hard threshold workout in my case. I would never quote my wkg based upon a value that can swing that much based upon rpe in one workout.
Just to give you another perspective that was not mentioned here before:
I also think that gaining that much is unlikely but we have to keep in mind that TR AI FTP it’s based around performing a threshold 3 workout… I think of 3×9 minutes intervals.
So if you are high glycolytic athlete (for example, the kind of athlete that get an overestimated FTP based on 95% of 20 minute test and should better use 90 to 92%) and have your strength and power duration curve at 5 to 15 minutes … the gains here could be a bit faster and a bit higher and that could explain or drive the TR AI FTP value because it would see that you can perform very good at those durations and intervals.
And then the confusion comes because we mix up the TR FTP with 60 minute prediction. But the good thing is that TR prescribes based on your power curve.
The only problem with the TR AI FTP value could be that if you are a super glycolytic athlete and really, really strong in those 10 minutes intervals, you may drive the AI FTP value so high that there is a problem with the threshold floor (95%) or sweet spot floor (88%) when TR prescribes you threshold workout or sweet spot workout with longer intervals. But I’m not sure about that.
That’s an Interesting post. It reminded me of one I saw the other day that was expressing concern that there didn’t seem to be a clear framework whereby TR would seek to build out your TTE at or around FTP over the course or a block. Which is a legitimate concern, of course.
If you’re already well trained, making FTP 10w higher might take a whole season, but increasing your TTE from 30 minutes to 60 minutes might only take 8 weeks, and will make a lot more difference on race day.
You know it’s not always clear cut. I do rate workouts on the same scale. But sometimes, it’s a toss up. If the workout had been truly hard, I would have rated it so. I have had zero problem rating workouts VH and ME even in this training block. In fact, I am trying to figure out why it gives me SO MANY of those. But this one was not like that.
I think the system is telling you that you need more rest based on the shorter time in zone. It may or may not be right, but I’d bet that is why it doing this. The system is trained not on a prescriptive pattern (add more time in zone or increase wards) but trained to give the workouts that based on the millions of TR rides analyzed that it believes will get you fitter and faster (notice that avoided saying raise you FTP ). So that the logic. You can choose to ignore and do your old structure. But I think that explains it.
I do think many are not good with knowing the outcome of their training.
We are also assuming the model is prescribing what is best for us. From my perspective any 1.0 iteration is going to still require improvements. There is logic to what the model is advising. We just have no idea if it is the best. TR changed our intensity for recovery weeks from before. Is this good or bad. We will see but I can say recovery weeks previously were very light on effort. I did think they were too long before but have no evidence other then how I felt.
). So that the logic. You can choose to ignore and do your old structure. But I think that explains it.