I have been training with TR and have been very happy with it. I’m not sure how long, but I think at least 4 years. I am not confident in the quality of the new AI FTP model/predictions.
After the AI FTP prediction was rolled out, it increased my FTP by a few watts, and then predicted I would gain 27 watts by the next prediction. It steadily decreased until the AI FTP calculation, and I got a -1 watt adjustment. The next day, it was predicted that I would get a 37-watt increase! It is going down the same path, decreasing my prediction. Today, in particular, I did a VO2 workout, completed it, and rated it “Easy”, leading to a decrease in the prediction. I feel like I have completed almost all the workouts successfully and did not rate them as overly hard. Unless my perception of how well I completed my workouts is way off, which it could be, this software performance would get me fired from my job. As my work schedule dictates, I sometimes have to move workouts around, but I generally get them in.
My account is public, so feel free to correct my perception of workout quality or show me where my training is so far from the normal realm that the AI model can’t predict my FTP very well. I am happy to be roasted on the next Podcast if it helps me and others learn how to train better.
It doesn’t look like you completed any Threshold workouts greater than a 2.6 in the last month. I think that is why your FTP isn’t increasing as fast as predicted. It looks like you skipped a couple of them.
The problem with consequently rating workouts Easy or Medium is that the Ai starts expecting you to do that. Sure, you rated your last workout Easy, but you also rated the previous VO2Max workout Easy, and the next one is a 7.3 WL workout predicted to be Moderate. For most of us, rating a VO2Max workout as Easy would probably give a big jump in the FTP prediction, but for you it’s as expected.
The other thing that sticks out to me is that your prediction is based on a very agressive ramp rate in workout level, but in the last weeks you have barely had any increase in WL at all. It makes sense that the prediction would trend to 0 if you don’t have progression in your workouts. The question is then why that happens if you find the workouts managable. What happens if you change your upcoming threshold workout to a WL 5 workout?
I understand that rating workouts as moderate and easy trains the model to expect that. If the workouts are completed, should they not progress to higher levels? This is also to your next point. What is the point of AI if I have to progress the workouts myself?
Yes, I did miss a couple workouts but that would not seem to account for a 27 → -1 FTP prediction change. If my workouts are rated Easy to Moderate why is the AI not increasing the workout level?
I’m just speculating here, but I suspect the Ai thinks you’re doing too much. Doing more than scheduled and training hard on yellow days is not following the Ai. So it makes sense to me that you will have to override it on other parts as well to get the “benefits” of the other changes you are doing.
Again, what happens if you change your next threshold workout to a harder one?
The main reason your prediction is on the high side is because you’ve set your training approach to Demanding.
So, in the simulation window you’re massively ramping up the difficulty of your Threshold workouts, but this is unrealistic given where you are, where you’ve been, and the fact that you recently skipped a threshold workout in a week when you only rode for less than 3 hours.
It doesn’t look like you are completing the workouts to get full marks from what I can tell. It appears you are able to do the sweet spot but not the threshold work. Maybe you just need to get used to your current FTP.
Can you give a bit of clarity on your comment? I have missed a couple of VO2 workouts, but I think the ones I have completed were done successfully. The last workout was VO2 4.9, and I completed it successfully; it was not very challenging. Reality check me on my assessment that I completed successfully. In the last few intervals, I did have a rest point due to stop signs. Will that have that much of an impact on being deemed successful? I do appreciate the feedback, and I am not obsessed with the FTP number, but I may not be progressing as much as I can. I did change to balanced based on the comments.
The Threshold workouts. They don’t look to be completed to the expected level. It is hard to tell though on someone else’s calendar. They are going to drive your ftp progression. If you complete your assigned workouts though I think your FTP will match the predictions, they look pretty challenging though.
I’ll be honest, I am a bit surprised looking at your suggested upcoming workouts that they’re all rated moderate and you’re in build phase. I’m not seeing any of the usual things we see here in this scenario (recent break, rapid TSS ramp etc.) that would cause this. Workouts look to be executed OK. There are a few missed workouts but it doesn’t quite add up to me. It would be good for one of the TR team to chip in here if they have time @eddie@Caro.Gomez-Villafane
Having had another look, I wonder if I’ve spotted the issue. @timdave13 - should you have power data for your workouts on the 18th and 27th of Feb (others too actually)? All your other outside rides seem to have it, but (from what I can see) these don’t which means there’s no KJ average power, power trace etc. and I wonder if this is throwing the model
Is it really? I would say it’s this weeks threshold workout that is wrong. He is finding the level 3 workouts moderate and haven’t strugled with anything due to intensity. In an ideal world the Ai should ramp him up to sufficiently difficult workouts. Also, the workout on March 21 is back down to WL 3.2 because he’s predicted a 13.3% FTP increase.
Can only see TSS, IF, and the line chart. And like a threshold 3.0. But can’t always see the workout name for workouts in the past. So can’t easily tell what was supposed to be the WL.
It also could be the pain meds as I recover from surgery. Bored out of my mind today.
Depends on what platform you’re on. You can see more on a laptop/browser, but you have to keep copying the username into the URL because every time you change page it bumps you back to your own profile.
I spotted the missing data I mentioned above because the TSS summary for the weeks wasn’t adding up and then I took a look in the activities list
If I look at the period between 21-Jan and 18-Feb. My training seems reasonably consistent with pretty good completion, and net FTP decrease. Again, not focused on the number but the direction. It seems like there would be at least some improvement.
I think you’ve missed my point. From what I can see there’s no power trace on those workouts. I can see it on most of the rest of them. This might just be an issue of what an outside user can see, but the rest of your workouts look fine.
I’m wondering if there’s an issue with the imported data that’s throwing the model, not your execution. Or did you not record them with power?
Can you post a screenshot of your workout from 18/02? What I see has no HR or power data. There’s a few workouts like this. Some have TSS, some don’t
