is this not the pure definition of slow steady progress? You also seem to have ignored what i put in bold, which is why I posted this. This thread was not asking for training critique per-say, it was a critique of how the AI responds me as an athlete. I went through my training history before posting this, and it’s fully public so anyone can see what I am capable of doing, and that I am fully aware of what an aerobic base training program looks like if you look at my race seasons.
I even stated in the my original post I’ve seen pretty good improvement in my power, but that that’s with me not following the AI’s power targets. I feel like the last couple posts are pretty disingenuous, as i never stated I had plateaued, what I said was the TR AI had plateaued, and if I followed it I certainly would not be where I am now, because I’m doing more work than it suggests I do.
Yes I can certainly see that since last Nov you have made steady improvement but week on week your consistency is not optimal. You are giving back a lot of fitness by being as inconsistent day to day as you are and I suspect the AI is trying to figure out how to get you started again each time you take 3 or 4 days off the bike which you have done a couple of times in the last months. Missing a few days here or there, no big deal, but your training pattern is…well there’s no pattern really. I would expect you to have a tough time building any adaptive fitness that way and the AI is going to struggle to put you where you need to be if it keeps getting mixed messages and can’t build upon day to day and week to week consistency. That’s all I’m saying.
There are very few times where i am off the bike that long. It’s like the z2 weeks are the only things you guys are focusing on at the expense of everything else on the bike. The times I do take more time off are after major mtb training rides (2h+ long) where it’s important I recover so i can hit the numbers necessary. It’s also completely ignoring that I am climbing 2-4h per week. Different type of work than cycling to be true, and isn’t in the AI, but as far as MY fitness and fatigue it’s obviously very important.
Week to week consistency does not impact AIFTP’s calculations. If it had picked a different number that should in theory have changed my workout levels. I could agree it doesn’t know exactly what workout level to pick right now… but I am completing the workouts it gives me at or above target regularly, while rating the difficulty about where it should expect the difficulty to be.
Another thing I noted this afternoon thinking about it…I adjusted my 4iiii power meter up 5% at the end of April because the only guess I had for the AI issues was maybe the delta between the PMs was causing issues and confusing it. In my opinion its clear the AI has not adjusted to this, which should be at least a 10w change from the low end boundary if we assume it has a confidence interval for what power/FTP the athlete has.
I think you have to take it back to the basics with training, like others has said consistency is key, that includes the crappy hot endurance rides in de-load week. Good rest and fuel. It all counts.
Also that damn outdoor workout bug can be having an effect with you AIFTP but we have to wait on TR for that one.
As far as the 4iiii powermeter, is there anyway that you can test it against another powermeter or trainer to see how much the difference is?
I ran into something what your going thought with some Stages PM back in the day.
I do not need training advice. Nor am I stagnant. My objective threshold powers are up in the dozens, possibly 20-30W, over the last several months. I would have dreamed of doing 300 W repeats back in march. I am not having any problems with my growth in fitness. I started very out of shape, i am now in okay shape, probably around where my default fitness tends to land. (~280 w ftp).
This thread is about how the AI Training program TR released this year is not reflecting any of that improvement, in any way, nor do I believe it is tracking that training in a way that would allow that improvement to continue. I am ignoring the ML models power recommendations, and that’s the only reason I believe i am improving, with the model is not reflecting that either 1. AI FTP should probably be higher, or 2. my WL’s need to increas. No matter what input i put into the system, in terms of power, I still get the same workouts. This is with OBJECTIVE measures of power improvement over weeks and weeks of training and improvement.
Please stop offering me training advice. I do not want it, nor do I need it. It is not what I made this thread about, which is why it is titled “Okay but also not”. Because i HAVE improved, just not really due to TR, or only partially due to TR.
I am aware of a bug that the AI starts high and ends up converging towards the end of a 28 day cycle on a much lower number. I am however not sure if that means the AIFTP is wrong at the end, nor that it can’t land on a different number. It just isn’t the massive 5-10% gains it seems to be feeding us outdoor riders a lot of the time.
How about doing a ramp test to make sure that your AI FTP is in the right ballpark?
I just did a ramp test test the day before my AI FTP prediction date. I pretty much hit the AI FTP number (-2W: 6-7 more seconds of mental toughness and I would have matched it to the Watt). The AI algorithm immediately snapped to the new number and started a new 28 day prediction cycle. The surprising thing was that the -5% recalibration effect from January just disappeared … apparently my AI FTP is now aligned to my ramp test FTP (twice in a row now).
I could, but the old AI FTP was actually spot on perfect for me. I am in the category where the ‘new AIFTP’ started me lower. So for example when i started this year my FTP was 240-250, and the new aiftp put me at 220 (lol) which was giving me WL’s of 9 and 10 to get even a basic workout in. Because my VO2 power is so shit, and I also tend to respond very well to long steady work, I would say on first impressions having a lower ‘new ftp’ makes sense, because I just can’t hit super high %s of my FTP, like at my max fitness with an FTP of ~300, my 5 min PR is i think 350 W, which was an absolute all out effort. I’m not sure if it was a bug or not, but i reached out to support and we put it back at 240-250 where it’s basically stayed.
And it would be very nice to have that old AI FTP model to use, because I calibrated against it a couple of times in 2022/23 when it released and at worst it was off by a watt or two (e.g. it wasn’t off).
This doesn’t seem at all helpful - a sign that your AI FTP under the new regime was way under.
Maybe you need to spend some time on your VO2 rather than zone 2/3? You need to train your weaknesses as well as your strengths.
This is very strange. You are saying that you can do 350W for 5 minutes and that is all-out, at your absolute max and you couldn’t have gone any harder at all, yet you can continue for another 15 minutes at that power without issue? I’ve never seen a power curve like that.
100% agree. Yet now my AI FTP is matching my ramp test FTP, and I can’t explain why. Did I change or did the prediction model change? Does training at a low-balled AI FTP cause your FTP to converge to the AI FTP? (… in my case losing 5%?)
Still, looking at your calendar some week-to-week TSS consistency would go a long way …
You don’t appear to be following TR’s program much at all, nor have any real consistency or method of an organized program on your own. The lack of rides you have in the .55-.75 IF is astonishing. Looking at your history, I can’t see how AI would be able to accurate help much, nor should there be any sort of expectation of progress based on what you’ve been doing.
You don’t appear to be following TR’s program much at all, nor have any real consistency or method of an organized program on your own. The lack of rides you have in the .55-.75 IF is astonishing. Looking at your history, I can’t see how AI would be able to accurate help much, nor should there be any sort of expectation of progress based on what you’ve been doing.
when’s the last time you looked at a low volume plan on tr’s website? I think it’d actually pretty obvious to see exactly what plan I am following if you took even a minute to read the post I opened with, and looked through my calendar for another minute.
And, who would have thought, if your FTP is artificially low that your IF’s end up being wrong. Incredible!
It honestly doesn’t look like a plan. I’m guessing the ai is trying to fix that. When I thought my ftp was too low I ramp tested and tried to show it who was boss
Or the OP could just manually set their FTP and see what TR prescribes in terms of workouts.
If it looks sensible, do that block. See where AIFTP Detection puts them at the end. If it looks sensible, accept it as a new FTP and then see what TR prescribes for the next block. If that looks sensible, all is good. If that second block doesn’t look good, set a manual FTP again and repeat.
It seems to me that the approach of Overachieving Within Workouts To Encourage The AI To Increase AIFTP doesn’t really work.
Another thing to consider is the unstructured ad-hoc riding outside of a LV plan - nothing wrong with that, but I think you get better results if you can schedule them within your calendar to some extent. That way the AI plan is expecting, say, a 3 hour solo ride on a Sunday, and assigns appropriate workouts afterwards within the plan structure, rather than suddenly dialling everything down after you’ve accumulated a load of TSS it hadn’t planned for.
I think it may have been a mistake for TR to break away from the old AIFTP which everyone seemed happy with - even if you actually get better training with the new number.
My new AI FTP is way too high - but my actual training has been absolutely on-point even with a few injuries, a crash and general life getting in the way.
But my suspicion is that my training would have been just as good if I was using the old AIFTP number which aligned pretty well with traditional testing for me so I’m not convinced that the small increase in quality of sessions was worth upsetting people so much.
