Completed every workout apart from 1 over under session I stopped pedalling for about 10 seconds before the last over.
My first FTP prediction started as going from 288-302 and ended up at 294 with only 1 workout being marked as all our effort and it was the final workout before the easy week, everything else was hard that was supposed to be hard and easy that was supposed to be easy.
I’ve increased my volume of my plan from 2 hard rides a week to 3 as was recommended, I’ve also increased volume from around 5 hours to around 7.
I am completing all workouts and achieving the targets set and even sometimes going over but my prediction and ai detection for FTP keep going down as the plan progresses, this prediction is for a 2w increase, what am I doing wrong?
Nothing I’m too concerned about there other than there are quite a few yellow days…. they’re landing on endurance days but it’s not really ideal to have that many… i think you may be carrying a bit too much accumulated fatigue.
I can see that you are not great always respecting your “off” days - that probably the only thing to improve.
That said you’re actually seeing quite a bit of improvement in fitness considering that your seem to have been training with the same volume for at least a year and a half - not surprising if you are close to a plateau.
3% improvement in FTP in a couple of months is not to be sniffed at - well done
The thing that I don’t understand is that both FTP predictions at the beginning of each block was over 300, but both times they have reduced even though I’m hitting all the power targets and completing each ride.
Going from 288 at the start which was an AI detection anyway to 294 I was still happy but slightly disappointed that I didn’t get to over 300 which it was predicting for the first 2 or 3 weeks. But now it’s predicating I’m going from 294 to 296 after the initial prediction being over 300 again.
I just thought I must have been doing something wrong that I can’t see, maybe the AI does just think I’m over training?
Getting a FTP-increase and then immediately rating your next hard workout maximum effort is a big red flag for the Ai. It has resulted in you having a very conservative Threshold progression which is probably the main reason that your FTP-prediction has dropped. If you feel that your training is going well and that you’re handling your Threshold workouts better now, there isn’t anything wrong with changing to a bit harder workout for your upcoming Threshold workout.
You are doing more volume than you were at this time last year and likely at higher power numbers than you previously were. I’d be super pumped about that. Close to 300 watt FTP is no joke.
As others have said FTP gains don’t come quickly and usually the gain you see is from the 4 weeks prior, so the work you are doing now won’t show up in your fitness for another month. Incremental gains are still gains and they compound over time.
My FTP is around 284 and my prediction for 2nd month of build at the start was 305 or 309. I’m ending up at 286 because I had to take a week off for work. Still feel stronger and know I can add more to my fitness. But I also get that I missed the gains I could have had due to no fault but my own, well really we can blame work.
It’s a lot harder to get gains when you are fit vs when you are first starting out and that’s something we all need to remember.
You can always check your volume on your plan and see if TR thinks you can handle more, that may help with your fitness as well.
The AI will throttle your training based on fatigue separately to the yellow/red days.
Think of the yellow and red days as a system that aims to stop obvious overtraining, the AI is more subtle and will fine tune based on your current level of fatigue.
So you may have ended up being prescribed slightly easier workouts than you would have done if you had managed your fatigue better… maybe… it’s also true that predicting the future is hard and that the predicted FTP has just become more accurate as the date approached?
Like i said before, you have stuck the the plan pretty well other than maybe over-doing it, if we assume that TR is perfect then the only thing I could blame you for based on that training history is not respecting your off days; and you are maybe also guilty of not making those easy dynamic endurance rides truly easy based on the observation that you have yellow days after them sometimes?
As mentioned above. It seems like you are not respecting your 1 complete rest day on several weeks. I would make sure it do that. Otherwise, slow consistent growth is what good durable fitness is made of.
Could be that the initial FTP bump to 294w was a bit keen…. so the fact that the AI now thinks you have turned it around and is thinking of bumping you up further to 296w actually demonstrates more linear progress over the course of 2 months than it appears at first glance.
So you still might get your ~3% increase over 2 months… but in hindsight it might have been better to give you two consecutive ~1.5% increases rather than a ~2% followed by a 1%?
Either way the system looks like it’s locked you in and has been giving you workouts at the correct difficulty throughout - other than that initial threshold workout that Torneng mentioned.
Totally agree with that. In the end, I use the AI FTP Predictor as a metric but do not obsess over it and just do the workout and see how I feel over time. What matters to me is feeling stronger over time on hard workout as the duration in high zone progress or when more intervals are added. The FTP number by itself is a plus.
I’ve been wondering if showing the AI FTP predicted number is worth it at the end. It makes people complains about the fluctuation of this number as the days pass. I kind preferred to have the surprise every 28 days and compare it with how strong how I feel on the bike compared to before.
I imagine the max effort early on lowered your threshold progression. Did it appear to slow from what it was previously. It’s hard to see the levels in the one month at a time screen shots.
Have you played with the sliders, maybe ramp down endurance and slide up threshold some.
100% agree with this. I’ve had the feeling recently that the posts have moved from “what the heck happened to my FTP # with this new AI FTP business”, to “why is my predicted AI FTP # movlng all over the place?”
Perhaps AI FTP prediction could run in the background but the user would only see it 10-14 days out from an AI FTP detection? At least then the AI FTP prediction might be converging on a value that might more closely match the AI FTP detection and there’ll likely be less variability in the estimate?
For me personally I’ve been taking the AI FTP prediction number immediately after I run a detection, turning it off, and then checking ~7 days out from the new detection. Rest of the time I just get the workouts done and regardless of my AI FTP # I’m feeling a lot stronger now than 2 months ago.
It did seem like the final hard workout before the rest week of the first block is where the predicted gains slowed down a lot, I still completed the effort but I marked it as all our because it did feel like it was an all out effort and I wanted to be truthful on my RPE.
It seems like from that ride onwards I’ve done the same vo2 workout and the same over under workout every week, before that my workouts seemed to be more varied and progressive.
Everything I’ve seen about yellow days is that they’re fine to train on, just avoid intensity which is what OP is doing here. If they were red, absolutely.
Absolutely fine to train on a yellow day - but its not normal to see that many and the quantity indicates high fatigue - its likely that the AI will also be nerfing their workouts if fatigue is high….
RLGL does not affect AI workout picking - it just triggers the switch to endurance or rest days - it’s a blunt instrument.
AI is more nuanced and will make small adjustments on the fly based on performance, fatigue and other factors.
They are separate like TR says.
However, you can use the visual output of RLGL to make assumptions on what the AI is doing. i.e. if RLGL is showing a load of fatigue then it’s likely that the AI is seeing fatigue too.
I’m not implying that there is causality… just that you can use RLGL as an indication.
I wouldn’t be surprised if they roll up the “blunt instrument” fatigue protection into the AI system in the future but currently the AI doesn’t really work on its own.
If you turn off fatigue management the AI will currently continue to try and prescribe you the best workout it can.
For example, if you had a week where you just placed 7x AI threshold workouts in a row, the AI will give you workouts you can do.
What this will look like is an unproductive week where Mondays workout will be at your current level but then the workouts will just get easier and easier throughout the week until the AI cant find any workouts you can do without failing.
The none AI powered fatigue management is still required to stop this situation happening.
TL;DR the AI looks at fatigue when prescribing workouts - but wont necessarily protect you from it in the most productive way.
