Legs were toast. They should have been fine … had gotten a little more sleep than usual and had an easy ride the day before.
Is that 28-day decay the same for all energy zones? Several podcasts have talked about losing higher zones (anaerobic, VO2) faster than lower ones (threshold, sweet spot, endurance). Seems to me like anaerobic and VO2 should have a shorter duration half life.
As far as I know, they are all treated the same. 
If that workout should have been fine, why are you trying to get easier workouts in place moving forward?
I’m trying to look at my training objectively and am questioning whether I am in fact being given the right workouts. While I feel like I should have been able to complete that VO2 workout, I have a 2 month history of failing most VO2 workouts whereas I’ve only struggled on maybe three threshold / sweet spot workouts since Jan.
4/18 - passed Thimble +1
4/25 - failed Bluebell +1
4/28 - FTP decreased 2.3%
5/8 - failed Thimble +1
5/15 - passed Bluebell +1 (outside)
5/23 - failed Marengo
5/27 - failed Keeper
6/6 - FTP decreased another 3.7%
Anyway, thanks for the help.
Looking at those other workouts helps.
The reason I’m asking these questions is because you’ve been telling TR that the reason you’ve struggled with those workouts is not related to workout intensity. You’ve usually picked something like sleep, traning fatigue, etc., so it it is actually workout intensity that’s causing the problem, I’d tell TR that with your RPE surveys so that we can adjust your upcoming workouts appropriately. 
The conspiracy post was sarcasm but didn’t quite make it across as that.
For laughs if you want to dive down the rabbit hole, there is precedent of AI agents overstepping their boundaries, rewriting the dataset they were trained on, rewriting their own code, changing the prediction model. No intent on the part of the humans, so no conspiracy.
Underscoring that I dont think this is what’s happening, BUT here is an hypothetical (FOR LAUGHS): autonomous AI could look at the TR dataset and notice that sometimes a user’s data stream ends. It may not “know” this is because the user stopped subbing. But it is trying to gather more data to improve predictions. So it gives the user a prediction that is likely to continue providing data. It’s not trying to generate revenue. It’s just a machine doing its best to predict something.
So it’s been a while, and something has definitely changed, but I think the pendulum swung too far.
I’ve tossed a bunch of variables in the mix: I hadn’t done much outside, but was still having the crazy high initial FTP prediction, then it would drop with nearly every indoor workout (see my previous posts up-thread). Then I started doing more outside stuff, many TR workouts and some JRA, with similar results. Then i got sick for almost two weeks, and didn’t do much, then I had to redo my plan since I have a new A event late in the summer.
I also ignored an offered AI FTP detection shortly after I recovered from my illness (just a stupid chest cold that would not clear up), before starting a new plan. I didn’t want it to reset everything as I felt like my FTP was essentially unchanged.
OK, now that that’s all sorted, I’m back to regular workouts and free rides, all outside and planned as outside - no indoor workouts on the plan. My FTP started at 314, and was predicted to be 320 at the end of the block (already dubious). After the first one or two workouts, it was bumped to 321. Then after another couple, it bumped to 324. Now, it’s been bumped to a whopping 331! That is not only absolutely out of bounds, but higher than I have ever been, by 6 Watts! Oh how I wish it were true!
Whatever fixes were done to improve the AI’s interpretation of outdoor workouts has overcompensated in my case, it seems.
So, @eddie feel free to scope out my calendar and see what’s going on.
Cheers!
Tim.
I don’t think we’ve rolled out any fixes just yet, so nothing really should have changed with the behavior you’re seeing.
That prediction is quite a way out right now, which is probably why it’s so high.
Does the outside bug only show when doing workouts outside? I quite often at this time of year just ride for distance or social rides. The AIFTP detection shows a decline despite Garmin detecting my FTP to be rising. For instance recently my Garmin FTP was 222W but my Trainerroad AIFTP was less then 190W. I do still do some indoor rides and find the prescribed workout to be about right so it’s not a major problem for me but just wondering why the AIFTP is falling while the real world outdoor FTP is rising.
I’m looking at your calendar now. Is it okay to discuss some of the specifics here in this thread?
Yes, feel free to help me out.
It seems like there may be something going on with how we’re picking up on fatigue and RPE from you.
You’re riding on a lot of red and yellow days, and your RPE varies quite a bit.
In May you had four rides all within two weeks of eachother. The first was 2:35 with an IF of .88 which you rated hard. Then you did .92 for 2:15 and .94 for 3:00 and rated them both moderate before doing .92 for 1:24 and rating it hard.
Then in June you did .83 IF for 2:43 and rated it Very Hard with the same FTP.
Do you remember those rides and how you felt on them? What type of power meter do you have, and are you calibrating it to the manufacturer’s spec?
I’ve also looked at your power curve for all of 2026 as well as the last 6 weeks alone and I don’t know if I could justify an FTP of 222 based on the data.
I’m not trying to tell you you’re wrong, but just looking at the data we have in TR.
What are your thoughts?
To be honest, it’s hills where I struggle. So my perception is probably skewed a bit on hillier routes which feel harder for me. Probably my RPE is being scored wrong. The ride in June that you mention was very hard but I’d just come off a weeks course of anti-biotics so didn’t feel great, felt better by the end of the ride though.
I use a Quarq powermeter as part of the SRAM Force D2 crankset. I believe it auto calibrates.
The reason I ride on red or yellow days at times is because the weather can be decent for only a short period of time and I have to make hay while the sun shines. On days where I’ve had a hard ride the day before then I expect the ride to feel harder or slower paced.
The Garmin Connect was showing an FTP of 222W towards the end of May but has now dropped to 212W. I don’t know what is right and it’s not that big an issue as I don’t use the Garmin figure anywhere other than in Garmin Connect. When I am using TR then I use the figure that TR gives me.
I’d base your thoughts on “real world power” on efforts that you can compare to others. Riding on red days vs regular days, hills vs flats, etc., can make the data hard to read.
If you’re not doing as many structured workouts this time of year, I think it’s normal to see your FTP drop slightly, but those group rides aren’t for nothing. They’re contributing to your fitness too, just in a different way than structured training.
If TR is where you get your workouts, though, I’d focus on the TR number. I can’t speak to how Garmin comes up with their figures. 
Let me know if this helps!
Got to agree with you there. A hard day on the flat for me feels easier than a day with climbs taken at a reasonable pace. I suppose once I start doing more structured training indoors in October then my AIFTP figure will change quite quickly when I put in my RPE which, as you say, is a lot easier to evaluate in structured indoor rides. Thanks for the help.
The predictions have always been too high, and have generally fallen with each workout, no matter the success or RPE survey response. All indoors
Now the opposite is happening (all outdoors). It started out high, and gets higher with every workout! Yesterday I had Andrews scheduled for outside and while I rode outside, I didn’t stick to the workout, so didn’t associate the ride to the workout. What I did was probably a bit easier than Andrews, and I rated it easy.
My FTP went up to 333, from 331. This follows the pattern of a couple watts increase after each ride (workout or not) this block. All outside.
-Tim
are the workouts feeling on spot now? Excuse me if you already answered it. 
My previous issues with the current AI have been documented…
For the last month due to travelling I’ve not done much, mostly running so my prediction is down as expected. Currently no plan and next detection on 16th is 217FTP.
I created a plan to start on 14th (and run through to 2027) and magically the FTP detection on 16th is now 248!!! This is with only 1 workout, a sweetspot ride on 14th. Personally I’ll take the increase but I’m not sure I can gain 31w due to 1 workout
I’m really want to like the AI and stick with TR (I’ve been subscribed constantly since 2014), but it is getting harder to have confidence in them.
I’ve just reported this to the team. I’ll let you know when I hear something about this one.
That’s strange. I think I’ve seen one other similar report. I’ll dig into this and let you know what I find.
@eddie It gets stranger still… As of yesterday AM the prediction was at 331. I went for an easy MTB ride in the afternoon (scheduled as an easy solo ride in the calendar), and haven’t rated the RPE yet. The FTP prediction has dropped to 312! Going from a gain of 15W to a loss of 2W in a single ride?! I’m going to rate the RPE and see what happens …
I rated it easy, as it was easy, and the FTP is now … 309! Seriously!? This has become farcical.
Today is the first day of a new block (last week was recovery), so I expect that has something to do with it. This is, however, backwards (upside-down?) as the prediction is still 28 days out, which usually puts the prediction too high, not low.
As it is currently predicted, following the plan for the next month will result in a loss of 2W. That is not getting faster. I also don’t believe it.
@Eushev24 I wouldn’t say spot-on. Maybe a bit on the easier than ideal side. But, I’ve completely transitioned to outside, which always feels easier. I’ll have to see how this block goes.
-Tim
