Firstly big thanks for opening up your training diary. Clearly you are a very strong cyclist and I can understand your bemusement at struggling with training perhaps for the first time ever.
It certainly looks like your ftp has been set on the high side. Interestingly it looks like you struggled with Sundays workout from quite early in the session. Was this a workout you were expecting to be moderate?
I have been a long time Trainerroad user and historically fail about one session a year. I have failed an aiftp session already this year because of intensity. In my case it was a 3x30 minute sweetspot session I was expecting to be at a moderate intensity. On reflection I didnāt prepare properly for the session and didnāt fuel properly either partly because I had already completed a hard threshold session that week so moderate should be no problem
My memory of failing workouts in the past resulted in a big step back in the set workouts and a gradual rebuild. I have been impressed at how quickly this version of Trainerroad got me back on track and Iām confident that within a fortnight you will be smashing your workouts out of the park again.
What I have also found useful with the latest rollout is the graph within the workout which compares the training session work against your power curve for the last 6 weeks. This should quickly highlight if the workout is realistic before you attempt it.
Good luck with your training - I bet this version of Trainerroad will soon have you at 5w/kg
I think you might be talking past each other here re: TTE at FTP. You might have missed the fact that the latest TR release redefined FTP. Now your TR AI FTP is just the power for which youād be predicted to find a TR Level 3 Threshold workout āhardā. For me, this doesnāt match my threshold at all. (Threshold being that inflection point on your power duration curve where 5% above feels markedly more painful and 5% below you can feel your legs recovering a littleā¦ eg what youād get from a Kolie Moore FTP test.)
I just tested my threshold and got 175W with 50 minute TTE. Immediately after the test, TR AI FTP detection gave me a 197W āFTPā. Because I havenāt been doing much work at this power lately, Iām not convinced I could hold 197W for even 10 minutes.
I think this means the new AI FTP would push athletes like us (who end up with inflated FTPs with the new system vs old) towards workouts with more intensity, which might work for some people! It wouldnāt work for me right now and it sounds like it isnāt working for you.
I think TR just had your training benchmark set too high and is now dialing back based on your workouts and feedback. Not much to criticize on your end ā¦ maybe you should have dialed down your workouts even more to an appropriate level to safe yourself from failing or going all out until AI FTP catches up on you. But at least this seems to happen now.
Iāve had some failed workouts too and have been on TR since the 89$/yr days and feel itās been ātoo hardā. Especially since itās supposed to better about that. But Iāve thrown in extra stuffā¦.maybe thatās part (or all) of the problem. Itās about 4 weeks in so a new aiftp number will be out soon, my plan is to follow it exactly and see what happens. I feel like the extra stuff has helped in the past but maybe itās a terrible idea with the new system? Hell, maybe it was a bad idea with the old system too!
I had a similar experience. AI seemed to prioritize pushing intensity and did so bumping my AIFTP (the overnight change). The solution for me was to use manual FTP. Iām mostly happy with the workout selection afterwards, although Iāve been overruling it fairly often the past week.
Edit: @waf89 Iāve failed workouts in my first and third weeks of January (intensity) and have some notes in my calendar tracking the FTP changes. I think Iām back on track now. If youāre curious, take a look:
Iāve taken a step back with basketball (1 session less per week) and lowered my FTP. Also, Iāve recognized that I donāt handle intensity well, so I spend time carefully picking workouts with less RBI and more TiZ compared to what the AI suggests, which is usually more intensity. Interestingly, the model thinks that I have less chance of failure for the workouts it suggests, while Iāve already failed a few. In contrast, I just finished a reported 5% failure chance workout (very ānot recommendedā) yesterday in suboptimal condition (underfuelled after a long day at work), so Iām now starting to think that it just doesnāt model me, personally, well enough, and Iām even more inclined to self-select my FTP and workouts.
It does seem to learn, but very slowly. My self-selected FTP is 353, which still feels relevant based on my recent sweetspot and threshold workouts. AI thought I was at 366 at first. I was due a new detection today, which came at 365. So either itās convinced in my ability and Iām detraining in its eyes (while hitting an all-time 1hr PR yesterday), or it is learning and is inching closer to my realistic FTP (projected to converge to it inā¦ 12 months? )
Seems to me the single failure of the AI here is in that first FTP estimation. I assume shoddy code that (somehow!) failed to take into account how little youād done recently, and how much of a negative effect that would have had.
If your initial estimate had been 20W lower as maybe it should have been, maybe all would have been well.
Iāll be interested to see what a TR admin has to say once theyāve reviewed this case.
There only seems to be a small percentage of people like us in this category. My only solution is to lower your FTP to get the intensity and tiz you are looking for. Iāve got level 5 threshold and level 7-8 SS back in my calendar and they are at the correct effort and tiz for this time of the year IMO.
I have one observation here. You mention you you āhave lifted weights on a Monday and smashed out VO2 on a Tuesday for almost a decade with steady gains.ā
This new plan prescribes you longish SS or TH workouts for Tuesday according to your calendar which may be the problem here.
From my own experience I have no problem doing VO2 work the next day after my heavy squat day, I can tolarate basically anything for 2-5 mins, but longer intervals can be really hard. On the other hand TH workouts after the deadlift focused day are okay.
I am not sure how you structure your lifting days, but maybe it worths a considaration to switch up the excercise order during your lifting workouts.
Itās not independent of FTP, but itās not FTP first either.
We look at wattage first, but your FTP can dictate what level of workout weād recommend.
If your FTP is set too high, youāll be stuck with really low-level workouts, albeit with similar wattage, whereas a low FTP will give you higher-level workouts with more time in zone, but again with similar wattage.
I think a lot of people missed the point here. I think the question is why did TR predict he would get to 357 so aggressively and if that is because he has a habit of extending workouts with z2, then that is something TR should look at as it seems to be influencing the AI to think he was capable of a much harder workout when the Z2 wasnāt hard work. I do this as well. I ride for an hour, and then will often extend that to 1:20 to cool down, chill out, etc. Iāve look at the progression ratings for the workouts Iāve completed and noticed that I got more than 100% credit for not just the endurance part of the workout, but also the other zones. Here is an example of a workout I did recently, Blackcap. I completed the workout, increased the intensity of the cooldown and then extended the cooldown a bit. It gave me credit for a 3.9 threshold workout even though it was designed as a 3.7. Said it was a 4.1 Tempo, even though it was just a 3.5. I did no additional threshold or tempo work, but I can see how the system seems to think I did.
My absolutely uniformed guess is that normalized power is being used to calculate athlete levels, rather than power. So even though you didnāt spend any more time in the threshold power zone, you did spend more time in the threshold normalized power zone.
From what Iāve read, athlete levels arenāt used by the AI to choose workouts (and therefore to detect FTP). But I would guess that normalized power is used and maybe that can lead to some anomalies (although I doubt itās significant enough to be the main cause of the massive discrepancy we saw here).
I think this is a very fair point and sequencing around strength vs SS is something Iām considering revisiting because, as you say, Iāve historically knocked out shorter VO2 efforts the following day which (like you) I find pretty tolerable
Thanks for this. I had a decent workout today after TR revised down my FTP 11w today.
I nonetheless agree there are a couple of points from my case that remain pretty unclear:
(1) how could TR have thought the appropriate January reward for a 2 week holiday was a 9w FTP bump
(2) having trained for many seasons, I think 329w was v unrealistic following my preceding 3 months of training. This does make me question the FTP calculations being used.
(3) the originally predicted ramp rate to what - for me - would have been an almost all time career peak of 357w after just 11 structured workouts / a single block of training / by mid-Feb was really quite wild.
That said - the FTP drop seems to have recalibrated me quite well. I am in agreement with basically everything Kathleen has said about this.
Like you, I also donāt understand what is going on with progression levels - exactly as you describe, I extended my workout to get additional z2 training in today. Despite having executed the intervals exactly as prescribed, not exceeding the wattage, the result of my added z2 was that I was given a level 6.7 SS for what was a 5.2 SS workout. Bit weird. Your thought is a good one - is this whatās warping it?? @eddie - could I get your thoughts on whatās going on with the levels here?
Which makes sense. If youāre not falling off the bike at the end of the final interval itās a reasonable assumption that youāre a more capable athlete.
Are you entering your strength training sets accurately? I see strength days (and many on red days at that) where youāre completing 20ish working sets and then rolling into a hard interval workout the next day. If thatās the case, then Iām not surprised youāre failing.
Thanks - Iām grateful for the comment but please do consider the full thread. The change correlating with failures is an overnight one corresponding perfectly and immediately with the introduction of TR AI. I have strength trained the day before hard workouts with steady progress, good gains, successful workouts and essentially no session failures for 10 years. So failures after those sessions do surprise me. I havenāt come on here to say āhey, here is how I train, I fail my workouts, please offer generalized observations on my programming and sequencing of workoutsā. The question is squarely one about what has changed to make my workouts so difficult from the instant that TR AI was introduced. I think the latter half of this thread has identified some key questions. The main failing seems to have been (1) the hugely optimistic January 2026 FTP increase, which raises questions about the underlying TR calculations because it couldnāt be supported by my recent training history / preceding holiday and (2) the curiously aggressive and unrealistic ramp rate prescribed. Iām v grateful to the TR team for digging into the curiosities of the Progression Levels which seem to be skewed in quite radical and unexpected ways by extended cool downs under the new software.
Your comment about working sets is a fair one and I think the TR definition of a working set is quite unusual. I count all sets that are hard working sets and excluding warm up sets. My sessions are typically full body with a barbell. Just a short rotation through legs, chest and back and an ancillary or two (4x8 or 5x10 or similar) and itās very easy to clock 20 working sets (eg 2 leg stations, 1 chest, 1 back, 1 ancillary and some core - 4 working sets per station).
Fair but still so many factors. Age, sleep, etc. Thereās a point as we age where what used to work simply doesnāt anymore (at least not in the same way). I canāt imagine doing 20 working barbell sets on a red day and having a great workout outcome on the following day (if the sets are true working sets and it sounds as though they are). Have you considered moving the lifts to the same day as your rides? Nate, John, Alex, etc. have all talked on recent podcasts about keeping the hard days hard and the easy days easy. This tweak made a big difference for me. I think lots of us had pretty drastic FTP shifts during the rollout (mine dropped 7 Watts), but it has since normalized.
is that normalized power is being used to calculate athlete levels, rather than power. So even though you didnāt spend any more time in the threshold power zone, you did spend more time in the threshold normalized power zone.