Something to consider - the changes (assuming there were any) could have just made the system more accurate. Sure, athletes aren’t getting the ego boost they want, but if the results are more accurate, that is really all that matters.
I have no dog in this fight, but in my product categories, we see this all the time with consumers and their blood pressure…“My new BP monitor is not as accurate as my last one…it is JUNK!”. No, your new BP monitor is ‘different’ than your old one, and is likely MORE accurate, but you are used to the inaccuracy of your old unit.
I’ll throw my hat into this ring. I’ve noticed the same behaviour with AIFTPD as well. I think the intelligence has advanced and is keeping FTP predictions very conservative - perhaps in another effort to keep burnouts down? Just comparing the start of this season to last - which are extremely close in terms of TSS / CTL / weekly hours, there is a clear difference in the model’s predictions.
What is also interesting is that I always follow up running AIFTPD with a 20 min test as I still prefer my levels set that way. The progression as per the 20 min test for last season and this one are nearly identical, which makes sense given the work was more or less the same. However, AIFTPD has been way less generous, in the realm of 20W less generous.
Possibly, I’m not upset with the results and generally I find AI FTP works really well for me.
Its just a training number afterall. I am pretty in tune with expected results and only in the recent past have I been surprised with the results after years of using it.
In the ops case i would be dissapointed to reach those PLs with surveys marked moderate / hard.
I would love to see TR forefront other metrics in the app that showcase progress. They have PLs, but my understanding is that they’re more prescriptive than descriptive.
This might be the black box element of ML where it’s determining a change in FTP and can’t (?) communicate why, which would be very helpful guidance. Ostensibly it could give you that guidance going into the past or for the future.
If we could re-run AIFTP with old data and compare it to the old AIFTP predictions then we may know whether everyone is getting less fit/the custom plans arent working OR whether the AIFTP detection has become more conservative due to ML
So my AIFTPs stopped making sense too. I was getting nice 2 to3 pct bumps on the std low and medium volume plans. Then i started plan builder plan, agressive, upping the volume, completing every workout, first month a 1w bump, second month a 1w loss (moderate rpe surveys for ss, sometimes hard for threshold, but these are 1:45h in lenght). Progression levels have increased, setting best power records, heartrate lower, im completely mistified. My compliance was close to 100pct. I missed 1 ride in the recovery/deload week. Are you on a planbuilder plan? Im also puzzled why trainerroad has not gotten back to this topic.
I also experienced the same - to be honest I got itchy fingers after expecting a bit of a bump only to get 2.4% decrease.
I am generally a trust the process/ software, but I didn’t accept that reduction as I’m nailing all the sweetspot work which is key to my first A event of the season in a few weeks. I will run with it the next time, but if there’s been a tweak I’d be interested to know.
Ultimately, it doesn’t really matter, but I had kinda set a target as a goal!
for sure that makes sense. I just dont know if the AI being more conservative reduces burnout because it doesnt change your PLs and the PLs keep climbing. This week I am doing Threshold 7.5 and SS 7.9 haha. Feels bad after a recovery week. So the workouts keep getting harder no matter what the FTP is.
I am glad I am not the only one haha. Seems to be quite a few people that noticed this.
I do dissagree that it doesnt matter. I think it does matter. At the end of the day, we pay a certain amount of money for a software that promises something so if the software is changing I think we need to at least be informed. Like you said, many of us set a target and if completing workouts doesnt get you closer to the target… that sucks.
It matters mentally for me (and maybe many of us). After 4 months indoors, motivation is at rock bottom after 2 solid months of training and a reduced FTP.
Thanks for bringing this up as I get why this feels frustrating! That said, we’ve taken a look at your account and all looks well!
I know a few concerns came up about whether not accepting the last AI FTP increase or doing workouts on Zwift could have affected your detection (the Zwift workouts were my concern) but we verified neither of these would negatively impact AI FTP. All your structured and unstructured rides with power data are accounted for, so your results reflect your full training history.
However, I wanted to point out that several athletes in this thread made a great point: FTP isn’t the only marker of progress. You hit power PRs and moved up in your Progression Levels—clear signs that you’re getting stronger. Your ability to handle more work at higher levels shows real improvements in fitness, even if AI FTP Detection didn’t show an increase this time. TrainerRoad is built to keep you progressing, and these improvements show that it’s working.
Your consistency is paying off, and it’s awesome to see so many positive markers of progress in your training. If workouts are feeling productive, that’s a great confirmation that you’re on track. If you do feel like things are getting too comfortable, some athletes in this thread suggested using Workout Alternates to add a little more challenge while staying within the structure of your plan.
Yea, it is what it is it seems. No real answer to the many of us who are reducing their TR AI detected FTP and W/kg. All good, I have other positives to look at as others and Caro pointed out. I will complete this next training block, do a 20 minute test (outdoors since indoor is absolutely brutal) and then head outdoors for the season.
I will reasses in July if I will renew my yearly sub but, as of right now, I will not be.
Great chats and inputs from everyone there though, including Caro and TR team. Thanks!
I mean it does depend if those 1 hour efforts were all out. My current FTP is 288 W, my 1 hr PR is 259 W (not an all out 1 hour effort). If I hit a 1 hour PR at something like 270 W, that doesn’t justify increasing my FTP, and could easily happen with a decrease depending on other factors.
If your 1 hour power PR is very close to your FTP, I see what you’re saying, but I can’t imagine most have done a 1 hour flat out effort.
The big issue I have is that when you complete all training, and do not have RPE issues, i.e. you flag the majority of higher intensity sessions as moderate. How can your FTP decrease? The only causes would be that 1) you were on the wrong plan, it wasnt prescribing the right training stimulus and you actually got slower, and this should have been predictable. I.e. in my case plan builder would have prescribed me the wrong plan, and it would have been nice to know had it told me that by completing all of this you are getting slower. This is particularly concerning because I selected the the “aggressive option”. 2) AIFTP is wrong. I’m starting to lean towards this option, and the trainerroad response in this topic is not very in depth/satisfactory.
I have made one other observation on the custom plan that may be related. After the first month with only a 1w gain, i started rating all my trainings 1 step easier the second month. In the second month however, there were no adaptions suggested by the plan. I.e. it was still suggesting workouts with minor PL increases (+0.2) or there about. I actually in the last 2 weeks started swapping in +0.5 workout alternates with the same profile, just the harder versions and rating those as moderate. With a stretch +1.2 threshold one right before the AIFTP detection. Still a -1w result.
The most concerning part to me is that in the custom plan it doesnt seem to challenge me with workout progressions based on RPE either. Now my PLs have also gotten quite high, and ironically some of the SS workouts have in their description that if you can complete these then you should probably adjust your FTP up, however AIFTP adjusted me down… go figure. To me it seems there is a general breakdown in how it assesses fitness, prescribes the next workout and then the more easily quantifiable AIFTP number.
I’m definitely willing to take a look at any others’ careers who might feel a bit confused about their FTP Detections.
@ivegotabike, the OP did set three new power PRs this year – 256 in mid February, 257 in late February, and 260 just a couple of days ago. Their latest accepted AI FTP was 288 back in mid-January which I feel is really reasonable given the type of training they’re doing. FTP =/= 1 hour power for every athlete all the time.
@belciugo, if you have any other concerns, please let us know! We’d be happy to dig in further and chat some more if you’d like. Our goal is to ensure that everyone is happy with the product and has faith in how it works.
@napoleon1981, the same goes for you. It definitely seems like it might be worth taking a closer look at what’s going on with your plan. Let me know if you’d like to chat about it!
If anyone’s not comfortable discussing things on the thread, we can always chat via DM or even open support ticket if you’d rather. Just let me know! Again, we’re here to help!
The sickness I mentioned was prior this 9 week training block. This one felt good. I mentioned that because my FTP had dropped from 292 to 288 and then it went up to 290 5 weeks ago (i rejected the 290 because it was close).
At this point, the TR looked at my data and theres no clear reason why it is dropping. It is what it is.
Thanks @eddiegrinwald i have no more questions. You guys have already looked at the data.
I did get better at riding close/at FTP so thats a progression but perhaps the FTP Building Plan i signed up to do needs some adjusting since it is relatively new. I believe is tailored to get you better at riding at FTP (for races maybe) rather than raising your ceiling, which is what my goal was. Additionally, I am assuming it was confirmed that moving to zwift had nothing to do with it. I think there is something to be said there as the zwift ERG is quite a bit more forgiving than the TR ERG. The TR ERG is more precise so you stay on power more but it is more taxing. Perhaps the AI saw some NP discrepancies during interval despite overall interval power being ok.
Either way, I think they TR team needs to notice that quite a few people have had this concern. I know TR is always working on improving and innovating so maybe this should be a point of discussion behind the curtains. Lots of podcasts talk about how its the user’s fault that they missed workouts or did a wild group ride etc. Sometimes that isn’t the case and it should be looked at long term. I dont expect a fix now but just consider that something isnt perfect with the way the algorigthm is using PLs and RPE to calculate things.
Lots of good discussions here which is much appriciated.
If AIFTP is giving me/others the same, or reduced, FTP we need to know if it is related to a system change/ML change (so out of our control and thus we can continue following the plan as prescribed) OR something we have done (overtraining, failing to recover, working the wrong zones, doing too much outside etc)
I.e. FTP is not just a number. It is an important guide
Lots of users are experiencing decreases in AIFTP. I believe I posted up thread, and I have no insider knowledge here, but it seems like there may have been some “under the hood” adjustments that are dialing in FTP to a more accurate number. At the end of the day a few watts isn’t a big deal. I mean anyone can always go crush a 40-60 minute effort to see if you can surpass it.
