This is where I think there’s no substitute for real world testing. IMHO it doesn’t need to be an “FTP test”, for about the last 10 years I’ve done the same early season 10 mile TT and that’s always a good benchmark for where I’m at.
Based on that I can then make an informed decision about whether I’m making progress. I can then decide if the plan I’m following is working for me etc.
“just wasting time” - I know what you mean, but hey we’re on bikes, staying fit… it’s not a bad use of time 
Thank you for looking at it. Yes, I have contacted CS at TR twice this year. They refuse to respond with anything other than a cut and paste response. My last question was entirely unanswered.
Just a personal note, and not shade on anyone else. I’ve loaded up my race schedule heavy this year. And for me personally, I know my training is off despite following the TR plan for my events since November. I’m big on personal responsibility. It is time that I take responsibility for my progress. These upcoming events are going to be a sufferfest if I don’t. I’m a long time kids coach, and I’m on an amature team. I’ve brought lots of people to this platform. I’ve identified and concisely notified them of bugs that they have fixed. I’ve advocated for them and their product many times. I have household members who also pay for subscriptions and use TR. All that is to say, I tried. But if I keep down this path it’s all on me.
I have hidden it as it’s been way off at the start of the 4 week prediction cycle. I’ll try it again when it comes out of Beta.
I think it relies pretty heavily on your reported RPE.
I would mark a vo2max as “hard” or even “very hard” and the estimate would drop. But the reality was, even though yes vo2max workouts feel hard, they are SUPPOSED to feel hard. So the algorithm actually wants you to call it “moderate” if you would’ve been able to complete additional intervals.
That subjective difference IMO accounts for an enormous amount of slop in the system. The more clearly TR can articulate how to properly fill out RPE, the more accurate the AI will be able to predict.
I love you all no matter what TR says your FTP is.
I played around with this. I’m very close to an FTP update. I had predicted FTP (Beta) turned on. I did a sweet spot work out and it was very hard for me. I clicked very hard, and it dropped me from a 0% (257-257) gain over the last 14 workouts, with 2 workouts to go, to a loss of 5 watts (257-252). I went back in a picked hard and it jumped back up to a 0% change (257-257). I then edited it again and picked easy and it jumped up 5 watts (257-262). It is entirely impossible that someone that has been doing structured interval training for years, and has been on a continuous plan since November with no significant deviations is going to move 5 watts FTP from one ride, let alone the answer to a survey question about that one ride.
My wife and I had a conversation about this, and she had been discouraged that her FTP was dropping despite following her more aggressive plan even more strictly than I have for months. We did a 100 minute Z2 ride today and she could get a 3 watt predicted FTP (Beta) boost by picking easy vs moderate. She has been working hard for months, with the “press the button” FTP update in the previous version being a major motivator. She is very irked that one survey answer on a almost entirely inconsequential ride (at this point) can move the needle that much. It makes the whole thing seem fake and cheap.
We are both looking for more real metrics to gauge our fitness and progress. Plus we are using this, potentially bogus, value to scale all of our workouts. We even discussed returning to the much dreaded FTP tests. I may interrupt my training next week to do one outside of TR because I want the usual known algorithm math that I can manually verify.
I think this is where the issue sits with most people.
The AI is simply trying to find an optimal ”FTP” figure to base your training around, as the workouts are based on % of “FTP”s.
This is not your “fitness”, 95% of your 20min max power or your hour power….. or whatever benchmark.
If your RPE is higher than expected, then it makes sense that this can be interpreted as the “FTP” figure that the workout was based on was too high to get an optimal workout.
This is different than your fitness going up / down or sideways based on this workout.
So while an individual workout (missed, perfectly executed or otherwise) is unlikely to shift your real fitness much, the workout will help form the AI’s (or a real coaches) understanding on how to pick the best set of workouts going forward.
But FTP has been used as a Fitness benchmark and not just a figure to base training around for so long, most people are struggling to detach this from just a training anchor.
If TR just changed the workouts to be based on a % of say “Trainer road functional power (TRFP)" and gave a second figure for FTP and or a 20 / 60min power, these could adjust independently and avoid much of the discord and heartache.
Then for example your estimated FTP / fitness benchmark (based on a ramp test or otherwise) can go up as you might expect from 4 weeks of work. But from a workout scaling perspective, TR can adjust your TRFP up / down or sideways to get the next optimal set of workouts independently.
The AI is continually adjusting your FTP based on how you do on workouts. I did an over under workout yesterday, level 4.4. My heart rate went very high, and I rated it as very hard. I completed the workout successfully, no breaks or reductions in intensity.
In the “old days” I would have interpreted that as good progress and a boost towards getting stronger.
The AI engine decided differently, and dropped my predicted FTP by 4 watts. I understand why. If my FTP really was what it was predicting before the workout, then why did the workout feel so hard?
Now - I’ve been a little sick recently - cold, not sleeping well, high resting HR - so that’s a contributing reason to why the workout was hard.
But all-in, I think the AI engine made the right call in this instance.
Arguing to TR AI that your FTP should be higher.
As per forum guidelines, sharing calendars or screens of training is a big help here.
What does aggressive plan mean here? If talking about the sliders, more aggressive is not a benefit to most users and is worth playing with to see what the prediction does.
I personally prefer the near instant feedback in planning and execution over waiting 4 weeks to find out my training wasn’t quite right. If you told a coach your easy ride felt a little harder than expected, what would the coach do? (Probably) Exactly what the software did.
My calendar is public. My wife isn’t the one that made the post. It meets forum guidelines.
I also said her training is more aggressive than mine, I didn’t say she used the aggressive slider setting in plan builder.
But the big question: If the AIFTP is inflated, are we allowed to say that’s our FTP??
100% yes, unless you got a higher value at some point in the past with a dodgy Shimano power meter 
It’s a transformer based sequence model based on Nate’s post he’s made in the forum.
While this is some what true it’s still not fully explainable like traditional statistical models.
It’s really annoying their software has gone from bad to worse! If you disable the prediction the feont page widget should go away seems like UI 101 and they can’t get it right
I have noticed some wacky prediction behaviour. When TR calculated a FTP at the end of the 28 day period it was way under what other platforms have stated my FTP should be.
After returning from roughly 9 weeks away overseas, my FTP was set at 209. Prior it was 298 (November 27th)
Since then I have made some big gains, and have moved up from D cat on Zwift to C cat, and won a C stage in the Zwift games and then came 4th on Stage 5. During those races I average 3.0 and 3.2 w/kg for the full races of around 1 hour. I weight 91.3kg.
During the final race (Zwift Games Stage 5) the NP was 322, and the IF was 1.54.
Despite these races Trainer Roads predicted, and then assigned a FTP of 252.
Intervals is saying 282 FTP
Zwift is saying 302 FTP at the end of the race.
Zwift Climb at the end of the three step sisters climb said “295W”
zFTP is: 299
Training Peaks is saying 301 FTP
I assume I am missing something. Prior to leaving on my trip I felt the FTP within TR was accurate, and workouts were fairly well prescribed. These new changes and my current situation has left me worried. I’m willing to put in the work for training, just want to make sure its the best value for my time.
I am considering just setting the FTP to somewhere between what Intervals is recommending and zFTP, i.e. 290-291.
My Zwift Power Profile: ZwiftPower - Login
Now they keep pounding on “there are so many definitions of ftp, there isn’t a right or wrong way to define”
my .02
for YEARS, this forum has been riddled with threads about TR dishing out too much intensity.
NOW, there appears to be a way to dial in the appropriate intensity for a rider, which usually means their FTP will be dialed back to “where it should be”.
and the backlash ensues about diminished FTP values. for crying out loud…its either too much intensity with an inflated FTP, or a decreased FTP and workouts that are doable, at the correct intensity.
Can’t have it both ways. check your ego at the door, folks.
now do all the people who’s AIFTP is too high and the resulting workouts are too hard.
Hmm, good point. I have no explanation for that, except to borrow an old acronym…HTFU
All kidding aside, I guess those individuals need to wrestle through those workouts, answer the post workout survey honestly, and allow AI to do its thing??