I can see myself doing this as well. Out of curiosity, was the final AIFTP in line with predictions, or was it slightly higher due to shedding fatigue after a recovery week?
Is it?
Sorry - I’ve never seen a plan like that before 
Definitely reach out to support to make sure you have picked the best settings in plan builder to achieve your goals.
I wouldn’t worry too much about what your predicted FTP says though until a few more people have actually reached their FTP prediction days and we can assess how accurate it is.
Plan builder lets you put solo or group rides in instead of intervals. I not entirely convinced it works though because it often tells me to do 220 TSS in 3hours, which is basically 7x20 @ 90% … which I could do maybe once but not week in week out.
For me it was slightly lower 233 vs 236 but I’ve been extending my rides to complete the Tour de Zwift, and 3 watts is close enough for me.
Over all I’m still a bit confused myself because while there’s a lot of posts how the new system have has some unrealistic FTP but so far for me it’s been realistic.
I don’t know if it’s because lately I’ve been force to do all my rides indoors on the trainer.
I’ve gone from 242 at the end of my season, to 222 after the break and at the start of the Beta testing now to 233.
My workouts have felt spot on and the new system looks like it’s not playing with my progression.
I don’t know what will happen once I start riding longer outdoors again but I’ve seen that this new system does a better job analyzing unstructured rides that I’ve done on Zwift.
Can you replace them with “dynamic endurance” days?
That would make more sense to me than “hard” solo rides.
I’m in a similar position. My new AIFTP is very much in line with my expectations, and I’m confident that AI will suggest appropriate workouts. I’m much less confident in the prediction capabilities as things currently stand, so will just ignore if for now and knock out a 4 week training block and see how things are looking after that. I can then sync the timing of future FTP detections with recovery weeks - which I’m hoping will lead to more believable predictions.
I just wish I could move the detection date now, rather than having to ignore the expected reminders about there being a new FTP detection part way through a training block.
It’s a conundrum you don’t really want … trust the plan, the prediction is wrong and you get better. OR trust the plan the prediction is right and you’ve wasted 4 weeks when maybe you could have done something more productive.
Realistically I don’t get paid to ride so it’s the same as self coaching experimentation at this point.
The real problem right now is this just feels like a big black box that spits out a number with no insight and you can’t take any action on.
I’ve been using TrainerRoad for many years, so I know what gains to expect from a training block. At the moment, the AI FTP prediction isn’t aligning with my expectations. This did cause me some concern and confusion at first, which also seems to be the same for you?
I’ve had a few days to think about it now, and am happy to ignore the prediction. I’ll accept the workout suggestions because they look reasonable (and aren’t that much different than previous plans).
If you keep doing what you’ve been doing up to this point, then you’ll know what the real world results are likely to be, despite what the prediction says. At this point, I’m trusting the prediction less than the long range weather forecast
As an experiment and to remove a variable …
I’d be tempted to replace the Sunday hard rides with TR workouts that mirror as best as you can manage the TSS and nature of those rides.
That way you can confirm whether or not the unexpected behaviour is caused by that type of ride being interpreted in a funky way by the AI.
Hey @trebor !
I am just getting to read this thread. I took a look at your training plan and it looks like you deleted the plan?
I liked @Helvellyn idea of swaping out your Sunday’s ride and then unpinning the workout from your screenshot.
I have a feeling that starting with less volume (especially on Sunday) would be beneficial as it’s triggering much needed rest. So trying to maintain a balance between training and recovery is key to see fitness gains. .
@Caro.Gomez-Villafane Yes, I did indeed. Sorry for wasting your time and thanks for looking.
Before doing that I did try reducing the Sunday ride time, the Sunday ride intenstity level, both ride time & intensity and then actually just deleting the Sunday right altogether. In all cases there was a predicted FTP drop.
I guess I just got to the point (based on the info in TR) that the plan was not going to help me achieve my goal. And in that case, no point sticking with it.
The lack of insight into the “why” behind the numbers is a bit unfortunate.
I went through exactly what you’ve been going through completely deleted a plan started over tried this tried that every configuration showed my FTP was gonna go down
Through trial and error I finally came up with something that shows I’m supposed to have a 2% increase in 28 days
I put in master’s plan
Balanced
4 days per week
Let TrainerRoad AI give me my complete workouts didn’t add or subtract any workouts duration nothing
I’ve been messing around with my plan and the predicted FTP and I think this could be a good tool.
While I don’t think that the Predicted FTP number is accurate, I find that it’s good to go off what changes will give you the biggest jump.
On Sunday it was 244 and went down to 238 for some reason, with some changes it’s back up to 241 that good enough for me and looks like a normal TR increase.
So far TR Check Volume donesn’t want me to increase my Volume even though it has see in the past that I’ve done that higher volume before.
I’ve found that starting with a lower Volume the first week and ramping up the time from there has gotten me the best Predictions. Just have to watch out for what workouts triggers Red days, I’ve lower those times to avoid those days.
Also keeping that fourth Recovery week lower time wise.
I’ve set Sunday’s Endurance Ride to Dynamic but and let TR do it’s thing.
I’ll be keeping and eye on those, for now in the winter I’m ok with those shorter rides but as the weather gets better if it doesn’t scale upwards I’ll just add in the time and take the FTP hit, which coming from Old School thinking I can’t see how riding your bike would make you worst at riding your bike. 
This is a real shame.
It’s just a prediction.
I would have hoped that people would at least stick with it for 28 days to see what happens.
I’m not convinced that the predicted FTP feature is a great idea - people seem to be overreacting to it - such a shame because it’s such a cool tool.
Not wasting my time! If you want to DM me we can work on building a plan together.
There as many reasons an AI FTP Prediction could be going down per Nate’s post.
A little bummed you didn’t try it to see how it goes. The AI FTP Prediction is like the weather forecast, it becomes more accurate as you get closer to it. So just because it was showing a decrease off the bat, doesn’t mean that’s the final FTP you’re going to get 3 weeks out 
Thanks. I have listened to both the podcasts. Still don’t really get the value of a number that fluctuates in an essentially (from my perspective) random manner. Without insight how can you take any corrective action?
fully appreciate it’s not your problem to solve and it is what it is.
But you shouldn’t let it put you off training entirely 
Can you or someone at TR confirm the window that the AI prediction uses for the prediction? Is it 28 days? From today to the prediction date? Back 28 days and forward to the prediction?
It might be help people understand why the predictions change.
Each prediction is looking back at your training history, including incorporating everything you’ve done/planned to do since the last detection.