PancreasFail, have you considered not telling TR about other activities and just letting the cycling speak for itself?
Just spitballing, maybe the AI is dialing back the cycling levels because it is calculating additional training stress instead of detecting it through workout results.
Also not sure how the breaks during recovery between intervals are interpreted — that’s usually a sign the interval was very hard if the break is of necessity.
Old AI FTP (Dec) : 289
New AI FTP (Jan 21) : 296
Initial AI FTP Prediction : 312 ** Prediction was originally 306 based on my existing plan. I moved around a rest day to get it up to 310. I added an extra V02 workout before my recovery week to get the prediction up to 312 but that quickly came back down to 310 and stayed there for the last 3 weeks of my build block.
New AI FTP Detection (Feb 18) : 310
New Prediction for March 18 : 324
Old AiFTP detection (January 17): 301 W
New AiFTP detection (January 21): 309 W
Initial AiFTP prediction: 333 W
Actual AiFTP prediction: 331 W
The prediction had an early drop to 331, and after that alternated between 329 and 331. I’ve had 100% compliance to the plan in regards of the amount and type of workouts, but had to slightly dial back a couple of the last hard workouts. That had no (apparent) effect on the prediction. One VO2Max workout was rated very hard which at the time dropped the prediction from 331 to 329, but for some reason it crept up to 331 again in the last week even though that was when I dialed back a couple of workouts.
New AiFTP prediction: 340 W.
331 watt is already a bit high for me, looking at the upcoming workouts. It has me doing 5 minute intervals at a power I struggled to do for 2 minutes only last week. I probably have to manually dial back the first workouts or fail them to show the Ai that it thinks too highly of me.
A bonus quirk for anybody reading all the way down here. I tried dialing back the first (upcoming) VO2Max workout to something I think I might be able to handle, just to see how the prediction would change. Going from 4x5 at 110% to 4x5 at 106% made the prediction go from 340 W to 364 W! Then I did my scheduled recovery workout and now my prediction is 367 W.
I know the prediction will plummet once I start doing my hard workouts, but at some point TR should probably introduce a sanity check for the predictions.
Just came back from a 6ish week break over Dec and Jan
November 5th: 334 (old AI)
Jan 19th: 313 (old AI)
Jan 22nd: 301 (New AI)
Initial AI Predicted: 338
Feb 19th Result: 322
New prediction: 346
Skipped 1 sweet spot session and changed one to endurance due to illness in the second week
I have swapped a couple of sweetspot workouts for zwift races and had a week off in the states doing nothing so probably 70% of the workouts i’ve done.
Overall after the inital getting used to the TRFTP (trainerroad FTP) which they should call it, i’m good with the system, that number isn’t set in stone and moves around depending on what you do, it’s a nice indicator but the game changer has been the Normalised power chart so you can see each workout pushing you or not.
So far really happy with the projection system, not too worried about the fluctuations, the workouts every day have been just right.
What did you end up doing? I feel like my AI FTP is now too high. I can either turn down the intensity a percent or two and rate the workouts harder, or I can manually dial back my FTP. Trying to figure out the best route forward.
Changing your FTP has no impact on how hard your workouts are - it may affect time in zone a little (lower FTP more time in zone) - but the actual difficulty of the workouts is defined by the AIs assessment of your fitness - which is independent of your FTP.
Personally I’m progressing nicely so far with my AIFTP set FTP so I’ve stuck with it.
Agree with @kevistraining here.
I got an aiftp bump from 205 to 241 at the end of my recovery week and I had a feeling it was too much of a jump and was concerned I’d struggle when the next block started.
I’ve just done my first TH workout in the new block and it felt surprisingly doable on the day.
I will say I was upbeat and ready for a challenge today which no doubt helped. Had my lizard brain had more of a say, I may have talked myself into less positive mindset.
I like the workout comparison that shows how and where the workout will push, and knowing it was asking for 10 ish watts more than my previous pb a couple of weeks ago helped me rationalise it.
So apparently I ran into a bug. I emailed support and they had me basically force FTP detection to run again. That gave a new FTP of 420, but my FTP predictor is still broken and “needs more data.” oh well. It was apparently wrong for the whole last block anyways if it was predicting 400 and my increase was actually 415-420.
With that said, there’s no way I can do 420 for more than 10mins right now. But I doubted it last time and it turns out I completed all workouts and progressions fine, so I’ll do the same here and see what happens and maybe also print out a t-shirt and socks that say my FTP is 420.
Just had my 28day FTP Detection 253w (4.05 w/kg) and they are predicting 260w (4.16 w/kg) in 28d time. My good bike (with a power meter) will be out soon for group rides, so I wonder if that’ll actually be achievable 
Quite a bit of shuffling in my non-cycling activities to stay on track, and rarely heeded red and yellow days, hard to keep sessions short to avoid collapsing ftp prediction.
I’m on a plan with balanced approach in base phases right now with low volume (3 rides/week planned, 2 TR workouts and an outdoor solo ride on the weekend … 3.25 hours per week). I missed a workout the second week, and shortened my solo ride. Week after that was my recovery week. I skipped the solo ride that week. Next week I missed one of the workouts and extended the solo ride from 1:15 to around 1:50.
This whole time the prediction for the AI FTP kept dropping. No surprises when I’m missing workouts. It was down to a 1 W predicted increase after 3 weeks. Starting week 4 and I did the first workout and rated it moderate rather than what it thought should have been hard for me. Previously all the workouts in the past 3 weeks had been rated hard except for the recovery endurance workouts. With that one workout and rating my predicted FTP jumped up something like 5 W … interesting. That was the beginning of last week and it was 2 days before I could hit the detect new FTP button.
Then I missed that next workout as I started some travel for a few days. Missed my weekend ride too after getting back home. Back on track with workouts this week, but waiting to pull the trigger on the AI FTP prediction button until my rest week (next week). I don’t like updating my FTP in the middle of a block. I know it’s not supposed to affect the workouts TR gives me because the AI doesn’t use my FTP or whatever, but I’d just like to wait for a rest week.
Currently its giving me a prediction of 238 W (+3% from 231) in 28 days. I’ll hit the detect button either at the beginning or end of next week (my rest week). I’m expecting it will only bump it up a couple watts, and maybe I should be mad about that after ~6 weeks of training, but I’m good with whatever. I’d expect more if I had been more consistent and I had been in a build phase, but since the AI FTP detection on 1/22 I’ve missed 3 of the 11 planned TR workouts, and 2 of the 5 planned solo rides.
I’m also not expecting the solo rides to add much in terms of FTP increase. I got a MTB this year and I’m using those solo rides mostly for getting to know my new bike, relearning how to ride on trails, and utilizing new muscle groups (who knew you need arms and core for mtb!?) rather than just a workout … and that will come with some tradeoffs.
december 15, restarted TR after using a coach for the past year (ill likely resume with him but he has some life things going on). 217 which felt insanely easy as my previous setting on training peaks was 260.
jan 19 aiftp: 232
feb 19: 238, predicting 250 3/19.
this still feels insanely easy. Im trying to trust the process but the only rides i have marked as hard were races.
I am not 100% sure on the dates. But otherwise as follows.
My FTP was at 301w and after AI FTP detection jumped to 305w.
I had Build block starting and prediction was 317w by the end.
I did 100% of my workouts, nothing extra. 100% focus. Some of the workouts I gave very hard but AI was expecting hard as most likely (prediction: 60% hard, 40% very hard).
My AI FTP prediction was going down every week and after build block that was a success in my mind my FTP prediction gave me 303w.
To be honest I dont really understand this feature. Maybe I will turn AI FTP prediction off because it does not seem very reliable. I created ticket for support so lets see.
Sounds like it’s working perfectly 
Starting AI FTP: 181W
AI Prediction: 213W (still 1 Workout left, should come Tue 3rd)
Result (March 3rd): 211W
I swapped the last workout with an outside ride, hence the slightly lower number.
Now, the next predicition is 226W.
So far workouts seem really good, it had me starting with 4 workouts a week but that was a bit too much for me. I am not at 3 with one of them being an endurance workout and the other 2 threshold and Vo2 Max workouts.
Initial FTP 309 W
Predicted 319 W
Current prediction in 5 days: 310 W
Quite a few workouts were too hard and a few that weren’t were downgraded for adjusting the TV (causing pauses). Not very impressed so far.
Did AI FTP detection after this block, it claimed I went from 309W→317W, and claims the next block will get me to 325W.
Would be nice if true (>4W/kg), but I don’t really believe anything of it. I had to take larger pauses for every workout close to FTP.
How did you rate those workouts? I think for best results they should be All Out / Max Effort. I think that should corral the FTP increases.