For those of you playing with the predictive FTP function by adding/deleting/modifying your calendar, post your findings here?
Here’s mine today:
7 days (5 workouts, 7:45 riding hours) left until the next prediction date
Changing a threshold 4.2 (Centennial, 108 TSS, 0.85 IF) to harder suggested 5.1 (Mount Goode -1, 106 TSS, 0.84 IF) drops my prediction from 265W FTP to 255W 
Both predicted as easy (63% vs 57% respectively)
I’ll bite.
5 days (two 1 hr workouts) remaining. If I rate today’s O/U hard, the prediction is 280 W, if very hard it becomes 275 W. (The two remaining workouts are the same in both cases.)
What do you want to train at next month?
It makes sense that your FTP would be a bit higher if you rated it Hard vs Very Hard.
I got a cold a missed 8 days–5 workouts (3 hard and 2 easy). AI was going to give me 10 watts this 4-week block, but now I’m slated for only 5 watts in 6 days. Of course, we will see how recovered I fee on my first ride back tonight.
I had Pneumonia in November-December ftp dropped to 284, joined the beta a week or so before it launched to the public at large. I do about 4 bike rides a week all TR on a trainer right now. The AI ramped up the training to one of the steeper ramp rates I’ve had, got to 293 about a month ago. Estimated to get to 305 (near my highest ever) by 2/14/26 (Sat). Played around a bit with the settings as i went along this month changing endurance, tempo etc. progression levels and Workout placements during the week. This past week went down to 303 which i was still happy with. Played around for another 20min late last week. Got the AI FTP prediction to 310 (by merging the weeks workouts in to 2x double days) which would be my highest FTP ever. Just did my last WO and it bumped up to 312 after completing. For what it’s worth I think the post ride survey is confusing. 1-5 doesn’t make sense to me. 1-10 or 1-100 i could do. I have to read the description of what 1-5 means pretty much every WO to stay consistent. It might just be a me problem. Anyway, Playing around with the model can really change things up and it really adapts each completed work out so what it said would drop your FTP from 305 to 293 one day, might actually bump your estimated ftp to 312 a day or two later after its seen (I assume ) the work you’ve been able to do. At least thats been my experience. Adaptive training was life changing for me back when it came out, Red light green lights been helpful and so far this new model has been even better. So far my experience has been positive.
Saturday FTP came in at 312 and with 5 min of the new month of AI green star twinkle training the next FTP is projected to be 325 which would be nuts. That said i have a 4yr old so the idea of not getting sick 2 mths in a row is almost laughable. TBD.
Really though, playing around with the layout of your calendar and adding a little here/ there really seems to make a large difference.
It similar for me, I took the 5w dunt last night though as my workout did feel very hard.
Actual human coach: How did your over/unders go today?
Me: They were a little bit trickier than we thought they would be, based on the last few sessions I’ve done.
Actual human coach: Cool. Well, because they’re really sensitive to the number used to centre the intervals, let’s go ahead and lower that number by 5 watts for next time, just to be sure that the unders are where they should be to get the stimulus we’re after.
So … is anyone going to get angry with their actual human coach for that adjustment? Anyone?
Mine was interesting today.
Ended my AI FTP cycle, Went from 238 > 245 (originally was 250 at the start of the cycle). Happy with that.
It immediately set a new predicted AI FTP of 255 in 20 workouts time.
Changed nothing, refreshed the page a few mins later, 251W in 20 workouts time.
I think what I am learning with this is to pay no attention to the predicted FTP until closer to the time, like a few days away
AI Prediction cycle ends tomorrow. Came from 306W, went down to 297 with the new AI. Used to be predicted for 310W until I extended an endurance workout by 15min easy riding (marked easy), which brought it down to 307. Added another 45min Endurance for a today (marked yellow) to bring it back to 309 - only to have it drop back to 307 after the work was done (evaluated as 3 in Zwift, easy in TR)
That does not sound right at all.
Of course, you have a good point. Although in this case, it was more like:
Me: Quite well. Actually felt a wee bit easier than last time. I was quite afraid of them before the workout, but it went better than I expected. Could likely do one more U/O (i.e. 2+2 minutes)
Context: The “last time” was two weeks ago, 4x9, 95-99% (TH 3.4), this time 3x12, 95-105% (TH 4.5). Last time I was not sure whether to go with hard or very hard, decided on very hard, and the following suggested workouts were ok.
But I have to say, based on how I feel, I would prefer to use 277 W from Tuesday onwards, instead of the suggested 275 (hard) or 280 (very hard). My initial AI FTP detection (using the new model) was 265 on January 20.
I can’t think why……
They’re the one who said this…
It’s amazing that the system works at all for them - but each to their own…
I cant comprehend how there’s nothing between “moderate” and failing a workout.
To add another data point: 6 days left on the plan. At the start of the plan the prediction was 214. After adding an upcoming work trip with no time on the bike, prediction dropped to 212. My workout surveys have aligned closely with the percentages, I’ve accepted all workout adjustments, and had no failures. No manual changes to plan volume (only manual adjustment was moving 2 easy workouts a day forward to accommodate a morning flight). Over the last few days the prediction has been moving between 210-211.
This is about as consistent as I’d expect when following the plan to a T. I do think I’ll have to stop looking at the predicted FTP every day though 
or maybe just hide it (you can do that, right?). I’m too susceptible to obsessing about changes that aren’t significant.
Honestly - this just builds my faith in the AI. It has learned this user and responded accordingly. Wow.
Trainer Road in its next overhaul needs some sort of personality test to figure out what UX to give to each user. Give some people knobs that don’t do anything. Others a single graphic that just shows their next workout.
Clear signs of overthinking.
I ended up doing the original prescribed (Centennial) and I rated it easy.
Could have done another few rounds of intervals if needed (maybe we need an add interval feature along with the add rest?).
New prediction took me down a watt to 264W.
If 6x 6-8mins @100% = “Easy”….
…what about 1x45m@50%? Is that also “Easy”?
Can you see how applying such a lack of gradations between different effort levels might be less than optimal for a system that makes significant use of RPE feedback….? 
people are missing something fundamental;
training stress + recovery = adaptation.
during the training stress phase your actual FTP may not change much at all. however your ability to express those watts may change a lot based on acute then chronic fatigue. sometimes the point of a block is to ratchet up the stress, and workouts naturally go from hard to very hard. it does not mean that your physiology has detuned chronically. some rest/ easy days, and the freshness comes back. FTP may then resurface to the north, if you are in the ballpark with your work/ rest. Today’s ftp is silly, but we all played the game of prediction dynamics at the start.