They should drop it. Enough content around yt and the internet of new cyclists entirely spinning around FTP and then seeing themselves dropped in a first race or event cause their 5min power e.g. and the ability to recover from surges is a somewhat different story than sinking into a ramp or 20m test.
Just nitpicking, but I believe this changed a little over the last months when another change around the Adaptive training was released. There was quite some buzz around users getting workouts selected which were way beyond their PLs but turned out to be achievable. Part of why this happened should have been a better analysis and incorporation of unstructured rides.
1 Like
deleted, wrong thread
2 Likes
Updated today to the new ai platform and am a bit puzzled what the logic is. My FTP went from 276 to 268 (who cares) but my predicted FTP in 28 days is a big drop to 252 and oddly the next three weeks workouts progressively get easier instead of harder during a build phase. It’s like it’s purposely down training me?
I received the new AI update today as well and was very surprised when it suggested an ftp increase from 288 up to 299. This doesn’t seem realistic to me as I’ve done very little intensity since November when my local cyclocross season ended. Also in the past whenever I’ve done a test and started up training again in the winter after a little break and lack of structure, my ftp number has always gone down a few watts.
I’m not inclined to believe this bump as all I’ve done the last three weeks are some moderate sweet spot and threshold workouts with some endurance riding around them. If I keep my current number, will his mess the whole thing up?
1 Like
This is explained in the very first post of this thread
7 Likes
I’m a bit concerned with the new AI FTP detection. I feel it is highly inaccurate. My last recent AI FTP detection was 343 watts, which would required a 20 minutes FTP test of 361 watts. I have no confidence that I could hold anything more than 350 watts for 20 minutes on a really good day. I already thought this detection was inflated.
Now, TR increased my FTP to 356 watts which is ridiculous. It means holding 375 watts for 20 minutes. @Nate_Pearson The new AI FTP is designed purely for TR training data or it is aimed to be accurate for real world cycling?
1 Like
11 Likes
2 Likes
I see a lot of recent activity in this thread 
So yeah maybe storm is coming? lol
AI is aiming to reduce my FTP which I normally wouldn’t mind but it’s from 267 → 249, which seems like a very big step. I usually cross-reference my TR “estimated” value and Xert, and they are very close and that is the case today too.
I completed all my TR workouts just fine, and sometimes would even increase intensity during the workouts.
What’s the consensus, should I just accept it? cc @Nate_Pearson
1 Like
I had a similar experience today. 7.7% drop. I decided to accept it for now and see what happens over the next 30 days.
Now I will go pout for the rest of the evening and then get back to training.
5 Likes
I feel a lot of this confusion could have been avoided by not calling it FTP.
It’s not FTP, that’s fine but calling it that just creates confusion for those who didn’t read a very specific thread in the forums.
No reason you couldn’t keep both around.
FTP estimate and TR Watt level or something. Yeah you need a better name, but if it’s not FTP then you’re just making people confused.
15 Likes
Did a IRL ramp ftp and vo2max test with a coach mid-Dec who said I don’t have to go all out as it’s base…it was 285w.
Since then, previous-gen AI FTP said something in 270s and I’ve declined it and kept 285….today, the new AI FTP says 251.
Recall that the previous AI FTPs were different from IRL ramp tests by less than 10w…and was even different by only a few watts after one IRL test…..didn’t accept the new 251w as I’m thinking that the workouts will be too easy.
I have been dealing with a head cold for the past couple of weeks, too, so I’m wondering if my numbers (higher HR per watt?) are off, if not accepting the prior-gen TrAIFTPs, or if being on a calorie restricted diet have been throwing off the new-gen TrAIFTP.
1 Like
That makes two of use then. Kinda difficult because it’s a drop after a EOY 277→267 drop, which I accepted as a “build phase start value” due to minimal cycling around the holiday season and snowboarding in December.
To be clear, I was never using this value as anything other than auxiliary indicator of progress and fitness level. I care more about power curve, and power over specific durations (1/5/10/20/40 minutes), understandably, I could never hold my TR FTP (or Xert TP) for a full hour.
I’m just used to certain imbalance of progression levels (never high values in Threshold categories, reasonably good endurance/VO2, anaerobic), and my concern is that my calendar is gonna be messed up.
And yes, it is: I’m forecasted to make an exciting 249→247 move in the next 28 days.
I’m looking at my calendar, and I just see that those workouts won’t do it. VO2 days have to be harder, what’s prescribed as Threshold is lower than my SS I completed today…
@Nate_Pearson or someone with TR expertise, do I just go and manually override all workouts for the next 30 days to see a positive forecast?
1 Like
3 Likes
IDK what role, if any, HR plays in AI FTP detection; personally, I constantly overheat indoors due to lack of A/C and reduced cooling overall. 267→249 made my calendar look way too relaxing to a degree I can’t follow it, so knowing that - yea I wouldn’t be accepting 285→251 drop.
p.s. My VO2 intervals are like olde lower threshold now 
When you’re building a plan, you can individually adjust difficulty of each zone (make endurance and anaerobic demanding, and threshold balanced/easy for example). Also, you could just replace the first couple workouts with something harder and rate them as easier and AI will pick up on it and adjust future workouts.
However, while I was building my plan, AI thought I’d do better with a balanced approach vs demanding. Play around with it and see if adjusting volume/intensity changes their prediction. I will say prediction isn’t written in stone, but if one approach you see -10W prediction in 28days and a different approach you see +10W, then I know which one I’d pick.
3 Likes
FastCat coaching has FTFP, whatever that is, I think Wahoo Fitness uses something different also guess for TrainerRoad TRFTP or stick with AIFTP but know it’s not the same as Universal FTP.
1 Like
Thanks @BenB ! I’m more concerned with the messed up calendar than the nominal value itself. I’ll try to fool AI by riding higher pl workouts and rating them “easy” as you suggested, but honestly I just don’t like the process of now telling AI what I want to see on the calendar. Kinda defeats the purpose of AI but its current recommendations don’t look productive at all.