That’s not what this means. You’re not doing FTP for 12 minutes, you’re doing it 3x. That’s a big difference. The first may be easy, but the third may be challenging. Many would argue the 60 minute definition of FTP is a bit arbitrary. There are many capabilities that have to be trained to hold your FTP for 60 minutes, so many postulated that anywhere between 35 minutes and 60 minutes is a good enough estimate. I personally don’t think you should train yourself specifically to complete an FTP test, but also folks with an under-developed aerobic base may sell themselves short if they select a too-low FTP for the sole purpose of being able to hold something for 60 minutes. Marion isn’t 35 minutes straight at FTP, but it’s not a terrible effort either. But main point being, it’s not your 12 minute power if you have to do it 3 times.
Thank you!
I think everyone really needs to wrap their heads around the “best PL” and “productive PL” concept. If you are consistently and productively doing 60 minute sweet spot workouts, but then go out with really high motivation and hammer out a 100 minute sweet spot, check the box to exclude that effort from your Progression Levels!
It wasn’t defined that way. It was defined as being closest to maximum lactate steady state (MLSS), which is the border between stable and unstable physiology.
To repeat, I’ll blow up in less than 10 minutes if I go 30W above ftp, versus 2+ hours if I ride 30W under ftp. My well-paced “blow up in 20 minutes” are about 4-6% above ftp. Thats me, your numbers will be slightly different.
This suggestion was to dial in an initial FTP value, if you have a good sense of what you think it should be but are not sure. Do Marion, if it feels hard but doable you’re probably in the right ballpark and AI FTP will pay particular attention to such a long steady state effort. I assume AI FTP also uses HR data and HR drift over time, so it doesn’t need an all-out effort and most users would prefer not to execute one. And once AI FTP has learned from your Marion (and other) efforts, it should be consistent with no further gaming of the system going forward.
What was the 257w AI FTP based on? Unstructured outdoor riding, TR workouts, other? As I said above, IME AI FTP is not very accurate when based on unstructured outdoor work.
Ooft. Imagine doing your “easy” days 15 or so watts higher than you should be.
Forever.
No, thanks.
That’s why it matters, and why saying PLs and AI will keep you right is potentially reckless. I’m not convinced they will.
that is an interesting topic, and I only see rapid gains after taking a break. Smaller jumps are more abrupt, and I use RPE and power-to-HR to inform and adapt on the fly. Long ago (before power meter) learned to feel out ftp. Not using erg helps, adjust efforts to feel.
Its based on me doing my structured workouts outside
Pretty similar to being done inside.
I just want to know what AI FTP is trying to tell me.
As someone said above, AI FTP is a black box. You need to file a support ticket and ask.
But how does a black box help you with training?
Is the perfect answer 
When given a good starting point, that black box will use machine learning to review your power data and map you to an estimated ftp (based on other people with similar power data). So when it works, it works well.
However the algorithm requires a good initial estimate. You’ve got one, so manually set your FTP with it. Manually input your field tested FTP, because you are seeding the AI FTP with a good estimate.
You also need to realize that Adaptive Training (AT) wants to use PLs for regular, normal, productive workouts. TR uses that to provide a more linear progression in workout difficulty over a training block (or two or more).
For that reason think of PLs as Training Levels (TLs).
Therefore you need to understand that if you do a heroic effort, for example the last 4 weeks you’ve been doing some high quality training at 4x15 or 3x20 sweet spot, and then you go out on a major climb and do 1x120 and feed that to TR as a completed 1x120 SS workout, well, you have bumped your PLs which will increase your SS Training Levels to something unsustainable.
Apparently AI FTP might try adjusting your FTP to help compensate, at least it appears to do that sometimes based on forum posts.
Are you using the same power meter for all your rides and workouts? There’s about a 20 W difference between my Kickr Core and my Assiomas. If AI FTP is mixing data sources that could be a reason you see the discrepancies. And that’s going to lead to a different solution to shit problem than adjusting PLs or whatever.
Exact same bike, same power meter and all done outside on the exact same terrain
Did you look at the “Cycling records” tab of your TR career page? Any outliers that could have caused it? (e.g. 310w for 5 mins or 290w for 10 mins?)
No outliers
Clearly TR is at fault here (@ZackeryWeimer might give his opinion here).
Train with the 235 value. @WindWarrior gave you all the advice. Always better having FTP lower / conservative than training against too high value.
2 other things to check:
- One reason that AI FTP is high could be that in the past you set a high value in TR because that influences the AI FTP output.
- Maybe you filled some surveys with ratings like moderate/hard when it was more of an all out effort etc.
yup. this is the correct response. might as well end the thread.
That doesn’t tell me what AI FTP is
Strong disagree. FTP 20 watts north of reality is bad.
You’ll be doing threshold when you should be doing sweetspot. You won’t be able to do long threshold intervals because you’ll be 20 watts over FTP. Over/unders will crush you. You’ll be doing all your easy riding 15 watts too hard.
IMO.