typo? 285->287 is just a 1% increase
Correct - that AI FTP came the day after doing an all-out hour effort workout (and rating it as such) and only achieving 262W over an hour. Whether 285 or 287 - both are just way off.
My point is that it is just more evidence that even with the AI FTP, it’s still heavily biased towards scaling short efforts and providing an inflated value. This is even directly in the face of clear data to the contrary, from an outdoor structured workout, including user input on the survey saying it was all-out, all nearing the conclusion of a specialty plan that was designed to prepare one for 1 hour efforts.
Seems like a place they really need to re-tune a bit…
2 Likes
How consistent are you in staying in the TT position? I feel AIFTP similarly over rates my FTP because I’m not consistent (its heavily biased by rides where I’ve not been aero) which in turns leads me to coming out of the TT position more 
This year it’s about 75% of my riding and pretty much all my power PRs this year down to 3 mins or so are in that position - I actually like it way more than climbing on my road bike now, and those climb efforts are kind of weak this year as a result 
So from that perspective, I think AT is getting pretty consistent data in recent weeks - majority off one bike, where the best efforts are being made.
1 Like
297 to 297 and did your PL got reduced -2 or 3?
I can’t remember now but I don’t think there was any change, or if there was it was insignificant, i.e. 6.2 down to 6.1
Sorry if this has been covered previously…
I’ve got a ramp test scheduled today but don’t want to do a test, I also don’t want to do an AI FTP update.
Is there a way to get the system to give me the workout that it would have if I had clicked to accept my AI FTP estimate without accepting the AI FTP it proposes?
No, I don’t think that getting a workout swap without accepting the FTP offering is possible.
You will have to pick your own workout to do instead.
Ok, thanks @mcneese.chad.
Seems like something that could definitely be a feature considering the team have stated that it is often not the best idea to always ramp/AI estimate at certain points in a plan.
Well, they have a clear plan to set the AIFTPD to an “Automatic” setting for most users in the future. This will end up watching for, and then offering FTP changes when the system sees enough change. That will happen more to follow the rider’s actual fitness progression, vs a simple structure based purely on a time projection (the usual 4-6 weeks it is now).
So, the future once that is implemented will be a different situation than we currently experience. As such, we won’t have Ramp Tests on our calendar in any way. AT will presumably be giving us pure workouts on the entire calendar which is the real solution that makes more sense, to me at least.
2 Likes
AI FTP initially pegged me at 300w, then 305w, then 311w. Did a 20 minute yo-yo attack in a crit last night and my garmin edge 830 said my ftp is now 312w. I generally do the mid volume crit plan and use the types of workouts in the plan as a guide but try to push PL breakthroughs.
I was just due for a ramp test and decided to check my AI FTP calculation. It seemed on point with the progression I expected. I did however go on to complete the ramp test anyway and it scored within 1 FTP from the AI calculation, impressive stuff! I expect to use the AI calculation going forward, unless I’m in the mood for self torture…
5 Likes
So you did 260W for an hour workout with 1min breaks sprinkled throughout? Sounds like an FTP in the mid 280s isn’t unreasonable…?
Instead of taking the 1 min rests at < 130W, as I was really interested in seeing a potential hour power target, I did the “rests” in the low 200s. But this was still no free lunch, because I couldn’t quite hit the low end of the range for all of the 10 min interval targets (probably because I wasn’t really “resting”) - so from that perspective, it was basically an all out hour effort as much as I could do.
Put another way, avg 262 / NP 265… so I wouldn’t doubt that 265 could be doable perfectly paced on a good day. But definitely not 285 - huge huge difference for that duration!
Still leaves me wondering - what’s really better training? Accomplishing hour efforts in the 250-260 range by manually overriding, or using an inflated TR-FTP in the 280s and getting workouts that are only shorter intervals but above real hour FTP?
One final data point, had our state 40k TT and I ended up at 269W for just under 57mins. NP 271 - well paced, everything I could give it. Very happy with the result, but obviously a huge gap to AI FTP @287 which I’ve never come close to achieving.
Which brings me back to the point - a TR 40k plan set at 287 when your real ftp is 269 will never bring you too the continuous hour type of workouts you expect to see in the 40k plan. It sets you up to fail workouts in the 6+ PL range , and then AT reverts back to intervals and long progression workouts from Z2-Z4 in the PL 4-5 range.
Not a mainstream use case, but something I think TR needs to take a look at for these plans. With a more correct ftp, the PLs would have advanced enough to have served up the expected near continuous hour workouts approaching the end of the plan.
Or, the AT on this plan needs to replace with lower %ftp long continuous efforts consistent with the goals of the plan to prepare for 60 min of work when late in the plan.
5 Likes
Awesome effort 
I could be wrong but I think IMC my AIFTP is influenced by my road bike power and its about 25w above what Ive put out in the TT position for an hour on the turbore and about 35w higher than my best on-road TT this year. I did about 14months ago exceed my AIFTP in a TT but I doubt that once off is still influencing things.
1 Like
Funny, I was just replying to a more active thread you had on this subject with Nate 
Agree that if you have stronger road efforts going into the dataset, that will definitely skew any TT prediction since they don’t model FTP by bike or position (yet).
In my case, all my best efforts this year and most of my training is on my TT bike, making the data pretty clean (and making my road bike climbing really suck ) so this is why I think there is really some room to improve how athletes focusing on longer power are managed in these algorithms.
3 Likes
I wrote about a month ago that AI FTP after my last block was on point down to a single wart. I verified it by completing a ramp test to confirm.
However after this recent block I felt like the AI FTP estimation was way too low, compared to how strong I felt on the bike. I decided once again to verify with ramp test and scored 7% higher than AI had estimated, an 11% increase overall from last month. This is quite significant, so for now I will stick to doing ramp tests.
Will report back next month.
1 Like
Lol, Finally AI FTP has downgraded my FTP slightly; it has up to now been increasing slightly. I still think its biased upwards by my road biking position and I’d struggle to maintain the oxygen intake/ courage required to sustain it long term on the TT bike but hey ho, that’s a ‘first world problem’ but at least its getting closer to more realistic 
I’ve decided to do what I was doing before AI FTP set my FTP at the level I think it should be in the TT position and leave it (only tweaking occasionally) rather than letting AI FTP keep it at a level I feel is too high . It seemed to be working better for me before that way (its actually within one or two watts of what my garmin keeps on telling me anyway) but then again maybe this year I have done too many work outs outside and dropped Wednesdays for a chaingang (paceline) so perhaps I’m looking for the answer in the wrong place