You can do it now, once. 100(1 - (Actual FTP/AIFT))
Using the numbers from earlier …
.. 100(1-(320-300)) = -6.67
It’s not going to be the same for every workout though. It really depends on what the AI picks for you. Sometimes you might get away with just a 2% decrease. For some workouts, your prepicked 6% decrease will change the intended adaptation, too. If you tell it your preferred FTP in advance, it will pick good workouts. That much I’ve seen.
1 Like
I’m with @eddie on that one as I think this isolated example here (not questioning your other points) feels more like an individual RPE rating problem and you should have rated very hard (and based more on the pure interval feeling and not so much with the long rest periods in mind.)
“due to the long rest periods and my relatively high 4-8min power” => high glycolic / anaerobic capacity athlete … given enough rest to refill the energy system … these athletes can suffer through many hard intervals. (This type of athlete maybe needs to change the RPE rating)
Could also describe why these athletes suffer more in long sweet spot which relentlessly burns the energy system down.
Maybe this scenario is not enough reflected in the ML models. Would love if the AI had some kind of individual expert slider in the settings to adjust for that (be it FTP threshold level normalization or whatever)
(This scenario reminds me a bit of the individual VO2max range of athletes but from another perspective)
1 Like
It’s only a problem if you rely on AIFP to set the system value going forward. The old AIFTP would usually get me really close to what I consider my physiological FTP, but that wasn’t perfect either and I’d sometimes set in manually. For me, it’s not a hard thing to dial it in within ~5 watts during the season when I’m training. No need for an formal test, I know by doing some tight over/under intervals or 20’ efforts just under the number. I don’t know or care if my definition of FTP matches anyone else’s definition, but it’s a physiological point for me that I’ve learned to recognize over many years of training. Having a number in the system that is 15-20 watts higher than that basically shifts my north star. And I can totally accept that making an FTP value my north star is dumb and flawed in many ways, but that’s where I’m at.
2 Likes
I feel I’m in this camp, if you look at my power ratings I’m skewed pretty far to the left of the table (excellent short duration power).
But to disagree on a point, I’m not burned down by sweet spot if the sweet spot is set accurately. In my history, it’s very easy to see my longer duration powers and try to pick something reasonable. Setting my sweet spot power to my 20- minute power seems overly optimistic of my capabilities and probably not what someone would be prescribed normally.
Before hopping back on the TR platform, I set some workouts for SS based on my 1 hour power in a recent Zwift group ride (heh, felt like a race), and my threshold intervals based on 20- minute power, and had some very productive workouts. But, the risk here is the zones are too easy and thus not as efficient over time.
I really like the idea of some adjustment option here. But yeah, we are somewhere off the side of the bell-curve here and can understand it being more difficult to adapt the general model to.
I agree with you and this circles back to the fact that all is good with manual FTP (if you have some reasonable approximation), which was never debated. I’m generally interested in seeing the automation work well enough for everybody though. Constant improvement.
I’m also basing it off the question if I could do another interval, and the answer would be yes - hence why I did not rate it very hard. This seems to be embedded in the hard vs very hard decision, at least to him. A better question after an over/under workout would be: Did the unders feel like unders, and, did the overs feel too far above FTP?
Like many others, I do have issues with the rating system, but that’s another topic entirely. As I mentioned, Nate specifically in the podcase a few days ago, that the rating would work for each athlete as long as you were consistently applying it. He even went so far as to say you could rate everything easy, but he kind of walked back from that. His point was, the difference between one or the other shouldn’t matter much. I should try and dig up the transcript
Unrelated side note:
I changed the two 90 minute Threshold workouts I have between now and 14th February from Hard Starts into Over/Unders, and each workout change took 1 watt off my predicted increase.
I just thought that was … interesting.
I changed them right back, obvs!!! 
This was a bug, it’s been fixed now.
5 Likes
My guess is only upsides. My conclusion is the system is half baked. It lowered my FTP from 306 to 298, glitched to 302 a few days later, then I did a 90 minute Zwift race up Ven-top, holding 305w for 40 minutes, dropping to 298 at 60 minutes, and 289 at 90 minutes. All that while keeping my heart rate average at 162, just under my LTHR of 169. Thanks to this effort my intervals.icu eFTP is 304w based on 26 minutes at 317w, and Zwift post race said congratulations, your FTP is 305w now. I did later change it back to to 302w to honor TR, but found I can no longer backfill workouts to reflect that my race effort closely aligns with a level 9 threshold workout.
Edit:
As far as my profile. I’m a 41 year old who has been doing NYC crit races with CRCA in central park and prospect park since 2016. My training volume in the span has been all over the place.
In my opinion, that is pretty darn close. Assuming you feel 305 is right, are you thinking 302 (or even 298) reflects “half baked”? They are all just estimates.
8 Likes
Yea the new model is just proving that it can do napkin math FTP calculations, instead of juicing it bit so my anaerobic intervals stay spicy for my crit racing calendar.
It’s been quite a while since TR assigned anaerobic intervals equally to everyone. If you are rating the anaerobic work as easier than the system thinks it should be it should quickly assign much harder anaerobic work irrespective of what your ftp is.
If it just raised ftp then that would make the workouts around ftp harder than they should be.
3 Likes
I know we’re all passionate folks, but let’s keep these discussions positive. I’m guilty myself of getting a little riled up sometimes!
2 Likes
These are all the same number 
10 Likes
Yea in a vacuum you are correct. However before last night my all time 1 hour PB was 281w, so how did TR arrive at 302w?
It ai, presumably looking at your previous workouts and RPE on that.
Sounds like in your case it was fairly accurate to an actual max effort even without you having done that max effort which is the idea behind it right?
2 Likes
Well if anything still have 4 months left on my annual sub. If other platforms are giving the same numbers for less, then TR will be a hard sell to my wife.
I think tr would argue their platform is much more than just an ftp estimation tool.
But if that’s all you use it for then yea, probably not worthwhile.
However if you use it to plan out your training for an event and let it adjust training for you based on your feedback then maybe it’s more worthwhile?
7 Likes
On update (+7w from the day before’s AI FTP detection), Wheel -2 which is 3.6. Rated Very Hard whereas would normally rate it Hard.
Predicted FTP +4w more (so in one month +11w total), first workout is Wheel -2 again.
1 Like