No I’m not…I have done exactly that two times. I thought AIFTP was low and I did FTP tests. One ramp test and another all-out race effort up Alpe de Zwift.
If you think AIFTP is wrong, then take the steps necessary to have a correct FTP, which is critical to a proper training program.
Out.
Yes you miss the point. You are talking about something i was not asking for, but thanks and bye.
The only way to know this is to test or compare your pdc.
If you don’t trust ai FTP (and until you can verify its accuracy I think doubt is good) the only way to check if it’s correct that your ftp has decreased is by properly testing said ftp.
Would you believe ai FTP if it said you had gained the same watts it’s saying you lost?
That would be more aligned with the direction i thought things were going yes, i wouls have likely not posted about that. If i press the gas padel in my car i would not be surprised to go faster. I however would be trying to figure out what is going on when i press it and i go slower. Espescially when the car moves faster but the speedo shows a decline. I would question the speedo.
And you could verify if the speedometer is correct by, well, testing it.
A mod should just sticky one of these threads so we can point back to the next time there’s an argument about it.
Ok, TED talk coming:
@napoleon1981 FTP is a measure of a physiological event. It changes daily. A couple percent is normal, a couple watts is within the margin of error on your power meter.
Your “fitness” can increase without touching where you cross the rise in blood lactate. RPE and other metrics can indeed be signs your ftp MIGHT have went up at least with newbie gains, but it’s in no way a sure thing. No one knows exactly how the TR brain works, but it’s clear they plot these things against their dataset.
The software is giving you it’s best modeled guess as you your ftp. Maybe it’s right, maybe it isn’t. Maybe your PM accuracy changed. Lot’s of variables. At the end of the day it’s trying to measure a very specific event happening in your body. You can verify it if you want, but do two ftp tests a week apart and you’ll get two different numbers as well. I know how this goes because I’ve been there.
If it feels close, keep going. If it’s the software getting more accurate that’ll show going forward. The AI dropped mine like 2% a couple weeks ago even though I know it’s “wrong”. (not lab tested) I had just done a KM test with some fatigue in my legs and I beat their number pretty easily. This was all inside. Outside I can probably go 3-5% over that even. This is all easily rectified by using the + and - buttons in TR or Zwift. Feel good? Go for 105%, feel crap? 95% that day. All good.
If you’re trying to specifically raise ftp, and it just won’t budge, you’re at a plateau. If that’s the case, congrats! You’re no longer a new cyclist.
As you get more well trained you’ll realize how specific your training has to get. AI doesn’t mean taking your hands off the wheel. It’s a tool to help with specificity IMO. Also, train now usually works pretty well for me if it’s a toss up on how I want my intervals served up.
All that to say, my perspective is as a cat 1/2 racer. Might be different than yours, that’s ok too. If you just want to see that ftp number go up and up and up, well, that train eventually stops. Unless you’re Pogi. If you are Pogi, well then, can I have your autograph? Wanna come help me coach NICA?
All in good spirits, this isn’t easy for any of us!
Again, im not concerned, im totally fine with it all, im going to test, i was going to test, the point was to build a huge aerobic base to build from. I also know this is where the ftp gains come from, also as a PhD scientist familiar with data the aiftp result made me smille bc it silly. I didnt come here for the help on my training, i just encountered issues with the product giving a clearly false trend.
And for the people holding their breath for the ramptest results, you are going to have to hold it for a bit, ill do it at the start of the build.
From this topic, I can see that TR cannot win.
I see that a rider’s monkey brain really like an FTP bump. Nice dopamine hit! Party on!!!
But if they get a PL bump, they think ‘oh shite, it’s going to get harder’. Monkey brain doesn’t think - ‘Awesome, I must have gotten stronger’.
And a minus -2 watt FTP decrease is like “WTF?” to riders when we all know that 2 or even 5 watts is noise.
My advice to TR:
Give way more +1-2 watts FTP fist bumps to people - the more the better. They’ll love it. Never give them less than a 5 watt decrease because it just demoralizes riders and a handful of watts is pretty meaningless anyway.
Seems like the aiftp was wrong. Seems like the base gains were also not tapped out like all my measurable data was suggesting. Settled? Or is a ramp test not a good medium to test ftp, thats a fun rabbit hole to go down on now… for reference early feb my ramp test was 267 with a 268 ftp. Right when the plan builder plan started. So its not that im a good ramp tester or that results were biased in the past between aiftp and ramp tests for me.
It is interesting that there is a lot of noise about AI FTP detection now on the forum, when it’s been very quiet about it since launch, considering how fundamental FTP is to training.
I can’t remember which episode, but on a recent podcast @Nate_Pearson and @Jonathan mentioned they were reworking AI FTP detection to no longer need 10 TrainerRoad workouts to work. Nate even went so far as to call it AI FTP Detection 2.0.
I wonder if a new model is rolling out that behaves differently from the old one.
Edit: here is the podcast. Great ftp discussion starting at 11:55 but the part I thought of starts at 22:50
Nice ramp test.
My guess is your one hour PR of 257w was probably dragging down the AIFTP.
The one hour PR of 257 was from last year June or July i believe. That was in the AIFTP all the way up to now and previously it wasnt dragging it down.
I wouldn’t trust a ramp test either. just because it gave you some more watts does not mean it is more accurate than what ai FTP said.
haha, see. I tested with the same ramp test 267 early feb, i completed everything to the T, power records, lower HR, lower RPE up to today, and now with the same test im at 280. I expected AI FTP to get me to 275. BTW tis was also un unrested ramp test. I think everything points to the ramp test being correct and not the AIFTP decline (even if its subtle). Its probably a hard pill to swallow for some of people here vigorously defending a blackbox algorithm with no transparency to how it works.
I am not defending either of them, there are several issues with using a ramp test to determine ftp and I stopped using TR well before the introduction of ai FTP.
There are plenty of threads discussing issues with the ramp test and why longer form testing is much more accurate if you want to dig into those.
fantastic….now you have a FTP you believe is accurate.
Go forth and train.
I think comparing a ramp test vs a ramp test is generally acceptable. Just like aiftp vs aiftp over time should be comparible. All im saying is the aiftp is not matching how i feel and actual real world data in identical testing approaches.
That again wasnt the point, the point was there is an issue with aiftp detection currently. Get it? Here is actual real world data supporting the point i was trying to make. Its not about the numbers. Aiftp if trending the wrong way over time gets people in the wrong power zones for training and this is a problem if true.
Anecdote vs. data.
A handful of people questioning AIFTP doesn’t mean there is an “issue” with AIFTP.
Again, if you feel you aren’t getting accurate numbers, there are ways to address it….as you apparently did.
Go forth and train.
