AI FTP detection vs Ramp Test vs. 20 min test

Science denial? :man_shrugging: pick one or the other. I’m a fan of ftp, I’ve got some issues with critical power. Start believing you can easily determine the narrow range of power (~10W) that puts you between doing long 1-3 hour upper tempo efforts and “my HR and breathing keep ramping up, and now several minutes later I can’t breathe and I’m going to blow up now.” It’s not rocket science. You can learn what it feels like to ride around the tipping point. The amount of time that you can hold ftp can be trained, and depends on your fitness and mental state to name two things. It’s only murky if you don’t understand some basic concepts.

1 Like

Same, aiFTP about +10 watts, which makes a massive difference if it is really around your threshold.

1 Like

Functional Threshold Pastry?


i’ve come to realise that for “training” purposes, FTP only matters for efforts in and around that intensity. by training i mean, actually trying to move the fitness needle forward.

FTP (the number) does not matter for recovery rides, endurance rides, tempo, Vo2 max, sprint efforts. does not matter, is irrelevant, and should be ignored. these should be based on feel, breathing rate, duration and the knowledge that adaptive signals are not always based on more watts.
FTP (the number) matters for calibrating longer, sustained efforts of mid to high intensity, where you want to increase your TTE. knowing how to pace 4 x 10, 4 x 15 etc properly to achieve stimulus without creating crippling fatigue requires a well honed FTP based on aerobic power. not the biggest number.
just ride a long effort and don’t rely on the crutch of AI, ramp, 2 x 8 or 95%. and don’t chase a new FTP every 4 weeks unless you are a noob.


Yeah, I would agree with that.

I dont really use FTP anymore. I only consider it as a starting point for threshold intervals, it soon becomes clear if you are over or under.

Z1, Z2 heart rate.

Tempo, long race watts / pace / HR under threshold HR
VO2max, best effort for the intervals.

Sprint… max.


Really, so the 1,001 posts on here where no one can agree on the definition of FTP is because everyone is just denying the science…or is it because the waters are murky?

Its all just fluff.

1 Like

Really. Not sure what happened to reading comprehension and STEM education. I tend to blame the Internet and mobile devices :man_shrugging:


So you’re saying that the 1,001 posts that cant agree on FTP are because no one can grasp the science or because its a little murky?

You’re saying that. I’m saying that I’ve clearly and repeatedly read there is a border between stable and unstable cycling intensity, you can field test it, and it’s called FTP. And it’s not defined by an exact timeframe (1 hour), rather it’s a time limited state at the top of stable physiology, just under unstable physiology.


I just realised I’ve become the exact thing I hate…part of the derailing of a thread into an FTP discussion. I feel dirty and disgusting. Sorry OP.


To bend the topic back to the op. :smiley: I think there is a gap between trainer driven efforts and real world efforts. Some people can really slam a hill based ftp test but die on a flat road. I know for me the wind is an absolute destroyer of my will. :smiley: The subtle shift of physical position, the mental strain of a race. Maybe heat dissipation is an issue, a too hot aero helmet?

I think it’s hard to compare across modalities, and a model that is based upon the ‘majority’ of your efforts will likely give you an answer that tries to match a ride that is of the ‘average’ type.

@mgrigat TR AI FTP and Xert and other analytical tools are using models. Sometimes models are wrong, in particular due to the efforts you are feeding the model. For example in recent years the Xert model consistently over-estimates my short power (1-10 second power), and like a see-saw, Xert under-estimates my threshold power. And depending on the efforts I feed WKO, Golden Cheetah, Strava, etc., the same applies. When I feed good data to the models, they do a good job estimating FTP.

Assuming your TTs were near-max efforts, those represent good data for estimating your FTP. And for various reasons, the models are wrong.

Yeah the AI FTP detection is somewhat dubious. I have literally done nothing for the last 4+ weeks and the AI detection showed an increase in my FTP. I don’t think so…

Welcome back to the forum. Not sure what your agenda is. However, it looks like your FTP went down 13.7%. Even if it were true, which it appears it is not, that would have to be an error for TR support to look at. I get that AIFTP is not accurate for everyone, but just making up stories isn’t helping anyone.


But isnt that set FTP not aiFTP?
I dont know but maybe they rejected aiFTP and put in their own selected FTP?

You would see similar on my account but your assumption / accusation would be incorrect.

  • Yeah, that pic seems to show a manual entry.

Here is what my AIFTPD one looks like from this week.


While I agree that there are worthwhile follow up questions to learn more from that user and what happened in their instance, I think it’s a bit much to jump to an accusation implying malice.

I do think having TR review the case is a good idea at the very least.

Per the calendar entry you dug up, it’s likely a manual entry done after discarding the AIFTPD offering, or they could have deleted it manually after accepting it. Definitely more than one way to get to that entry so making assumptions is not a great start.

Thought so but wasnt 100% sure.

1 Like

I had something close. After my break period, and feeling that my FTP was naturally going down, AIFTP put me at 292 if I’m not mistaken. I barely can hold 280.

And, by the way, I follow “WKO protocol”, 1min, 5min, so on and on. And my mFTP was 277. My eFTP in itervals 280. I felt that 280 is definitely acurrate for the moment.

Did you check those results with TR to see if they could explain the results?