I’ll clarify, they are not selling you the right to see the details and inner workings of their model. (Almost no companies do in scenarios like this.) They are selling you access to their training platform, and the right to use (or not use) the results of the AIFTP model as part of it. In fact, from an organizational value and IP perspective, they’d have to be CRAZY to just tell everyone all the details of how it works.
Don’t get me wrong, from an intellectual curiosity perspective, I’d love to know more because I do a lot of work with data and analytics professionally too, there’s just zero expectation that they would or should do that.
I’m quite aware I’m not purchasing the model when I speak of TR and selling the model. Just let us know what the model IS. I have a feeling that the cycling world doesn’t share technical details because what they are doing is not special at all, any they’re protecting IP that’s largely based on public models. My opinion, obviously.
You and I have different experiences and expectations. I don’t expect them to share their secret sauce. But I’m used to using software in the $$$$ range that have voluminous technical manuals and the technical manuals break down every model included in the software for audit. The performance of my structure depends on it.
But this is cycling software. No real consequence.
I definitely disagree, I don’t think we have any right to know what the model is, and shouldn’t have that expectation. If it is based on a public model, or not, it doesn’t matter. It’s just an excuse to complain for some people.
This isn’t mission critical business platform, engineering software, or running NASA, it’s a $20/month cycling training platform.
I think of FTP as a tool to determine the intensity of workouts. If the workouts feel right, it makes no difference to me if it’s quantified in watts or almond croissants.
It has been mentioned on podcasts and in the support articles:
You need accurate weight, age and at least 10 completed rides in TR. It has been implied that these data points as well as data from completed rides, you are then placed on the “curve” and given a FTP (based on data from millions of riders). As stated in posts above most are very close but as with anything there will be a few outliers.
Think about it… hundreds of thousands of users, millions of activities, all collecting data such as power, heart rate, etc. Add in weight and age and you are probably going to be within the standard deviation of that plot on the graph.
Which definition of FTP is everyone working to? No, they’re not all the same. And here we go, yet another thread falling into the FTP discussion
Within TR - FTP - Functional TRAINING Power - doesn’t necessarily have to align exactly with your FTP - Functional Threshold Power (depending on your definition (Oh god, here we go).
Again, none of that really helps. They’re using population data to predict individual performance. That’s like using 220-age for max heart rate.
I’ve thought about it. I have. That’s why I have questions. You say I would be “within the standard deviation “. What’s the standard deviation for TRs model?
Dragging this back to the original question, I don’t think it really matters. There is no definition of FTP where it’s a number that you can’t hold for 20 minutes. Especially for a TTer who is focused on producing exactly those sort of sustained threshold efforts and has a whole seasons worth of data containing multiple maximal and race efforts.
So either AI FTP in this case is off by maybe 10% (on the basis that the best 20 minute effort was ~96% of FTP where on the old 20 minute model you’d expect it to be 105% of FTP). Or there’s some power meter or calibration inconsistencies messing up the data.
Does seem pretty odd - could maybe see how you could throw AI FTP off if you did lots of shorter VO2 efforts but no maximal sustained efforts. But TTs and Zwift races are great data for estimating FTP accurately. Maybe HR bands are set way off so it’s not recognising those efforts as being maximal and thinks they’re sweet spot or something?
Who knows… I am just guessing here like everyone else. It is probably more like growth models when you take your kid to the doctor. But I get it… you are being obtuse on purpose.
For me FTP is just a training number. It is closer than my running days when we ran at whatever coach told us, then completely ignored it and went faster. Curious why this matters so much to you?
I’m actually just asking reasonable questions about technical details of something. You’re speculating on things. We’re not having the same conversation.
When is an implication (and subtle snark) speculation?
I’d suggest a direct email to TR as the best chance of success vs a PS level comment in a topic that TR may or may not ever read. Go straight to the source, but I wouldn’t hold my breath on a detailed answer that seems you’d like.
i suspect (based on some comments by TR people) that some of the AIFTP requires on user input, and not purely based on power curve analysis. Not sure how to test other than to have a new account, set FTP low and do a bunch of workouts and then get some other data with higher power to see if the AI FTP just incrementally increases the ftp recommendation or if it considers a higher power file (hope this makes sense!). my opinion is they analyze workout levels accomplished and if you complete certain workout levels with a given ftp they extrapolate to suggest you can do lower ones at a higher ftp. again, this is all conjecture on my part!
Its only murky if you ignore the science. There is a border between being in a physiological state of stable and unstable exercise. Some back the Critical Power concept, and some back Functional Threshold Power concept.
My personal summary of the science is that in either model, someone well trained for TTs, with a sufficient warmup and some level of freshness, should be able to generate power around that border for 30-40 minutes.
In the distant past when I first started interval training on the TT bike my FTP was bit higher than what I was sustaining regularly on a TT (with the odd exception). Then I realised in my 20 min tests and intervals I would raise out of the saddle when the going got tough, and often stand in the TT position and hover over the virtual nose of the saddle. I don’t do either outdoors. Concentrating on holding the outdoors position has resulted in lower FTPs (closer to my outdoors power) but I am going faster.
