Given that you have access to a large set of data now, do you have any aspirations to look at building an AI FTP model similar to TR to give an estimated FTP based on sub-max efforts? Potentially could be a chargeable feature.
1 Like
Is this not essentially the eFTP number that intervals.icu already provides? I find it very close to TR AI FTP and WKO5 mFTP calculations
2 Likes
I don’t think so. eFTP still needs a max effort of a specific length (that can be adjusted?). AI FTP can be calculated even if you never do a max effort. On TR you can just go about your business with training and never do a test but still have it calculate an FTP.
Unless I misunderstand eFTP of course.
That is true ref eFTP but I would argue that TR needs some level of effort otherwise it wouldn’t be able to give you an FTP? Eg. if I just rode z2 then how would it be able to predict that my FTP is anything other than within the z2 ranges?
Like any of these models they need to be fed with a certain amount of data to provide an “accurate” estimate of FTP in my eyes. I like the TR AI FTP and I check it every so often but with a rounded training plan you are probably feeding the model in intervals/ WKO5 or TR with the data it needs for most of them to be pretty similar - this is my experience anyway.
2 Likes
I think AI is able to estimate your FTP based on, age, weight, maximum heart rate and heart rate for power data.
1 Like
Do you mean you think TR AI FTP works in this way?
Looks like they are using modeling to estimate FTP. Not much different to other programs out there. At the end of the day, it’s an estimate of FTP.
From their support page:
AI FTP Detection accounts for much more with a broad range of data points.
1: ie Monod Scherrer 2 Parameter CP, Morton’s 3 Parameter CP
Do I need to use a heart rate monitor to get an accurate FTP Prediction? Nope! All you need is consistent structured training with a power meter or VirtualPower.
I use a different power meter for my outdoor rides than for my indoor rides.
2 Likes
I think I found their PoC of the model a while back on GitHub: GitHub - jrkinley/ftp_proba: Predicting Functional Threshold Power (FTP)
Needs some fixes as libaries got updated but I had a version running with the example data that provided in the notebook.
2 Likes