Trainerroad AI FTP Detection

I see TrainerRoad introduced automatic FTP detection last night…

I subscribe to TR as well as Xert and it gave me the EXACT same number as my xert TP. To say I was surprised would be an understatement. I wonder if they’re using the same number crunching method as xert?

I’m really intrigued by it too. Based on what TR say, they’re not basing the estimation on capacitive efforts at all. They’re looking at your recent history, against others your age, and I spose a mess of other data. It’s ironically what I thought Xert did way back when I started, before I understood it better.

I’m really tempted to finally try TR longer than a few days and see how it works. The estimation is amazing, if it works, and I like the idea of having a plan laid out. I would keep Xert and run them in parallel.


Setting Signature Decay Method to No Decay is FTP “detection”.


They’re using a machine learning algorithm. It’s actually pretty easy to do this and get good results from it. The main issues is that is a black box so they can’t go backwards, i.e. “I want this FTP. What do I need to do to get there?” It’s just one way and opaque to people as to what it’s doing.

We don’t use an ML algorithm. We’ve modeled fatigue and MPA. We use your signature not just to understand fitness changes but to also predict exact moments in a ride where you can longer continue.


Meh, went back and tried TR today, and it didn’t take me long to realize/remember why it won’t work. The FTP estimation is still one dimensional; you don’t know anything about your other fitness parameters. Also, TR (and other systems) can’t quantify work, except as TSS, and that’s to say nothing of Difficulty and other stats. I was really taking all those for granted.