Hi TrainerRoad Team,
first of all, I want to thank you for continuously pushing new ideas and building such a thoughtful training platform. As an AI developer and researcher myself, I truly appreciate how challenging it can be to introduce new concepts - so please keep up the great work!
I’ve been using TrainerRoad for quite a while now and I absolutely love it. Recently, I made some significant improvements to my sleep and recovery, and I realized that I had been under-fueling my workouts more than I thought. I know this is something you’ve emphasized for years - it just took me some time to fully implement it.
Since making these changes, my performance has improved a lot, which I’m really happy about. However, after planning my next training block, the AI-FTP Detection predicted a rather large increase (around +20%). My impression is that my recent data might be a bit “out of distribution,” and the model seems somewhat sensitive to it. I assume the system will adjust over the coming weeks and select appropriate workouts, but the initial prediction might raise some questions.
So I wanted to ask: is there anything I could do differently on my end? I’m not using Plan Builder because I prefer creating my own training structure - could that make my data less consistent for the model? If so, how could I better support the system with my current workflow?
Has anyone else experienced something similar?
Should I just accept my fate and go pro - or at least try not to die trying? 
Thanks again for all the work you put into TrainerRoad!
Best regards
