Does anyone know if AI FTP detection can be trusted if I’ve transitioned from more structure and 80-90% plan adherence to less structure and a mix of group rides and train now workouts? This change happened approx 3-4wks ago.
I suppose I’m asking what data gets included in the AI FTP detection modelling and if we know it to be less accurate if not following a plan with high adherence rates.
AI FTP Detection is not really dependent on a training plan from anything I’ve read. It reads TR workouts but also any other cycling ride, race and workouts imported to your TR Calendar. It’s actually possible to use and get decent results without any TR workouts.
Questioning results because increase in modelled FTP not expected given sporadic riding and a few failed workouts recently (with completed workout surveys showing “I did not pass” and “intensity”)
In that case, contacting TR support is probably best.
I have transitioned almost exclusively to outdoor rides at this point…and so far I would say AI FTP Detection has kept up with my efforts. Just updated yesterday and got another 5 watts, which is about what I would have expected.
It also, IMO, bumped up my FTP appropriately after a week of outdoor endurance rides during Spring Break, etc.
I’ve been using it all year and it has bumped me up ~15% total in FTP and each increase has felt “right”.
I use TrainNow exclusively (I do not follow a plan), and this year have ridden outside a total of 6 times. AI FTP detection most certainly works and has been accurate for me so far.
Simple version from some recent posts from TR CEO Nate:
- power data
- historical power data
and then it calculates FTP probability across a range of watts. For example it may give highest probability (likelihood) that your FTP is 250W, but there is still a 13% chance its 265W (from that link above).
Estimates can improve if you feed it power data with more efforts near your current fitness level.
As an aside, in a similar manner you can estimate FTP from 60-90 day power data just by looking for the ‘knee of the curve’ on a power duration curve, or ‘drop’ in power histogram. Knowledge of the efforts including in the data set, and previous ftp results, can help guide the results of your visual analysis.
Long and short, firstname.lastname@example.org can take a look and better comment on the specifics of your data and AI FTP estimate.