My bike is stuck in airport chaos and I am going to try and keep workouts going on my girlfriend’s bike with her kickr core, which raises some questions for me:
Normally, I use my power meter pedals for everything inside and outside. If I want to continue my training plan now, presumably I have to do a ramp test because I’ll rely on her kickr core. If my FTP goes up, progression levels will be screwed up by being set way down, right?
Now let’s say I get lucky and my bike is delivered in 3-4 days, I could set my FTP back to the value I trained with before, but again, progression levels will be set down.
More importantly, does this variation in power data have medium- and long-term effects on AI FTP Detection?
Finally, how likely is it that a well-calibrated kickr core is going to deviate from my pm so much that I have to retest at all? Does this vary by zone? (For example, today there’s a long zone 2 ride on the calendar; presumably the tolerances for an off-ftp would be much bigger for zone 2, so should I wait to test until a harder workout is in the cards hoping that I might have my bike by then?)
You could ask TR support and get a more definitive answer but I think I read somewhere that their AI Model would cope, dont sweat about it, save that for the KICKr
No need to change anything or test. I would use whatever power source you have as a guide and rely on your RPE. If it feel a little too easy/hard just adjust the % of each workout.
Does AT use the model/manufacturer of your power source as an input vector to the ML model?
If yes, I’m confident that “AT will take care” as you vaguely point out.
But if the answer is no, I’m a bit skeptical with AT and different power sources and therefore would want to bring different power sources to closely align.
Would be helpful and nice of you to get a concrete answer to that specific question.
And if it did, what would you want it to do with this info? Weight the inputs differently? Only use inputs from 1 source? Always discount “X” model power meter data because it reads high? I am just not sure how assumptions about this info could be used correctly or predictably.
Yes. Model should learn that for example power source X matches power source Y +6% (relative) or 15 watt (absolute) and take that into consideration for workout success/compliance. Obviously it gets tricky when putting out a single value for FTP which then might be enhanced with hints regarding power sources.
I just see it as way too many variables and almost impossible to identify a stable relationship between power source X and power source Y without controlled testing as someone like GPLama would do. Real world example, I have a Kickr and 5 different bikes with power meters providing input to my training. Without extensive controlled testing, no way any model is going to be able to accurately determine relationships between all. Luckily they all seem to read similar enough and I don’t sweat the small stuff. Which is what I would recommend for the OP.
Don’t mess with anything. Just do a spindown on the Wahoo then ride.
Unless there’s something seriously wrong with her trainer or your PM, the small differences in power readings mean nothing. Your body doesn’t care about a 1-5% difference. Power zones aren’t discrete, they are a continuum.
@cyclhist great question, it’s normal to worry if a different setup could be impacting your training experience. @MI-XC is correct that you don’t need to change anything or re-test for a new FTP result, but what you should do is make sure your devices (both your temporary setup and the setup you return to) are being calibrated so that Adaptive Training can lock you in.
Please don’t hesitate to reach out if you have any questions about that or need a hand with calibration as you move through training. We’re here to help!
I don’t want to share the inner workings of our model. We’re a small, bootstrapped team going against big competitors with bigger funding. While we may not have their resources, we have an awesome team that is building something truly unique, so we have to protect that.
What I can say is we didn’t build this feature devoid of the common context our athletes face in their day-to-day training. As of now, we’re confident Adaptive Training and AI FTP Detection will do a good job of handling most situations, and we are certain it will constantly improve.
I hope you can respect our need to be careful with sharing sensitive details of our product!
