Thanks for these bullets. This is just the sort of thing I was thinking about . I have plenty of data for the robots to start with, but I’m questioning if the data I have gives a shows the breadth of my profile comprehensively. I think I’m going to punch out some hard efforts before pressing go on AI.
For the record: The AI engine has worked really well for me in the past. I just haven’ been on TR for six months at least, and for the past few months, my time on the bike has been limited to z2 and SS workouts in ERG mode. Happy holidays, everyone!
It’s 10, but that could just as much be to stop new people signing up using the AIFTP and then getting a refund, 10 (at least tries to) makes sure you use the system, and give it a fair crack … as opposed to the only reason being is to calibrate the AI
Yeah, I wonder. I’ve asked before how does AI determine an FTP if those 10 workouts are done in ERG mode exactly as prescribed. How does it know that they weren’t totally easy or that they were at the limit? At that time, TR said it doesn’t use HR, so how does it know?
It’s been decided that this AIFTP is a LLM ? and not GOFAI ? , also under the contraversation Strava TOC, didn’t they state that you couldn’t use strava data in a LLM ?
Not taking it the wrong way. For the past few months, I haven’t barely ridden anything above SS. When I did, it was in ERG mode. And, I haven’t done any FTP-like testing. Put it all together and I was beginning to wonder if there would be a datagap in my profile…
FTP is just another number amongst many you can and should use. Therefore, I don’t think it matters so long as you always use the same testing method(s). AI or ramp test or 20 minutes or dual 8 minutes or the true one hour time trial. Consistency is key, just like training.
I tried AI ftp testing after a wee while of sporadic training and it put me down at 323 watts, anyways I thought this was stupid and went and did a 20 minute test today and held 404 watts. So honestly it feels like it was just stupid. I had done a chainy the week before that anyone would’ve looked at and seen was higher than their guess.
I’m coming off a broken foot . Train now gives me today Anaerobic Mendel , 5.4 level and I’m now at 1.2.
Ouch . Probably could have done it but…..
I weenied out and did a sweet spot instead.
My new plan starts next week. I don’t want it to kill me the first week.
the ai sucks ass. i took a multi year pause from tr, rejoined and the estimated ftp was like 20 watts off a ramp test THAT I COMPLETED (manually) A FEW DAYS BEFORE RESTARTING TR!!!
How can the ai be any good if it looks at a recent workout where i just did the exact ramp test protocol and it gave me a FTP 20W higher than what my ramp test workout said.
If you took a multi-year pause, then the AI has very little to go with. In fact, people with very little recent training tend to have trouble digging deep and underperform a ramp test.
Ramp tests in your case aren’t great at closely estimating FTP, and there’s a good chance the AI is doing a better than you think. It’s seen your situation thousands of times.
I’d probably start with AI and see how it feels. With that much recent Z2/SS and some gaps, ramp can overshoot or undershoot depending on how you test. If workouts feel off after a week or two, then do a ramp to recalibrate.
The spread on the ramp test “ftp” result is rather wide. I think I posted an image showing this up thread.
Have you verified either result with a workout like 2x20? I would expect the “ai” ftp to be more in the ballpark than a ramp test if given sufficient data but it’s not a bad idea to confirm if you have doubts.
I think now that I have 3 weeks of TR easy workouts on my healed foot TR is trying to see exactly where I’m at. Maybe it’s trying to set my FTP . Yesterday I was training stress is high recommended endurance. Did that. Today I get a 8.5 anaerobic - Thomas . I’m at 1.1 on anaerobic level. This is going to kill me.
