I wouldn’t question the accuracy of their statistics, but statistics can show a system to be very good “on average” while the system may not be good for a significant percentage of users. I’m not saying TR is trying to be misleading or that the system isn’t great for a bunch of people, but of course they are going to pull together statistics that tell a good story. My AIFTP runs 20+ watts higher than the prior AIFTP version (which aligned very closely with physiological FTP). I like lots of things about the new system, but the new AIFTP works very poorly for me and how I train. If it’s working great for 90+% of the user base, that might be good enough for TR to call it a win.
And while I’m totally on board with consistency being a primary driver of performance, “success rate of workouts” is not a proxy for performance improvement. I could absolutely lay out a plan that almost guarantees “success” in ever workout and result in a performance decrease. TR definitely prioritizes workout “success” over finding limits and I’m not saying that’s the wrong approach for the majority of users. I can appreciate how that gives many people (especially beginners) a positive experience with the motivation that comes with it. But I’m personally not a fan of flagging workouts as successful or failures based on whether it went exactly to script. You can absolutely have productive workouts that are flagged as failures and unproductive workouts that went perfectly.
I believe TR when they say that workout success rate is higher in 2.0 but, completion of workouts alone is not that important imo.
It’s fairly easy to assign workouts that the athlete will almost always be able to complete, but that does not mean that those workouts will be driving appropriate physiological adaptations.
I’ve done multiple lactate/vo2max lab tests at the local university and I largely agree with this statement. And I think it’s more interesting to look at results over time rather than an isolated test. The one interesting thing it can tell you about your physiology is looking at how training is affecting things like efficiency over time.
It’s been around 5 years since I last had a lab test and I’m curious how much my vo2max has dropped in that time (as I approach my late 50’s) despite my performance ticking up slightly over the years. I was already on the high side of the efficiency curve with a decent FTP as a percentage of my somewhat pedestrian vo2max. Honestly, if the test wasn’t such a brutal effort, I’d go do another one just to see what’s going on with my body in the last ~5 years. It’s bad enough doing an all out effort like that, the face mask makes it much worse for me, almost claustrophobic or drowning sensation.
Tr says that, but those workouts are categorized in to buckets that are based on ftp. Interval length, rest intervals, total interval durations based on this same number. They were created that way for a reason.
It does do that. TR AI predicts the difficulty and it doesn’t just suggest easy workouts. Also, in my experience it hits the right energy systems. Sweet spot workouts feel like sweet spot and very much unlike threshold.
Thats a pretty wild claim. There were plenty of people in the Beta saying their FTP was too high or too low and therefore their workouts felt inappropriate for the zone.
There was plenty of discussion about FTP, what it means, AI FTP vs. etc. I’m talking about letting TR AI base its training off one’s AI FTP. Not whether it agrees with other methods to determine FTP, It does not for me.
That’s great that it works for you. But In my experience it does not. I was in the beta too and have been monitoring the number since the beginning. It’s been varying degrees of too high the whole time for me to the point that it changes my zones.
Seriously, why can’t you just acknowledge that it’s not perfect and it doesn’t work for everyone. Your n=1 and quoting TR stats doesn’t change any of this and it’s getting tiresome.
@grwoolf’s point as I understand it was that AI FTP tends to overestimate FTP, and that the shift is so significant that it e. g. makes sweet spot into threshold workouts, etc.
My claim was that according to TR’s statistics, training with AI FTP is an improvement for the vast majority of athletes. My experience (not just me, but many other participants) starting from the closed beta bears that out. And that I don’t think it is warranted to make a blanket suggestion of nudging down whatever value AI FTP came up with by a few percentage points.
There are always outliers, though, and that might include you and others. Even if, hypothetically, it had an accuracy of 95 %, if TR had a user base of 50,000 athletes, it’d get it wrong for 2,500 athletes, which is a sizable number.
Yeah, but I’m not saying that’s the case for everyone. Clearly, there are people who feel it’s too low and people who feel it’s too high. To the point that workouts are not matching their stated intent.
Even if it’s working great for the vast majority (which could make the statistics look great), that doesn’t mean it’s working well for everyone. Whether those outliers are 1% or 10% of the population, I have no idea. I just know I’m not in the group that it’s working well for and there are others experiencing the same.
And even if it’s only off for a small percentage that can see be a large number of people. That works both ways, just because there are a lot of post with people complaining too high or too low that doesn’t mean it’s an inaccurate measurement. But also even if it is a very accurate measurement it might not be right for everyone. Just like some people get more or less accurate results on ramp or other tests.
He’s not talking about too easy/too hard. It’s about correctness of AIFTP driving the appropriate zone.
If you’re supposed to get a threshold workout, and it ends up being VO2 instead, it can still be appropriately difficult for a VO2 workout, it’s just not the intended zone.