While I did not quote your entire post, I quoted you directly without any edits so no misquoting, and I don’t see any corner that I’ve backed myself into, I’d say it’s the other way around.
You made an assertion that is flat out incorrect in some scenarios where pretty much every coach in the world would disagree with you without some serious caveats. FTP and MLSS are not the same as cycling fitness, period. And there are scenarios where no, they are not a good enough proxy either. And saying that not accepting an AIFTP number is pandering to TSS Trackers while maybe a slight joke shows either a lack of knowledge of training concepts or choosing to look at one side of the issue and not the other. While it’s true of the reverse, there are scenarios where NOT accepting the number is a good thing and accepting it is a bad thing.
The AI is finding paths to better performance in some scenarios, but worse in others. That’s what I’m saying so no, we are not arguing the same thing.
I understand the point you’re making, but a 35-watt drop (around 14%) seems excessive. My training and race data don’t suggest such a large decrease in fitness.
For example, I’ve done long rides with ~190 NP (5 hours), which would be about 90% IF if my FTP were really 212 W. That doesn’t seem physiologically realistic for several hours of riding.
So it feels more like the new model may be overcorrecting rather than reflecting actual fitness.
the day the new trainerroad ai hit, i happened, by coincidence, to have done a 20 minute ftp test. the 20 minute test gave me an estimated ftp of 177, which was the number that i was already using and have been using for a while. i am accustomed to the zones created by this ftp and a ftp of 170-177 felt right to me. when i logged into the trainerroad app after this test, the new ai gave me a new ftp of 163, which i did not accept. 4 weeks out from this day, the trainerroad app ai gave me a new ftp of 171. again, i kept 177. 2 weeks later, i did a zwift race lasting 53 minutes with a steady ave watts of 180. zwift told me that my ftp was 180; i kept my ftp at 177, a number that feels right. none of these numbers are that different, but the 163 number throws off my tempo and threshold zones.
Current AIFTP is 215 and projected to be 230 on 3/18. I haven’t done any other test, but I can hold 150-160 watts for an hour at a HR of 125-130 bpm, which lands solidly in the endurance zone. A recent 90 minute overall 4x9-minute threshold workout felt hard with HR climbing to 175-180 bpm during the efforts, so overall, I think the AIFTP number is accurate enough for me. What matters most for me is that I’m getting the perfect workout each time. I’m not over-trained, and I’ve remained injury/pain free for 6 months since starting TrainerRoad.
In the sense of the post you tagged me in, tte does matter. We were taking about the success rate of a certain workout with ai ftp.
As a triathlete, tte should definitely matter. There’s a ton of science out there about fuel substrate usage based on training history is a major contributor to tte. Running off the bike, after being tr trained was just horrible for me.
Conversely, tr now seems to think my workouts are not actually improving my fitness. Suggesting I would fail a workout i had just completed 2 weeks prior, but now the target is 5 ways higher. I had no failures in that block and most workouts in the high 4 to mid 5 range.
That was a thread created to reveal ftp is not hour power, by actual results.
I’m not particularly interested in time to exhaustion, never have been. Running after TR workouts has always been much more fun than running cold for me. So we’re different in those respects and I’m fine with that. My problem with running, is running itself, sadly!
Now, if you could give me an AIFTP for running, and adaptive, progressive overload AI workouts, then I’d be over the moon. Not sitting in bed trying not to get my shoes on for Parkrun right now…
I would be wary of using NP to try to assess ones FTP. I think average power over a period gives a much better indication. Ideally you want average power and normalised power to be close to each other. A sustained strady state effort.
