Hey all,
I was looking for a thread to see if everyone was sharing the results of their AI FTP detection and couldn’t find one, so I thought I would make one myself!
I wanted to see how everyone was doing that was using the AI detection and to share their circumstances that led to their result. I will start.
After a horrible few months of surgery, ER trips, and chickenpox, I started training proper towards the end of January. I had a good block and then a rest week. Instead of doing a test, I let the system guess for me.
It said 319-333.
I was a little surprised as this seemed like a huge jump, especially considering my circumstances. However, after a week and a half of training at the new level, I don’t think it could be more accurate If advanced aliens from the center of the Sun could harvest my legs and scientifically figure out my FTP to the smallest degree.
All of my workouts have felt exactly like they should, and last weekend I did the Alp in zwift and averaged 340 watts for the 42ish minutes. This feels like a very realistic assessment of my FTP, as I could have extended my ride to about 50ish minutes, which is what I loosely consider my FTP time since my focus is 40K time trials, at 333 Watts.
What have you all found for your detection? Post either percent change positive or negative, or the raw numbers. Either way, I’m really curious to see what everyone else is finding.
On a related note, I used that figure of 333 for my beginnin. Everesting prep for this year and that wattage felt exactly right for what I was trying to do.
Anyone else having really good luck, or are there some horror stories? Lol?
Users are putting their results here:
Indeed, plenty of results and related thoughts there already. I am not sure we need a separate topic, but I let this roll yesterday to see how it went.
Yeah, there are some good results talk in that thread and I definitely browsed it. However, the results seem to get buried in the bickering about AT and other things (I had to scroll a bunch to even find some).
I wanted to make a thread that was nothing but results talk to leave the theory crafting and arguing to the other thread 
@mcneese.chad I’m fine if you want to merge or delete this, but I was hoping to get a thread that could bypass all the general talk and get straight to results and nothing but.
Sure, I guessed that was the goal here and that’s why I left it in place. We can let this continue and try to corral the cats to keep it purely to results without the tangents 
My initial ramp test in Nov 2019 was 208. Long story short injuries, recuperation and life got in the way of cycling. When I started trying individual TR workouts again they were too hard for me to complete at the 208 ftp. I manually lowered it to 180. Found it easier and then at that level was too easy so I decided I would slowly start bumping it up manually as I really didn’t want to do another ramp test. Eventually I manually made a big jump from 210 or 215 to 240. This seemed to give me workouts that felt right. Then I saw the AI detection features o I decided to let that happen. The system gave me an FTP of 249. So far those workouts, only a few, have been right on target as far as pushing me just that little bit more that leads to improvements.
For me the system is working well.
I’ll start with “Perhaps I’ve never been properly dedicated to structured training” but my experience:
I’ve used TR since 2011, off and on with various plans. I generally only do low volume riding due to running and swimming also throughout the week and probably never take a true off season. Added volume and plan compliance change a little in summer months with outdoor riding etc…
My FTP tested through 8 or 20 minutes way-back-when, then the ramp test were usually pretty consistent (or at least within noise margins). This includes switching back and forth between a few bikes and PMs over the years. Let’s say the total range over the years is 300 - 340 W (I’m a larger rider around 87 kg).
Still in all of this, I fear and struggle with High-end Sweetspot and proper Threshold work. I’ve always dreaded workouts with intervals over 15 minutes. Maybe weird for someone who aims to do long course triathlon - I’ve always assumed my threshold could be over estimated for this (especially since ramp testing and “enjoying” VO2 type intervals more).
My AI FTP experience - I half hoped it would take me down 5% to make these easier, but it did not. It bumped me 1 watt from 330 to 331.
Maybe it’s a vote of confidence and my ultimate issue with longer intervals is mental, proper workout fueling, bike fit, proper recovery, motivation or something else 
FWIW my progression levels if interested (I specifically added an endurance and tempo ride recently and bumped those, they weren’t being specifically trained as part of a plan).
AI FTP gave me 351 on Feb 9 after 6 weeks of build.
I’ve been doing high volume sustained power build since then, with no missed workouts. I failed one workout because I tried to do a sweet spot ride after a 2.5 hour hard outdoor ride with a buddy (including a 1000+ watt and several 700+ watt efforts). Other than that, I havent failed any workouts, and I haven’t even had to stop and rest (even on over unders).
I did a ramp test and it was 1W lower than the AI FTP prediction.
Folks, take that ramp test on the same day for a proper comparison.
AI FTP Detected 266w for me when I first used it and 270w last week. I had been training of 260w all last season (no test). My Nov 2020 test gave me an FTP of between 266w and 270w (284w for 20mins) but my last test in Feb 2021 only gave me between 240w and 254w (267w for 20mins) so I dropped right back to 246w ( it might have been pre my acceptance to AT and I didn’t want to overtrain and I think I had a slight injury too) and upped it gradually to 260w based on feel 
AI FTP detection was released just a few days before my plan had me set to perform a ramp test. I had just finished a “custom” base phase of 8 weeks. In that phase I did one or two workouts per week of 4x8 @ 105%. The other three rides of the week were all easy endurance stuff.
FTP detection set me to 354w which was a 12 watt bump. Since I was headed into a build phase I was a little unsure about that number. I’m almost three weeks into build right now and have had no problems with any of the workouts.
I did an unscheduled gravel time trial today that was approximately 24 min full gas. Monday, I have a ramp scheduled, which I don’t plan on doing. Will this AI FTP estimate look at an unscheduled outside race/tt to estimate my ftp or will only look at scheduled workouts (inside)?
@dustin - that question is best asked on the main FTP Detection thread linked to in the second post.
Here’s what I posted there (prior to this thread starting):
A (little) bit of FTP history. Generally I’d find I could push most workouts around 2-3% higher than I tested via the Ramp Test so after a couple of weeks would manually adjust the value. The next Ramp Test would usually sit a Watt or two either side of what I’d basically guessed. My last Ramp Test (cough) a long time ago put me at 272W. At the time Intervals.icu’s eFTP variously put me at 270-275W. Then last summer Intervals suddenly pushed my eFTP to 292W. When I returned to indoor training I decided to set my FTP on TR at 288W where it’s now been all winter. No particular reason, it was a sort of “split the difference” between my last tested and that particular eFTP but with a tilt to the latter.
Yesterday was to be a Ramp Test so I hit the prediction button and got 297W. Hmm, +/- 2% or so and it would have been a wash. I’ll go with this suggested value and see how things go.
On Early Access day AI-FTP gave me 314W. I’d been training at 308W as confirmed with long testing.
After riding with 314W a few weeks I had a heap of confidence in the AI-FTP, right around my true threshold.
A week ago I started SSBHV for the first time and AI-FTP bumped me up again to 323W. Definitely above my Lactate Threshold but given there’s no threshold work in SSBHV I am keen to work my way into it.
My PL’s changed almost perfectly for SS (10->8.9) and I presume the other areas too. They are completely messed up now following Saturday’s ride but.
I’ll be interested to see how it goes in the next two detection opportunities and whether it gives me the right combination of number and workouts.
This is AI is spot on accurate! 344: 90% of my most recent 20 min race/TT effort.!
No more ramp testing for me!
Awesome work!
Good god. After a horrible few months, you’re still putting up these numbers? I want whatever bionic surgery, and ER trips/chickenpox ailments gave you these superhuman powers.
On a serious note - The new API has been spot on for me. It challenges me “just enough” for the next block, to continue making progress. And I’ve always been a very poor ramp tester, having to manually adjust my ftp when workouts were simply too easy.
I’m not sure how accurate mine is.
I had a big bump from 340 to 356 AI FTP, then did a few workouts, some of which i failed but still got bumped again to 359. Tbh, i feel like my FTP is much closer to 340.
Anyone else feel like their AI FTP is over inflated?
- There have been at least a few people that feel the same as you, and they can be seen in the main topic.
At the start of my current block (General Build Low Volume), my AI FTP was 205 (had been 207 before the rest week). I decided to do the ramp test and it was 195 – less than the 198 I started with, which seemed hard to believe. I chose to split the difference and ignore both, sticking with 198 so the progression continued.