FTP Estimate Confusion

Training
1

Hi TrainerRoad,

I’ve been having some wild swings in my FTP estimates from different sources and I’m not sure if my training is working and whether I’m training productively.

History:
I first started using TrainerRoad in early 2024 and used it until May or so, then took a break from it for the summer. Signed up again December 13, 2024. Since then I’ve had wildly inconsistent FTP estimates:

  • Dec 3, 2024: Took a ramp test on Zwift: 260W
  • Dec 13, 2024: Started TR training plan, increase FTP. TR AI Estimate: 281W
  • Dec 20, 2024: I failed a workout, TR adjusts FTP down to 275W
  • Jan 27, 2025: TR AI Estimate: 280W
  • Feb 5, 2025: TR Ramp test: 247W
  • Feb 5, 2025: Manually adjust FTP to 260W because excessively low and I was recovering from a cold
  • Feb 10, 2025: Start a new training plan, increasing FTP

Most recently, my workouts have been feeling a bit too easy. Yesterday I had a VO2max workout and I already rated it moderate.

So basically, I’m confused about where my true FTP is really at at the moment and whether my training is working at all. AI estimates always seemed to be a bit on the high side to the point that I was failing workouts. I’m inclined to discount the recent ramp test because I wasn’t feeling optimal around that time but it’s leaving me unsure about where to set my FTP for optimal training intensity.

On the other hand, in a real 1h threshold effort like the Alpe du Zwift my recent best was 246W on Dec 6 and only 222W on Feb 12 (didn’t ride the whole segment all-out after I blew up a bit on the first three bends).

1 Like
2

Pick a method (preferably the classic test or other long form test)and apply it consistently to gauge progress. Eventually AIFTP will produce more accurate results given sufficient training history.

2 Likes
3

Hey @hannesbrt, welcome to the forum! :partying_face:

It looks like the recent Ramp Test on February 5th is the major outlier here. Ramp Tests aren’t as good at getting an overall estimate of your FTP in a general sense if that makes sense. They’re better at taking a small snapshot of your fitness at one specific moment and also can be skewed based on whether your unique physiology tests well in a ramp scenario or not.

This is why we like AI FTP Detection.

Maybe that first AI Detection on December 13th was a bit high, but you showed that within 4-6 weeks of training, you were at that level. Sometimes it takes a little bit to get things dialed in perfectly for new users, and Progression/Workout Levels help to even things out along the way.

That February 5th Ramp Test came 11 days after your most recent hard workout, so while you may have been well-rested, it’s possible that the sharpness of your fitness had decreased quite a bit by that point. As a matter of fact, the week before that Ramp Test you had 48 TSS, and the week of you had 69. Additionally, you were still recovering from illness which further exaggerated the results.

The fact that your workouts now feel a bit too easy backs up my theory that your FTP is higher than what that last Ramp Test suggests. It seems like you (similar to myself) might not test super well with Ramp Tests, which is totally okay – there are other options. Again, that’s why we like AI FTP Detection. :grin:

Once your FTP is set and you’re consistent with your training, AI FTP Detection does a much better job of following your fitness than periodic Ramp Tests.

Similarly, it can be superior to other fitness tests like what you were going on on Alpe du Zwift because pacing is so critical but can be difficult to manage in those longer efforts.

I think that a good starting place to put your FTP right now is probably around 270. If you’re curious to test this further attempting a workout like Lola could be a good option since ERG mode will help you pace things properly and you can always adjust the intensity up or down slightly (only do this if you really need to though).

As @ingluis29 mentioned, moving forward, it’s really best to find the FTP estimator that works best for you and stick with it. Alternating between 3 or more methods will certainly give a variety of results, and what matters most is consistency and the proper difficulty of your workouts during training.

Let me know if this helps and if you have any other questions about this.

Good luck & happy training!

2 Likes
4

Rule of thumb - the only types of workout that are useful for calibrating your FTP are (best) Threshold and (OK) Sweetspot. Anything else is too far away from your FTP to tell you very much.

4 Likes
5

This might be derailing the convo a bit, but I have a question. What is AI FTP detection actually estimating?

Is it directly trying to estimate “the highest power you can maintain in a quasi steady state without fatiguing” (my understanding is this is actual FTP), or is it estimating a best case scenario for what result I could get on a ramp test if I was fully able to express my abilities?

I remember reading back when AI detection was released that it was validated on athletes’ training, then comparing what they got on ftp tests with the AI detection. This makes me wonder if AI FTP detection is trained to estimate an estimate (ramp test, or even 20 min test), rather than actual FTP.

While the ramp test is a great training estimator (repeatability, easy to do on erg, etc), it certainly isn’t FTP itself and assumes a lot about an athlete and their location on a bell curve (big caveat that I don’t really know anything and this is stuff I’ve read over the years lol)

6

Makes a lot of sense — thank you for that answer.

I started doing ramp tests after previously doing 20min tests because of the difficulty to pace them well. I like the idea of the AI FTP estimates, but admittedly, getting the results that I have been getting, I don’t trust them entirely yet and therefore throw in a real FTP test once in a while.

I’ll set my FTP to 270W for now and see how it feels. I can do another AI estimate in three days from now and let myself get surprised by whatever comes out then.

One suspicion I have for why my Dec 13 estimate was too high is that I had a ride from Oct 18 in my Strava history on which my power meter was malfunctioning and reading >2000W. I’ve excluded that ride from adaptive training since, but maybe it was biasing the results before.

7

I used to struggle with the 20 minute test. Now I do the longer form Kolie Moore test. It’s way more reliable and way easier to pace.

You could confirm your 270. I just use the baseline test. By 30 minutes into the test I know I can “feel” my FTP, meaning that if I go a little bit harder I wouldn’t make it very far but if I back off a bit I could go for a significant amount of time - “about an hour”. I usually just do 30-40 minutes because by then my FTP will have been sussed out. I make no effort to game the test like I used to do with the 20 minute test. I mean, everybody goes all out for the last few minutes of the 20 minute test, right?

Once you start doing a long form test, it just becomes a good FTP workout.

3 Likes
8

Like the 20 minute test, ramp test, it is an estimate of your FTP. In my experience a lot better estimate than either of those if you’re giving it good data. One that is even better though is a long-form test like the Kolie Moore tests where you can get to the point where you can feel it. I can ride for 20-30 minutes and tell you what my FTP is on that day, and it’s just a threshold workout.

Seriously though, let’s not get into the actual definition of FTP, that blows up every thread it’s talked about on. What is important about it is a basis to benchmark training zones, measure progress, etc.

1 Like