AI FTP Detection overestimating

I do hope it’s not fatal to users.

No it was still off. I believe but would need to check eftp in intervals for me was based on power at a shorter interval of something like 5 minutes. Currently they are 13 watts different. My 30 minute rides power is 10 watts over the eftp

My experience was similar and I posted something to that effect on the AI FTP announcement thread. I did a 40 minute Baseline TTE test and confirmed AI FTP was overestimating by about 6-7%. I chalked it up to the fact that I’ve been doing unstructured outdoor riding pretty much exclusively for the past 6 months, but TR Support responded suggesting I make sure to fill out my survey responses for my unstructured rides.

It was news to me that survey responses had any effect for unstructured rides since these don’t impact progression levels but I’ve been filling out my survey responses since and my most recent AI FTP detection was pretty much bang on with my most recent test.

I don’t know if this will solve the issue for everyone, but it could be something to try as it seems to have helped in my case.

5 Likes

Yeah, I guess I didn’t take the survey responses as seriously for unstructured rides in the past, but I usually fill them in. I guess it’s also important how I fill them in. At first (this has adjusted over time), I would answer relative to how the work out should feel (ie: if it was a stretch threshold workout and I felt like my heart was going to beat out of my chest when I was done, but I was able to finish, I would rate it moderate because that’s how I thought the workout would go). Since then I’ve adapted my responses to reflect how the workout actually went (the above example would get a hard or very hard rating now).

I don’t think I appreciated, until now, just how much weight these responses have.

This was the point I was trying to make previously. He had two data points, then went out to ride and see how accurate it was to validate either number.

The OP and my experience was that AiFTP was over-estimating and for others it might be under-estimating. so you just go and validate it in some way. doesn’t mean you need to do another FTP test to validate it - just go and see if the your threshold interval feels like threshold.

1 Like

I havent always looked to know if I can make that statement.

I don’t see the AIFTP necessarily as the power I can do for an hour or so. The AIFTP is whatever power level that makes the workouts at my current PL hard but doable. Only if my PL is high would I expect to be able to do the AIFTP level for 45-60 min.

2 Likes

Yes, this is true. The number AIFTPD is giving you will make the workouts hard but do-able for your progression in each zone. And those training for TTs or long climbs should be the ones concerned with raising their PL in the threshold zone to the point where they could hold this power for long durations.

But FTP should be a level at which your other training zones are based off of. These training zones are used because they trigger varying adaptations in your physiology. If you’re trying to train Threshold, and your FTP is incorrect, you could end up training VO2max or Tempo and not getting the adaptations you desire. The workouts would be doable sure, but 2x20 at tempo serves a different purpose than 2x20 at threshold

Does it matter though, if you are doing 2x20 at an intensity that feels hard but doable, what label you put that zone in?

1 Like

I get that, and you’re right. If you can do it, then you can do it; it doesn’t matter what it’s called.

But when it comes to planning your workouts and picking what zone you’re going to be targeting, it does matter. I guess it matters more for those working out with ERG mode, or just power targets in general.

By all means, if you’re working out with RPE-based intervals then FTP doesn’t even matter does it :wink:

I mostly use ERG mode. My thinking is that to get a training adaptation you need to be doing the intervals at the limit of what you can do. The length and structure of the intervals should then ensure that you are getting the adaptions you want.

Only the recovery workouts are done at levels that should not be stressing you. But I don’t think the AIFTP will ever be that far off that 50% FTP becomes too stressful.

1 Like

Unstructured rides raise some questions as to how to answer the survey too - for example this morning I went on a group ride that wasn’t particularly hard, but I put in a hard effort on the sprint. I rated this ride as moderate, but I wonder if that makes the system think the effort I put in wasn’t hard and might lead to overestimating in the future. Hopefully once unstructured rides start to affect progression levels we’ll have more answers about this!

1 Like

This is a misconception you share with many forum posters. I bet fatigue is the main issue that prevent athletes to perform at best. You would probably be far better of if you save some for your next workout.

8 Likes

Was just about to say this.

Let’s say your true ftp is 250w. Doing threshold intervals at 240w doesn’t make them useless. The same goes the other way around, sometimes it can be better to do them a bit harder. It’s all about how you feel that day. FTP isn’t set in stone.

2 Likes

Yes, it absolutely matters, because the physical adaptations are different if you are wildly off

Except that is not how it works: your physical adaptations vary significantly between zones and how your workout is structured. Even workouts with similar structure, but different power zones lead to different adaptations. “Narrow” over-unders (95 % under and 105 % overs) will induce adaptations so that your body gets better at getting rid of excess lactate. “Taller” over-unders (90 % under and 110 % overs) already have different adaptations.

Ditto if you do tempo rather than threshold: these are two very different kettle of fish. And just because you find both “difficult” doesn’t mean the adaptations are similar.

Truth. The more experience I get with structured training, the more I learn that it is mostly fatigue management, especially long-term fatigue management.

Over the long term, riding tempo that often will be a very bad training strategy. I know you didn’t intend to suggest it, I’m just adding this for emphasis and context.

IMHO that’s the wrong way to think about it. If your true FTP is 250 W, then your training zones should be set relative to 250 W, not relative to 240 W.

That’s quite different from recognizing your daily form and understanding when it is better to make the conscious decision to reduce the intensity of a workout or choose an easier workout.

1 Like

With AT and PL this doesn’t event matter. If ai ftp underestimated you ftp (in the example 240 vs250). After the first two workouts. If you fill in the survey correctly, you will get progressed much faster to more difficult workouts (let’s say 4*12 @ 102 or 104% instead of 4x12@100%)
So the stimulus on you body is exactly the same.
Same goes the other way around.

Buttom line is that I think ai ftp should only be used in conjunction with AT and PL.

1 Like

I was referring to threshold and VO2 max intervals. Also being at the limit is not literally at the limit such that it is not sustainable over the weeks. It should feel hard but doable and that is the level that the AIFTP puts me at.

I get that when doing endurance workouts those should not feel hard and having the FTP set too high would make you build up too much fatigue.

Was listening to a podcast the other day where it was mentioned as one of the advantages of polarized training that there is more margin for error in setting the FTP, since you are mostly well above or well below FTP. The wisdom of the old coach was: “If you train right on the edge, it is also easy to fall off!”

It seems to me you are conflating what being at the limit means and setting your zones correctly. You can do VO2max intervals or threshold for certain periods without going to your limits, especially your physical limits. If you can manage do 4 x 3 minutes at 120 % instead of 7 x 3 minutes, then the answer is not to shift your FTP downwards. Endurance and repeatability at a specific power level is a different dimension of fitness.

You should not make your workouts easy by shifting your zones downwards, because that changes the nature of many workouts. VO2max workouts are less of an issue, but there is a marked difference between threshold and sweet spot work, for example. Your FTP should be set as close as possible to your lactate threshold.

This is not something that you should change by changing your FTP. Instead, you should choose your intensity level of your workouts, including endurance workouts, consciously. For endurance work, it is usually better to stay longer and extend the ride. (But that isn’t an option for most of us since most of us are time constrained.)

2 Likes

This makes no sense to me.

Seiler claim upper tempo (sweet spot) to be included in his z3. I’m not aware of any specific adaptions that occur at threshold so I have replaced most threshold workouts with longer tempo workouts. These are easier to recover from and can be extended over longer intervals resulting in more work performed.

In my VO2max intervals i pace myself to not exceed ~95% max HR as there isn’t much to be gained buy targeting a higher HR; it just adds unnecessary fatigue which only delay recovery.

Edit: I just watched this video that brings up a few of the points I’m trying to make: https://youtu.be/StnxjISyeWg

Been a while since anyone commented on here, but I’ll post my question here anyway. I’ve just completed my first base block after a period of doing other sports and using Train Now intermittently. At the beginning of the block AI detected an FTP of 207. That was a touch high, but I knew it didn’t have enough data, so I left it, and just scaled down the threshold sessions by a few percent to allow me to complete them. Now AI tells me my new FTP is 214 which is definitely too high, especially with VO2 max sessions now starting. Did I feed the AI with the ‘wrong’ data by reducing the intensity of my threshold sessions but finishing them? The questionnaire recognised it though, and I gave “intensity” as the reason every time. Doesn’t that get fed into the AI?

1 Like