No the TR aiftp is about 20w too high
1 Like
That makes me feel a lot better. I was feeling like maybe I had some weird error on my account and I’m happy to see it’s not just me. I love that they are pushing this direction but it seems like it’s not accurate yet.
If you’re someone who does higher power outdoors for the same RPE/HR than you would indoor, and you go into a block of indoor riding using the same power meter you’ve been using, will TR see those lower numbers and suggest lower FTP?
I really want to use the Ai FTP button as looking at some of my progression levels it would seem that my FTP is starting to increase but I have Disaster day approaching and well I would rather have it a bit easier if I can lol ………
In all seriousness I look forward to using this soon, well done TR on the release.
I decided to peek. And will have little regret if I decide to bump intensity down 3% in the moment.
today was the first time in forever where I really attempted a long threshold interval. I was a bit time crunched so I aimed for 35mins. So while I’ve had access to AIFTP since the early release last Feb, I haven’t adjusted my ftp at all, and it’s been set at 295, my AI FTP readings are generally around +/- 5w and my most recent predictor is 299.
Anyhow, I did my workout in zwift on sim mode, and only watching 3sec power, so I wasn’t super influenced by averages. for 35min I avg 295 and I’d say it felt about right at that edge of doable/hard. I can pretty confidently say I prob could have easily done 5 additional min, and probably at 45min I’d really be feeling it, so I think my personal estimate has been spot on. Def not feeling (at the moment) that I’d want to push my luck up to 300. But I’m also at the tail end of a block and not super fresh so who knows if in a couple of weeks I’ll change my mind!
Not a lot of comments but it’s looking like there’s some people who are finding it very accurate and some people who are not. Would be nice to know if there’s something skewing it one way or the other.
What do you mean by this? Are you using Xert’s prediction and then using it for your TR workouts? Are you doing a specific workout to see how the FTP feels? Or something else altogether?
I look at the Xert FTP prediction then I do an FTP test and the two numbers match. I’ve done ramp tests in Zwift Sufferfest and also the Sufferfest 4D test, they all match what Xert predicts.
1 Like
A couple of thoughts then: Each test/prediction you listed requires maximal efforts (as far as I know, at least). TR’s doesn’t. If you are like me and struggle with max efforts, it could be that all of them are low.
I once did so poorly on a ramp test it would have had me doing sweet spot intervals at 193 watts when I was doing 3-4 hour endurance rides at 200 watts. this was before AI FTP so unfortunately I don’t know what it would predicted me at.
My current AI FTP from today was 279. I am almost positive I wouldn’t get that from a ramp test. I would guess I’d test at 250-260. However, I just scheduled a tempo workout with PL 7.1 that looks easy at my AI FTP . Which brings me to this quote.
For me, this rings true. It definitely sets my zones better than the ramp test.
That said, you could absolutely be right and be in the ~6% that would “struggle” in subsequent workouts using the AI FTP. It might also be worth trying some workouts at the AI FTP to see how it feels.
2 Likes
Surely ‘failed/struggled on next workout’ is a terrible metric for qualifying AiFTP as it’s not very granular and a single point in time. Surely what your need to check is plan or block compliance following different methods. Even if your FTP was 40w over, you can still fight through one or two workouts; but the cumulative fatigue will break you.
Similarly everyone on here is talking about a ramp test underreporting. I’ve no idea how that can happen unless you just sack it as soon as it gets a bit uncomfortable.
My AiFTP is 20-30w lower than I can push on a ramp at the moment and I’m finally able to complete a training block without failure.
1 Like
I’m unclear what the consensus is here. It sounds like @voldemort is essentially saying that the aiFTP is the correct FTP and the others are wrong. Is that a good summary? If it is, I’m not totally opposed to that idea but as mentioned by @Neuromancer I’m not sure how to evaluate that. You are just suggesting to try some workouts? Would you think that the aiFTP is transportable to another service (like zwift)?
@Neuromancer it sounds to me like you are saying that your aiFTP seems more accurate than a ramp test. Am I correct?
1 Like
I’m saying that AiFTP delivers me a training plan that is difficult but achievable with what I view as the correct amount of grit, but I think it errs on the side of caution and so tends towards more achievable Vs more accurate because consistency beats specificity every time.
Whereas the ramp would leave me broken within a week or two, irrelevant of my grit.
1 Like
To me that sounds like you think the aiFTP is a bit lower but better for it. That’s great but mine is 30 watts higher with aiFTP. I can’t imagine that would work well for me to use as interval targets.
dumb response that is both toxic and misses the point.
Yes, but also no - it doesn’t miss the point, thought I did delete the post because I was worried you wouldn’t get the joke.
TR’s modelled FTP is based upon a large data set of what human beings attached to power meters can do.
If someone wants to make a case that the model doesn’t work for them by a massive amount (c.15% assuming you have the median TR FTP) you’ve got to get past Ockham’s razor first.
I am similarly guilty: I like AiFTP because it delivers me something achievable. But does that give me a confirmation bias that it’s right?
1 Like
My confirmation bias would be it’s great to go around saying my FTP is 30 watts higher. It kind of looks like I can even do the workouts. That said, when multiple other sources say one thing and TR is the outlier, I have a hard time believing that. I suspect there’s something happenning where it’s not that accurate for some types of riders.
Or the multiple other sources are all subject to the same issue?
I’m trying to keep a level of humility because I’m aware the million ride model is probably better than my 2 random data points, no matter how uncomfortable the consequences.
For me that means that I’m probably not as fit as I thought I was, but I might now have a better shot at being faster.
Does anybody know if the AI FTP is influenced by survey responses? These are obviously quite subjective, but I could see how some systems might use the indicator that a tough workout has been ‘surveyed’ as easy and concluded that the FTP should go up a notch.
Just keen to get on top of understanding how trustworthy I can consider the AI to be or I should run a more traditional FTP test in parallel (hopefully no need!).
As I understand it it does take in survey responses, at least for TR workouts. I can only say n=1 that AI FTP and PL’s seem to be working ok for me, following TR plans (which is the most important thing for me, paying my sub to TR!)
Rather than do a test, you could sub in Lamarck which would give some indication.
2 Likes