I used the AI FTP almost exclusively for the last two years. I was able to do time 20min time trials at the FTP test pace which matched the AI estimate.
Had a wobble last summer and the estimate was (IMO) high - but I wasn’t doing any structured training at the time.
Assuming it’s not changed dramatically, i don’t see much reason not to trust it.
exactly this, no need to test anymore with aiFTP, thats the whole point!
But the watts you should target for these many different scenarios don’t have a constant relationship with any version of FTP and never did. They’re going to vary by individual, by the training you’ve been doing, by how deep into the race they come, how much work you’ve done at that point, how well you’ve been fueling, etc. Figuring them out is as much about experience, RPE and race simulation sessions (or just races) as it is about FTP.
Thanks for pointing out that workout. 29% failure rate and that isn’t even at TRs number which is higher
I’ll do a 20 minute test later during the year just for my curiosity.
I’ve always done that way at least once per season and the 20 minute test is the one I’m most comfortable with.
So far with the update I have no complaints with the workout and AIFTP TR has given me so whatever I get out of the test will just be more data for TR.
Those tests alway produce some good numbers because you can go all out.
When We Were Young And Stoopid
Haha. I don’t actually remember doing it, a 40km TT sim, but my post says I held 8W less than my FTP for 59:28!
Well done again to the few nutters, I wonder how many are still here?
@bbarrera
@captain_doughnutman
@Bioteknik
@4ibanez
@jarsson
@stevemz (and SO)
In most of those situations FTP is not very useful as your efforts are anything but steady state and you have incurred fatigue in an earlier stage of the race. After, say, 2 hours, for how long can you hold your
FTP? And should you even try to hold your FTP rather than punch way above your FTP (e. g. to make a breakaway stick or bridge to the front group) or stay in sweet spot to attack on a climb.
In race situations, listening to your body is much more important. If you know what your body feels like at sweet spot or at VO2max, you don’t need to monitor your power numbers. Sticking to power targets is much more important for individual efforts (e. g. TTs, hill climb TTs or very long races) to make sure you don’t overdo it.
This whole discussion about FTP testing and AI-computed FTP numbers always boils down to the same thing for me: you verify your number, however it was arrived at, with workouts. With time I have gotten very good at sussing out my power at MLSS. In the past, that was very close with the ramp test result for me, but if it were systematically lower or higher, I would have simply adjusted my FTP test result accordingly.
You can do the same here if you simply accept that your FTP is the basis for your workouts and once you have experience with the numbers you get from AI FTP, you can easily pace off of those.
To me every non-endurance workout is a test of my abilities and yields a data point. I don’t need an FTP test of any stripe, e. g. a threshold workout will give me a good indication of my abilities.
Every time I see this, I think of TR’s video on Justin Rossi’s attempt to win TT nationals: TTs were (are) his thing, and yet the suffering he describes after a few minutes is significant. It belies that every athlete should just be able to do threshold power for 60ish minutes, otherwise it is “not their FTP”. It takes mental and physical training to do this much power for 60 minutes (or however long your estimated time is).
In your argument you are implicitly equating FTP with your best 60-minute power. Given that there is huge variation (I typically see 40–70 minutes) for which athletes can hold their power at MLSS, that’s not necessarily a good indication.
My understanding is that functional stands for “obtained in a field test as opposed to a lab”. Scientifically, only numbers obtained with the same testing protocol are directly comparable. Famously, there are different variations of the 20-minute and ramp tests, for instance.
Putting aside the many problems with the definition of FTP and what it actually means, was your indoor FTP always the same as your outdoor FTP in the past? For me I was able to put my head down indoors and push out a bigger number than I could achieve outdoors - I think because I was able to just focus on the effort involved without worrying about staying upright, other riders, cars etc. However I know many (most?) people were the opposite - they could push higher numbers with a number on their back. I guess what I’m saying is that I used the indoor FTP for training and then experience gave me the boundaries of what I could do outdoors. If it’s bothering you I would do some sort of test outdoors that is closest to what you will do “in anger” to give you that experience
I say test it. More data points are good to have. It will be interesting if whatever test you do produces a different number, and then if ai ftp changes because of your effort. Cause it it does adjust, then id say theres a problem with the ai.
Let us know how it goes
Measuring FTP is an estimate of an individuals lactate threshold/MLSS. I hope that’s not a controversial take.
I also consider TRs AIFTP detection is an estimate of the same.
It doesn’t make sense to me to compare an estimate to another single estimate. If validating AIFTP estimation, at best, you could compare it to all of the different methods to estimate FTP for an given individual - more effort than it is worth in my opinion.
Comparing AIFTP (or any testing method) to an actual lactate test or some other more accurate way of measuring MLSS would be way more interesting. Something like train red, possibly. Anyone used one of these?
For me, an U workout soon after AIFTP detection gives me a good enough indication to be useful.
I don’t think it’s useful or wise to consider FTP measurement as an absolute, akin to the measurement of the speed of light in a vaccuum, which is peoples perception of it come across in discussion.
I agree. I’ve trained and ridden long enough that I have a really good sense of where that turnover point is between…..i can do this for 5 mins, I can do this for 10 mins, I can do this for an hour, I can do this for 3hrs and I can do this for 10+hrs. Sometimes I surprise myself, sometimes I underachieve, most of the time I’m pretty accurate.
I do not agree at all with the number that TR has given me recently for an FTP. It has supplied me with great workouts, but it is simply not FTP by either the original definition by coggan, or any widely adapted iteration that’s come and gone in the past 20 years…..all of which are generally a number you can sustain at a steady state for a significant amount of time. At least 30mins, up to 70mins.
Is it a problem if it gives me good workouts……maybe not? Except it does push every zone up, either on TR’s platform or if I use any classical zone model with TRs version of FTP. So now mid z2 workouts are upper z2. Upper z2 are into low tempo….etc, all the way up the scale. And if I input my TR FTP number into another platform like training peaks or intervals.icu, as one would do because both say FTP, then my ATL/CTL is gonna be 3-5% lower than it has been in the past. I’m doing the same amount of work in pure watts/kj’s, but as a percentage of my FTP its all lower. That might cause me to want to train more to achieve similar loads to what I’ve done historically, which would actually be overreaching and potentially into overtraining.
As I’ve said before, I think the easiest fix to all this mess is to baseline the FTP number to a higher PL workout. If you can’t do something like Stromlo +3 or Mount Hayes at your TR FTP, you have no business calling that an FTP. If you physically can’t do 4x12 minutes 1-2% below your FTP and rate it hard/very hard, then I’d like to hear the argument as to why that’s still your FTP. If you can’t do 4x8mins at 99%, is it still FTP? Why not just call your vo2 max your FTP at that point?
If the AI thinks that shorter, more intense workouts are what drives fitness increases the best/fastest, that’s fine……then it should assign more supra threshold workouts, not inflate everyones FTP and give them level 3 threshold workouts with an inflated FTP.
Just to add to that: Coggan alone has changed his mind on “the” definition of FTP over the years. Machine Learning (the way it is used now) wasn’t a thing nor was there a corpus of data comparable to what TR has, and a “definition” via ML algorithms wasn’t on anyone’s radar, because it wasn’t possible. That doesn’t mean a statistical approach based on a data set that an exercise scientist couldn’t even dream of creating can’t be more accurate or better.
Even within the “accepted definitions” — or rather, most commonly used FTP test protocols, there is a lot of variability.
You are assuming that AI FTP has a tendency to overestimate “FTP”. How do you know that? According to @Nate_Pearson about 2/3 will see a decrease in predicted FTP and less than 20 % an increase or a significant increase. There are several threads and posts like this where users see a decrease of AI FTP.
Importantly, a smart choice by TR was to simply make TR AI oblivious of all of this, because according to TR they are based on absolute power numbers and are not anchored in FTP.
I don’t think it is wise to use several training methodologies simultaneously. TR hasn’t used ATL and CTL ever (as far as I can remember at least), whereas other training approaches use CTL and CTL ramp rates, something that is quite a bit more primitive than what TR is using at the moment.
Yes but does anyone up to this point (including all of Coggans changing views) have a definition of FTP that says you can’t do 4X12 with 6mins recovery on a normal intense training day at less than your FTP, but that’s your FTP? That’s fundamentally what this new TR AI model is selling us.
TR hasn’t used ATL and CTL ever (as far as I can remember at least)
They give you a TSS for every workout and every week, with a rolling chart and have for the 10 years that I’ve been a subscriber. TSS is a function of your FTP. It’s a trash number if it’s based on an inaccurate/inflated FTP.
I’d guess those athletes FTPs were either WAY overestimated under the old model, or there may not be enough good data on them for Ai to predict them under either scenario.
I thought that it would only be seen as a workout, but Eddie corrected me here
A 50/50 shot at making it an hour sounds like a pretty traditional definition of FTP to me.
Did you mistype something here? Generally the easiest workouts you get after an FTP update are over unders for a couple sets of ~10mins, and they are expected to be rated hard mostly, not all out.
He really has not changed the definition since the wattage days.
Adding more data won’t change or update the definition of FTP.
No?
If you’re calling a number your FTP, I’m of the strong opinion you shouldn’t be working towards being able to do (something like) 4x12 at 99% of it with a bunch of rest (Mount Hayes)…..you should be able to do that on day one and find it hard. If you can’t, that’s not your FTP and your FTP is lower. If you need shorter, harder, different intervals, over/unders….etc to build that lower FTP, then those should be provided by TR…..but it shouldn’t change the fact your FTP is lower.
I’m working towards doing 4x20 at 110% of my current FTP. But that doesn’t mean my FTP is 10% higher than it is right now.
No, the purpose of AI FTP is to help select the right workouts from the library. I. e. it is used as a training aid to select the right workout from the library.
(Their scoring algorithm does not have a concept of FTP, but the workout library is based on FTP. AI FTP generates a number that then generates workouts based on absolute watts, which are scored with the TR AI algorithms (that lack a concept of FTP). Hence, FTP currently still enters, although I reckon this may change when TR implements dynamically generated workouts.)
Yes, but they do not provide any of the other Coggan/TP training metrics. Nor do they really do anything with that TSS number. I mostly use the bar chart to identify rest weeks and off weeks (mostly due to illness).
Those “old” FTP values were not solely determined by AI FTP v1, but also various tests and athletes’ own estimates of their FTP. I remember unnatural spikes around specific FTP values such as 300 W.
It is important to keep in mind that common lore based on individual cases may (statistically speaking) be false. I remember @Jonathan saying on the podcast that the ramp test was on average underestimating FTP even though if you looked on this forum, the impression you got was that it tends to overestimate.
Why do you assume that raising your FTP isn’t part of the improvements? Old-school athletes who were “raised” in the early 2000s on FTP tested much less than TR did (once every 2–3 months), so increases in FTP had to be baked in.
There is no “the” definition. Coggan’s changes have been documented. He proposed the 20-minute test as an FTP test protocol in what is AFAIK the first document where FTP was coined. Only to later disavow it (e. g. on this forum). He also waffled between “average power during a 40k TT” (which may take you much less than 60 minutes or longer), 60-minute best power and power you can hold for “approximately 60 minutes”. There are other FTP test protocols that people like to use (e. g. Kolie Moore’s).
Even lactate tests are not the gold standard you think of, because of the few data points and the fact that to my understanding 4 mmol is used as a proxy for MLSS even though there are many athletes who have a much higher natural lactate concentration at MLSS. Nothing is perfect and we ought to remember that.
Even if we had agreement on a scientific definition (in the sense of exercise physiology), that may differ from how “FTP” is used as a training aid to prescribe workouts (vs. a field test to approximately determine power at MLSS or a performance metric).
If someone starts TR without any prior knowledge of the concept of FTP, I reckon they would be none the wiser as the AI FTP value seems to work significantly better than what came before it in the context of structured training.
These are testing protocols? Not sure how those are relevant. His definition of ftp has been consistent.
I never said anything about lactate testing nor it being a gold standard 
