Before you make a declarative statement about how “wrong” someone is, you should probably have your facts straight.
Your argument says that ok if someone fails 1 minute in to a sweet spot interval at 500W FTP setting, that means it’s too high. Ok, fair enough… Then what? The system continues to throw spaghetti at the wall hoping to land on the right number eventually? How does it know that the power deliverables at VO2 max aren’t too low?
AT can exist without FTP? You base that on what? Your opinion? If that were the case, that would be the model. It’s not. What your suggestion quite literally flies in the face of widely accepted training science.
Yes. its pretty easy to go a long time between formal FTP testing if you have good power-to-HR data, and/or a good feel for, as Coach Chad puts it, “walking that lactic tightrope and maintaining the balance between hard work and too-hard work.”
Some athletes have reliable/repeatable power-to-HR data when doing threshold and even sweet spot efforts. That allows the athlete or coach to visually inspect a threshold effort and roughly estimate FTP. It also would allow a computer to do the same.
The TR workouts are based on Coggan power levels, which rely on having a reasonably accurate estimate of FTP to set training zones for endurance, tempo, sweet spot, and threshold. Above threshold the % of FTP can vary quite a bit between athletes, so you need to dial that in for yourself.
TR has a catalog of workouts with a wide range from easy to hard. Adaptive Training in combination with workouts indexed by Progression Levels helps people without a lot of training experience, or interest in self-coaching, to quickly find the right workouts. The lower progression levels generally have fewer, shorter intervals, and the higher progression levels generally have more and/or longer intervals. In addition for vo2max efforts the lower PLs have lower % FTP (e.g. 110%) while the higher PLs have higher % FTP (e.g. 127%).
The entire catalog was built around FTP, and going back to the article linked above, Coach Chad states “Because we use this measure of fitness as the basis for all of your training, it’s important that this number estimates your actual sustainable-power threshold as closely as possible.”
Therefore you want a good estimate of FTP, and then AT/PL will help with locating the best workout relative to your ability in terms of interval duration, number of intervals, and % for above threshold.
A good estimate of FTP can be determined many ways: for some simply using power-to-HR can provide a fairly accurate estimate, for others a ramp test, while more steady-state athletes might prefer a 20-minute test for several reasons, and others have found good results from using computer algorithms in tools like WKO and Xert.
Progression levels in and of themselves aren’t going to make FTP assessments obsolete. Computer algorithms have been available, and TR is working on their own.
Just a few thoughts.
1 Like
This is more or less the crux of what I am getting at and the limitation of my own proposition. Testing sucks. Power curves are better BUT less reliable in terms of ensuring the athletes have enough breadth of data to support the model. There is a population of athletes for whom this is highly effective and another for whom this would fall apart pretty rapidly. Maybe a toggle FTP estimation on/off to be recommended for people following a structured plan with high compliance. This assumes TR sorts out the plethora of other limiters discussed in this thread
1 Like
The flip side of decay is new strong efforts. Last week I had two key workouts with season PRs under 90 seconds. Like a see-saw, that caused WKO modeled FTP to drop from 262 to 244W. That isn’t an actual problem if you understand how the model works, it simply underscores what I already knew (time for some long efforts).
1 Like
The older I get, the more of a Luddite I become.
The more stuff I read around here, the more I am convinced that I am definitely not right for this crowd. I don’t seem to understand any of this stuff 
I use TR to do the thinking for me, hoping it will make me stronger. Probably not going to work (for me).
2 Likes
Training before power meters was way less effective than it is now, precisely because you can address different power zones in training with pin point precision. FTP is one of those points, a rather important one.
The fact that there are almost equivalent definitions for FTP doesn’t change that.
No, it wouldn’t. Under the hoods AT likely uses a 6- or 7-zone power model — that’s what TR’s workouts are based on after all. The anchor point is an athlete’s FTP, because in terms of training LT2 ≈ FTP. AT adapts the zones as fixed percentages of MAP, which is what is measured in the ramp test with good precision and unambiguously. In this way AT adapt the power points to the athlete, and FTP is an intrinsic part of the story.
Within the current testing scheme, predicting FTP is the same as predicting MAP (at least until TR changes how it computes FTP from MAP). Even if this is doable, I think it’d still make sense to do a ramp test regularly to test your progress.
When modeling anything there is an inherent trade-off between using a more sophisticated model with more parameters and a simpler model with less parameters. A more sophisticated model is not automatically better for many reasons. Just three important ones are that you need to be able to determine the parameters with good accuracy from the data that you have. If you cannot, then the additional effects the more sophisticated model includes cannot be resolved. The other important factor is that more parameters can allow you to fudge your model more easily to the data. And if you can predict outcomes with a simpler models, it tells you that the additional factors are not as influential as the few in the simpler model.
The third one has to do with how you get data in TR specifically: most of my power curve is generated from TR indoor ride data. So the power curve is determined by the workout TR chooses for me. Yes, I sometimes drop the hammer, but mostly on shorter efforts. Climbs or uninterrupted straight bits longer than 20 minutes are hard to come by where I live. So my power curve does not necessarily reflect my abilities.
The WKO (and other) power curve fits to your mean maximal power data, therefore you need to test more (ie at many durations) to populate it. If you’re not doing maximal efforts at ~5s, ~1m, ~5m, ~20+m every 90 days, the mFTP (and any other model data) are probably no better than a flubbed ramp test.
If you are just completing workouts from a TR plan, you are almost certainly not hitting maximums at any duration (until you do a ramp test 
).
5 Likes
Do the pros assess? I’ve read about some of them doing it, but I have no idea how widespread it is in the peloton. They certainly have plenty of chances to dig deep in races. It’s hard, but I like to assess. I don’t get much chance to race, so I like the occasional all-out effort
I always hear pros saying that “my numbers are good” so they are assessing in some way. I think a lot of Teams are doing Inscyd testing a few times per year or more - at least one would believe that from listening to coaches on podcasts.
My point about the 500w SS interval was simply to point out a computer doesn’t need to know an FTP to start making adjustments. And yes, it would keep throwing spaghetti at the wall. That is why I said it is not optimal. But just because it is not optimal, doesn’t mean it can’t be done. This is opposed to your argument, that without an FTP, AT would do nothing. It is similar to feeding a search algorithm an initial good guess, or nothing. I said originally that it would be optimal to have an FTP assessment to start (feeding it a good guess). Once you start progressing, it can update what power to base your workouts on from progression levels. In fact, you already pointed out that there is an equivalent FTP for different progression levels. That is hardly different from what I am arguing. I do not see why that is controversial, given then title of your post?
Sure, it is an opinion, but it is correct. If a computer cannot classify interval workouts into different clusters, there is no way one would be able to choose the right workout for us.
As a first thought, I would feed an algorithm the length of intervals and the percent change from work interval to the rest interval. a 30-30 has short intervals, short rests, and a big % change from work interval to rest interval. An endurance interval generally has long intervals with minimal % changes. If this does not get it right, you can add in additional data points. If there are multiple types of intervals, which one is time being spent at more, etc.? The key is that you can fix the power the workouts are based off to classify them.
Now that the workouts are classified - all the AT has to know is whether the system producing the power can sustain the power for the entirety of each interval in a workout, with the goal of maximizing the power it can sustain over time. Obviously, this is not optimal, as it would take time to get it dialed in. which is why I said it would be better to have an initial assessment..
This is not that different from what is currently being done. Except the workouts are (maybe?) classified by a coach. You do an FTP, then the AT sets your progression levels. All I am saying is that if you hit a progression level at a specific power output, the algorithm could increase the power you base the workouts in that cluster and lower the progression level for those types of workouts. As such, the necessity of doing another FTP test goes away. Correct me if I am wrong, but you are saying that rather than adjusting when hitting a progression level, you adjust when you have an adequate power curve. The differences are minimal and I do not see why this is controversial.
At this point, I don’t understand why your title is based off progression levels when you are arguing to use power curves? I mean, strava already estimates FTP based off a power curve. Pretty sure others have pointed out other platforms do that as well.
That is all I have to say about this though. I wasn’t trying to get into a debate. I just love math and data science so I think AT is cool and thought your post was interesting. I think with the right vision, AT can do a lot of cool things - such as force the obsolescence of FTP assessments. Have a good day, and good training to you.
That’s just a different way of testing FTP, just worse. The power curve is mostly made up of rides where you do not go all out. Indoors I follow TR workouts, so I am far from all out (apart from ramp tests). Outdoors, I think most people can only really test shorter power PRs accurately, because you don’t do >40-minute all out steady state sessions during a longer ride. So that means you need to estimate based on the data that you have, which introduces the same problem that you have in a ramp test. Just that the ramp test has the definite advantages of being (1) repeatable, (2) quick and (3) all out (if done right). The result, MAP, is completely accurate. It is not perfect and should not be seen as the only metric. Once you think of it as a MAP test rather than an FTP test first and foremost, a lot of the issues that have been discussed to death here disappear.
The question is whether doing something like a 4DP test produces better results when tailoring workouts to you. (4DP could be interesting to find out where you are good at, but I am focussing solely on adapting workouts.) TR’s claim is that they tried it and it did not produce better results. I have to take their word for this, and perhaps that will change over time. But I think progression levels do the same thing: if they see that you can hold e. g. a higher percentage of your FTP during VO2max workouts, AT will serve you harder VO2max workouts. And vice versa.
AT rests on a feedback loop: it starts from your MAP obtained from the ramp test and computes progression levels from your previous workouts, and then extrapolates from that dynamically. But you need to test the ML model periodically to assess the validity of the model. Clearly, the longer you extrapolate, the larger the variation and the error. That’s why testing your MAP is necessary to give you the best results.
Even when ramp testing becomes optional, I reckon your training progression will be more optimal if you do incorporate regular testing into your training.
1 Like
I think ERG could potentially be a big factor as well, because you’re more or less riding to a predetermined number. Certainly adjusting the intensity is an option, but it’s not quite the same as riding to feel- for better or worse, it’s more a question of “how well can you meet this target” vs. “what are you capable of today?”
1 Like
Easy example, lets say we have two people where both have a 300w FTP.
That doest mean that their aerobic and anaerobic engines are the same. Since FTP is only “one” value, saying that both of their SS would be 88% of their FTP might not be correct.
This effectively means that if we say that everyone with an FTP should do SS at 88%, there are gonna be a lot of people who get burned out, since 88% might be their threshold, and for others it might be low SS (this is just an example).
And thats why its important to look at the whole curve, and then base power levels/target depending on where the rider is at those different parts of the power curve.
Usually coaches look at this by determining critical power at different times, say CP3 (critical power at 3min), CP5, CP10, CP20 etc.
With that data you can then get more accurate levels based on that persons engine!
Hence why FTP is such a rubbish number.
Ex: someone with a low FTP might be really good at short surges, and effectively be faster than someone with a higher FTP that aren’t as explosive, if its a course that requires punchy accelerations.
I think people obsess about numbers these days (no offense, not specifically you, I myself might be even more prone to it!).
Short story: A friend of mine (who happens to be my best riding buddy) rides the trainer exclusively from November to February and doesn’t even follow a plan. He just does his endurance/tempo rides and the occasional hard effort. He uses no power meter outdoors and trains only based on HR for endurance/tempo/LTHR and “sustainable all-out” for different interval lengths (read: strava koms 
)
He sits at 4.5w/kg (320watts) FTP and holds almost 90% of that for 2hours with 10hours/week. (Has barely top end though, but is happy with that)
Yes, maybe he is genetically gifted and who knows how good he could be if he followed a structured plan with power.
I find myself making good gains (momentarily at 4w/kg 280watts) with a year round structure but
hold it … hold it … here it comes …
who knows how fit I was if I “just” rode consistently with a raw structure of mainly endurance and the odd effort?
Yes, training without power meters is less effective, but training with power is also eking out the last bit of potential (I would not want to miss my power meter out of curiosity and to track progress though).
I honestly think that having a power meter translates “just” in as much gains as really nailing 8-9h sleep every single night (instead of sleeping 7-8/hours with the occasional 6-7hours) or having a pin point nutrition plan where 18 out of 21 meals/week are exactly calculated (and I’m not talking about “just” being aware of eating how many calories and vegetables, fruit, proteins, carbs at any given time. That would be “just” HR-training)
Bottom line: Yes, training without power is less effective. Is it really that much less effective not to speak of “way less effective”? I doubt that.
I hate to be so negative and I really like the idea of using the possibilities of modern technology to seek improvement, but …
Would a pro rider use something like AT? No of course not, but he also has a top coach, who not only knows everything about training but also monitors the athletes life.
So the question for me personally really comes down to this: A (half cooked) AI or a shitty coach? I pick me
4 Likes
To be honest, I am not quite sure what your point is. Of course, sleep and other things are crucial, especially the farther up you go in fitness. Yes, you can get fit without a power meter, cyclists did that for decades. In particular, we also shouldn’t be making general recommendations based on anecdata. Proper sleep and other healthy habits are part and parcel of training, and it is a bit silly to have to choose between training with power and good, healthy habits.
The whole point of TR (with or without AT) is that coaches don’t scale, and that most coaches cost a lot more per month than TR costs in a year.
1 Like
You are right, my former post is indeed quite a mess, sorry for that …
Let me try again:
I guess I didn’t like the statement that “Training before power meters was way less effective than it is now”, which is just not true IMHO. Not for pro riders and certainly not for amateurs.
Training without power and good habits get you very very far (and I then used the anecdote as an example and examples are used to visualize a general phenomenon).
Training with power and excellent habits get you even further. I think we agree on that.
My point is we tend to overestimate the use of powerdata and underestimate the influence of excellent (instead of “just” good) habits.
As an ambitious amateur (as I would define it at about 4.5 w/kg) you don’t need excellent habits (good is good enough) and you don’t need a power meter (a HR strap is more then enough).
Therefore “way less effective” just hits a nerve with me.
(I still can’t imagine losing my pm though …)
1 Like
This is not how mine worked. I had a two watt increase from 302 to 304 and my progression levels went down, but only a little and kind of what I expected.
And that math. .1 is not .1%. I believe it’s 348/355 = .98. 1.0-.98 = .02, which is really 2%. 2% is a pretty good result in my book.
A ramp test doesn’t have you hit MMP at any duration either. Just because you set a power record, doesn’t mean that’s your MMP for that duration. I guarantee that most people could break all their ramp test power duration records if they went out and just banged out the hardest 2, 3, 5min they could straight through.
4 Likes
Absolutely. But how many people actually do that? I have noticed that since I started doing ramp tests on a semi-regular basis, my 90 day P-D curve almost always has a nice smooth hump from ~3 to ~10 minutes from my last ramp test…
2 Likes
TrainerRoad users? I’d bet practically none. Doesn’t mean they shouldn’t. It’s one of the biggest gaps in TR training - way too few all-out efforts (if any) at various durations.
5 Likes