I think this is spot on. I donāt get the sense they are trying to nail what your max fitness is and give you workouts at that all the time, but rather keep you at just the right amount of effort to remain productive and consistent. Thatās a pretty wide bar, and being conservative is better than going based on your best effort.
That said, Iād rather they indicate what your achievement level is by completing a workout, and then give me the right dose of workouts (be it lower or higher leveled ones depending on volume). That just doesnāt seem to be the way the system is operating.
That sure could be the case, and may explain the lack of changes in PL that I would have expected.
Related to the whole survey response, what is expected, what happens as a resultā¦ it is a mess in my head. I wrote a best guess comment that was my understanding of how things might work and it seems I was at least partly, if not fully off base.
Because of that, I really donāt know what to expect, and that is why I posted what I did. By my research, it would seem that the āworkā I did would equate to a 4.4 Threshold workout. Since I completed it, specifically at a 3-hard rating, I expected a bump up in my PL.
I have completed workouts at 4-Very Hard and gotten bumps in PL, so I donāt think that a 3-Hard rating means there will be no change.
As I tried to estimate in my likely failed post mentioned above, I think the rating must be viewed in context to the assigned workout level, and the actual results within the workout. It all seems very complex, and it boiling my brain. So at the very least, I am trying to understand what TR is doing, so I can adjust as needed.
In my case, I may have been ābetter offā to pick and perform the harder workout, with the same 3-Hard rating to get actual PL changes.
If the reason I didnāt get a bump is the fact the Super-Pass is not in place, fineā¦ as long as I know it will be addressed properly in the final tool.
If this result is āas intendedā, then I really donāt like it and would take this as an indication that I would change my direction when given a similar choice in the future.
So much of this AT use is ātrustā in the process, but itās hard to follow without at least a consideration of what should or should not happen in some cases.
Trying to game the system instead of following the plan? Adaptive training at the moment is built around TR planned progressions, correct? Seems like it might be better sticking with LV plan as-is and doing your extracurricular efforts outside the planned progressions.
Not gaming at all. I saw a WOL 2.4 on my schedule, know my PL of 3.5, and was feeling like I wanted a bit more than an easier, āAchievableā workout. Thursday is usually one of my hardest workout days of the week, per the regular TR schedule, and I wanted that vs one that was too easy.
This also has to be framed by recognizing that my PLs are still messed up, from the bad association issue I covered well above, not to mention the fact that I am also doing outside workout, but on a Wahoo, so I am not getting any credit for those, or the unstructured work I am also doing outside.
I have an incomplete picture in my current PLās, and am trying to still get what I know I can and need to do with my current goals in mind. I am ābeta testingā in ways and dealing with several ongoing issues that make the current state of AT well short of a "donāt mess with it proposition.
Yup, and as covered above, mine are not an accurate picture at the moment.
For the record, I am on a Mid Vol plan, as I have done for many years, successfully.
I am making the best of a bad situation. I figured that trying to elevate an actual TR inside workout would yield positive results. I chose one method to do that (via workout intensity vs workout swap) to try and compensate for the data losses I have with my outside workouts not registering, and my unstructured stuff also no in the mix.
The problem all stems from an incomplete and inaccurate picture of my current PLās.
Any time I āmissā a scheduled TR workout for any reason, AT assumes fail and drops PL, and all following workouts in that Zone. Not idea, and not correct for my current situation.
If I blindly take what AT is giving me, I am getting way lower level workouts than are actually appropriate for my real fitness.
I could swap to LV in attempt to minimize the issues for when I choose to do a workout outside. But then, when I choose to do one inside as an extra workout, I am left guessing to a degree again. So that is itās own issue, and adding workout outside in any fashion still gets ignored.
I donāt see a āfixā here that removes the guesswork for anyone that is doing any workout other than the current TR inside workouts, or outside via Garmin. Itās all quite messy.
I have no idea, but Iām assuming itās asking about the workout as designed, not caring that you exceeded that expectation. If you could exceed the targets of the workout, I would think it would not expect you to say the workout was āhard.ā It was in fact probably relatively easy, even though you made it harder by exceeding the targets.
If true, that is TERRIBLE programming and makes no sense to me. It HAS TO compare the planned workout to your actual performance or this just doesnāt work.
We already know that if we end up under power too much, pause, backpedal and such, that AT will trigger a āStruggleā survey based upon whatever criteria they set.
We also know that there is supposed to be something called a āSuper Passā. This implies that it is looking at the workout to recognize if and when you perform above the prescribed workout. I have no idea what that threshold is, or if it would apply directly in my case.
I was not necessarily expecting a super pass in my case, but I would expect that AT can see that I performed a workout that was above the prescription, frame it within my current Progression Levels and apply changes to my PL accordingly.
Yes, the defined workout was below my PL, and set as āAchievableā. Had I done that one, I expect I would have rated it at a 2-Moderate, and would have expect no change to my PL, since it was below my current level.
Again, makes no sense for me if the app ignores my actual performance with respect to the planned workout and my current PLās. Itās would be effectively blind if it didnāt take those factors in to place.
I am trying to follow the TR advice, which is roughly ārate it how it feelsā, and the effort I performed yesterday felt to me like more than 2-Moderate, less than 4-Very hardā¦ so, you knowā¦ 3-Hard.
For what itās worth, hereās what TR support told me when AT left a Threshold 1.1 session in my plan after Iād completed a Threshold 3.7 the previous week (this was at the end of April):
Itās unfortunate that you havenāt been given any adaptations to change Buttertubs. =(
This being the case, it seems unlikely that it will give you an adaptation, so if it still hasnāt on the day of your workout, we should use TrainNow to find a Threshold workout that you can replace it with. You can also replace it before the day of your workout, but if you decide to do that, make sure that it is a Threshold workout at your progression level. =)
AT sometimes needs a little help, so āfollowing the planā is a little more complicated than usual. The better we understand what the system is trying to do, the better we can help TR get AT out of beta, without tanking our training.
Why? AT knows what you actually do, so to me it makes more sense to rate the actual effort, not try to imagine what the target effort would have been.
The beta users could benefit from a little more detail on what the system is actually doing and what itās expecting from users, I think
I would think this could potentially trigger it to bump up your PL, actually. By recognizing it was easier than expected, wouldnāt AT then conclude the level needs to be higher and bump you up to keep you at the right level?
Sure, there is a scenario where a low survey rating would trigger or increase the resulting PL change, but I would only expect that in a case where you are doing a workout with a WOL at or below your current PL.
My case was inverted (WOL lower than PL), so I figure Iād have needed a 1-easy rating to hit that trigger, but I am purely guessing.
Yeah this is exactly the kind of guidance we need. Whatās the best way to feed the model? If weāre presented with a session that we know is below our capability is it better to
do the session as-is, complete the survey, let AT figure it out, or
do the session but increase the intensity/shorten the recoveries, generally make it harder, or
replace the session with what we think is an appropriate alternate?
And the same thing on the other side: If AT gives us a workout that will be too hard, is it better to fail it, make it easier, or replace it with something easier?
Here is the reply I got from support. I am still digesting the info.
I took a look at the workout youāre referring to as well as some backend data giving me a glimpse of the inner working of AT. It looks like this workout was correctly classified as a Super Pass, but it was too far below your original Threshold level of 2.4 to raise your level.
I agree with you that the power profile of The Priest +5 does look similar to your final power profile, but this isnāt quite how the backend processes workouts and decides how to alter progression levels. Additionally, rewarding a full point increase in progression level is a bit of a stretch after completing a level 2.4 workout. The goal with this is to keep athletes within their own abilities and that large of an increase is too much for a Super Pass.
To be clear, if you Super Passed a workout that was closer to your original progression level, you would have seen a slight increase.
It makes me wonder also if swapping using Alternates is indeed giving you credit or if itās doing the Plan Builder breaky thing and therefore isnāt working properly. I say this because even with outdoor Wahoo workouts, I can fill out the surveys still and see things that look like Iām getting credit, but the bug list suggests thatās a farce.
EDIT: Scratch the above. Just saw your follow-up from support.
I think even without such guidance, AT needs to factor in that people will do all three of the above and soooo much more. People do wild stuff and actively donāt follow rules and guidance in so many situations without realizing it. I would hate to be the person on the team tasked with understanding all the weird ways the user will āincorrectlyā interact with AT and create a plan for how AT addresses it. Thatās gotta be a crazy job.
Agreed! Iām sure if we asked a TR person what the best option is, theyād say āit doesnāt matter, when it launches AT will handle all three scenarios the same wayā, but thatās less helpful in the beta when all the pieces might not be working in concert yet.
This makes it sound like their model is pretty limited, and maybe they should make it clearer that if a workout looks too easy, you should search for an alternate thatās a higher level, rather than just working harder on the current workout to get the best possible adaptations.
Iāve swapped workouts twice this week using alternates and not alternates in plan builder and gotten +1.7 both times after exceeding targets slightly on stretch rated workouts and ranking them as hard.
Perhaps the key was they were both rated above my TR estimated progression level, and I mostly stuck to the workout as planned.
Edit to say that +1.7 brought my PL to exactly the level of the workout.
I suspect the issue with properly identifying workout level is a similar problem with classifying outside workouts.
I could be wrong, but from the discussion, it sounded like they have developed an equation based model for determining workout levels. AT has the capability for wiggle room in profile matching, but not to the degree that if you were to follow a profile from a higher level workout it would put you at that PL like it would have if you would have swapped it.
If all thatās right, then I would guess they are working on a NN to replicate their workout level classifier, along with the right amount of smoothing to feed in what you do and consistently rate it. At least thatās what Iād do.
