Hey all, I checked in with our developers on this, and it’s indeed a point of refinement currently being worked on in some updates to Adaptive Training’s logic. As for whether it’s a bug or intended, the best answer to both is “yes, sort of.” Ideally, your Progression Levels should adjust with your FTP to guide you smoothly into your newly more-intense workouts. I can personally attest that when it works, it’s incredible, and a game-changer. But a some athletes are finding, it can over or under-compensate, which is certainly not what we intend. Since the specific relationship between FTP and Progression Levels is nuanced and differs not only from zone-to-zone but from athlete-to-athlete, it’s a complicated and ongoing project to refine it, and is one of the major tasks the beta is helping us to work on. As always, we appreciate the ongoing feedback, and are particularly appreciative that everyone on the beta has been so constructive in their criticism and patient as we work towards perfecting AT.
This makes me a bit nervous to hear. Knowing the limitations of AI against the variability of how I feel on any given day, I’m hope we aren’t expecting too much on what the AT can do. One thing I’ve been wanting for a while is better guidance in the app itself on ranges of productivity for any given workout. I know there is guidance to reduce the intensity by 3-5% if struggling, but I’d prefer to just have an explicit target range that is appropriately +/- the nominal number for every interval rather than a single target number.
I’d trust a model to handle the larger scale targets, but I don’t expect it to be super dialed in on any given day or zone.
Hi Sean, having progressions automatically ‘flex’ targets is awesome in particular if you don’t have a lot of experience doing that yourself. I’m going to put on my captain obvious hat and point out that you are dealing with two adaptive systems (humans and the AT software), so it is completely reasonable that sometimes the entire system (athlete+AT) can under or over compensate. In fact it should be an expected outcome for some % of athletes using AT, and therefore the real question is how you intend to handle those adjustments (from the athlete’s point-of-view).
Because it seems if you expect a high level of perfection - say an extremely low % athletes seeing discord between two adaptive systems - one might conclude that everyone will be waiting a long time for it to leave beta. And that if you offer some convenient option to override, it might leave beta sooner.
Just thinking out loud.
For sure- all options are on the table. Once again, we appreciate every suggestion and bit of criticism/feedback/glowing testimonial! We read every post in here and take them all to heart.
Is there a table similar to this on an AT training info page somewhere? If not, then there should be. It is an excellent starting point for either creating your own evaluation criteria or just using it as-is if you feel it is applicable - I now have a laminated copy next to my trainer to aid in a more consistent ride evaluation.
(Note: I did a quick search on Ride Survey on the TR Training Blog to see if something similar existed but nothing immediately obvious was returned apologies if my search was inadequate.)
Could someone please remind me of the appropriate survey responses relative you how you did on a work out please. Also how they relate to the level of progression (eg easier sessions vs progressive etc.). Failed vs struggled.
There is so much noise and so many comments on this I can’t find what I am looking for. (I think @mcneese.chad summarised them at one point).
I am not on AT yet, but wanted to start noting how I was doing relative to the workouts I am currently completing and their digfficulty levels.
Thanks
There’s no right or wrong answer- don’t overthink it. Just be honest with it, and imagine someone was asking you “how hard was that workout?” What would you say? That’s the answer. There’s also no “expected” answer relative to the difficulty of a workout.
Sir, this is the TrainerRoad forum.
Clearly I underthought that response.
Does AT apply to a customized plan? Say for example I add a recovery week to SSB1, right after the 3rd week. Will I still get adaptation from AT?
Unfortunately - no. As of now it only works on plans that are put on your calendar by plan builder. So if you add a TR plan to the calendar manually or build your own plan it will not make any adaptations.
This is ‘in progress’ but not yet released.
I’m currently in the AT beta and doing a TR polarized plan - which can’t be added by plan builder so I get progression level updates but it doesn’t adjust my workouts
2 things. Does anyone know if you ‘push’ a week in your calendar when you are on a planbuilder plan does that break AT’s ability to modify workouts in future weeks?
#2, byoungxprt I am not sure the system will ever see that thing as endurance. It will see it as a some threshold with some sort of sweetspot intervals. I think what it means by endurance in this context is z2. That said, they have stated AT + custom workouts is work in progress so results are not well defined right now.
Great question. I’m likely to move a workout from next week into the weekend as I have nothing scheduled from plan builder. If I manually add a few workouts instead it sounds like they will not count towards any adaptation. Maybe I’ll just do some easy recovery sessions instead.
I disagree. Anything over 2 hours probably has some impact on endurance depending on where you are in your progression. This particular ride was 60% z2, 25% z1, and 15% z3+. Similar TSS workouts in the 2.5 hour range are rated around an Endurance 5 so it makes no sense that this workout would be an Endurance 0. But yes the reason I posted this is because I understand custom workouts are still being worked on and I think this type of thing should probably be looked at. I personally don’t do much endurance work on TR. That’s mostly done outside or in Zwift.
I get what you are saying, and dont disagree. I am more saying I dont think the computer will ever agree with you. If I were trying to design a workout that was endurance-ish that an algorithm would think was anything but, this is the sort of thing I would design. I also dont know a damn thing, just an opinion.
That helped. I had to refresh the page multiple times waiting for response to appear and it finally did. This took several minutes. Hopefully this will be fixed in future as otherwise it is not obvious when it is safe to close the app.
As for Android app, I fixed that too. What I did is I had a fresh start by clearing cache and data of the app. Now it works as intended and does not freeze at the end. I am also happy to see that fresh install does not require to wait like 20 minutes anymore until the whole career is synchronized with the app.
Is there really value to our subjective feelings? To me rides are pass or fail. There is no inbetween.
Yet someone else may have a list of how they answer and say use a different range of success or failure.
This just feels like noise when the actual ride can be quantified by time in the various zones.
They should do I think. My over under rides over the last couple of weeks ridden outside weren’t seen as successful by AT which was reflected by no changes to the VO2 max progression over those two weeks despite hitting the power targets.
A bit of bad weather this week so this I rode indoors and selected a different workout, not one in either the original plan or any suggested adaptions, based on the progression levels as a productive workout which did result in level changes.