What is AT trying to do here?

I wasn’t asking why AT wasn’t substituting my Threshold workouts for Endurance for me. I was trying to understand why it was adapting some workouts and not others or why it was adapting some workouts that it usually doesn’t.

I think I have more expectation on the Plan Builder. I’m not sure why at this point we don’t have a plan with two intensity days that can be used with Plan Builder. Polarized isn’t an option because a) you can’t use it with plan builder, b) it doesn’t adapt because of that and c) I think we all see the progression levels and realize it’s not a very good plan and d) I don’t have 2 hours a day during the week for those weekday endurance rides.

I would expect AT to look at my next workout and suggest and appropriate workout based on my progression level. It’s not doing that in a way that I can understand, thus the question.


That statement is what I think is the ideal. It is what I would like but it isnt what it presently does.

AT doesnt assess what workout to do from my own experience. It assesses what PL to do.

Err, you can. Create a plan using PB then you can swap out the Base and Build parts to a Polarised Plan by click on that part’s name in the calendar. In the dialog that pops up, click on the down arrow next to the plan name and at the bottom of that list will be the “Experimental” Polarised option.

I’ve not done the Polarised Plans whilst using AT so can’t comment on how well the two play together, I’ve seen comments that it works and others that it doesn’t.

It is making changes based on what is/was currently planned and how you’ve told it you’ve performed on similar workouts in the past. AT won’t suggest an endurance workout in place of a VO2max one, it will replace a VO2max workout with another VO2max that’s slightly harder or slightly easier depending on where you need to be in regard to the weekly progression.


See above.

Which plan is this? That is too much intensity IMHO. Off the cuff what I see is AT keeping the load (weekly TSS) the same, by easing off threshold and increasing zone2 work (intensity, not time). Said another way, its decreasing some of the high intensity and adding low intensity. You could also look at weekly kJ of work, instead of TSS. :man_shrugging:

Basic ole SweetSpot Base II prior to General Build - Mid Volume

TR has an interesting take on base!

Traditional Base has less intensity, could try that one.

IMO SSB works best LV - 3x SS workouts a week is nice. If you’ve more time in base, probably having less intensity might be good.

I recently abandoned my plan because I wasn’t able to adhere closely enough to really utilize AT effectively. Based on podcast advice, I started using TrainNow and that’s working much better for me. It’s very easy to see how AT works and you have the freedom to go crazy on workouts.
In the plan the progression didn’t seem steep enough, but with TN I did a few of my favourites and that bumped up my levels quite nicely. Or not-so-nicely

This is not true. To use Polarized plans with Plan Builder, create a plan of your liking. It will use SSB and one of the standard build plans. Then in the Calendar you can swap SSB1+2 for the 8-week Polarized plan and Build for the 6-week Polarized plan. It’ll then adjust with AT just like any other plan you use with Plan Builder.

i dont want more sweetspot. i want vo2 and threshold work, just two hard days, not four. its not a strange ask. i should be able to do that.

1 Like

can i down vote your comments? your aggressive approach to help isnt helpful. one, thats a workaround, obviously not what tr intended, so how is that better than what im doing and two, there were two other dealbreakers i mentioned with my response on the “polarized” plans.

OK - one of the LV build plans will give you close to that.

It’s not a strange ask at all, but it fits later in the progression in the TR plans. The LV plans go like this:

SSB1 - 2x SS, 1x Threshold
SSB2 - 1x VO2, 1x SS, 1X Threshold
Build - variable based on event - short power build is 2x VO2, 1x Threshold, sustained power build is 1x Threshold, 1x VO2, 1x Sweetspot

TR will put you through those steps in order, moving from sweetspot to higher intensity zones as the plan progresses.

As above if you just want 2 hard days, picking an LV plan and subbing one day out is going to give you better results than what you’re doing now.

1 Like

@bobmcstuuff, i know the plans. i know what im doing. ive been through them each many times. i know how to add z2, i know how to manually do my own programming. i’m trying to understand why AT isnt doing what i expect.

sorry, not trying to sound frustrated with you. its just sometimes hard to ask a question here without taking it in a different direction.

Have you asked support? Maybe something is amiss.


This is a guess:

AT was likely trained on success and fail use cases from folks following TR plans.

The success data were very likely either folks seeing NOOB gains or people who had dropped FTP from prior levels and were building back. The fails, depending on how you look at it, could have many subsets.

I doubt there were many riders coming into TR already well trained who saw significant improvements as high intensity low time plans don’t tend to work for those use cases.

Further…. Id wager that the most robust TR data will be from SSB 1 and 2 and either General or Sustained Build programming. I doubt many riders do the full base to build to speciality. If correct, the use cases and data sets are going to reflect that programming and be constrained.

AT is not “smart”, it’s an algorithm. If that algorithm is trained to see “success” as a specific pattern, then it is going to push (adapt) riders into that pattern. If you try to fight that with manual adjustments, you will diverge from both the TR training philosophy and from the success use case structure and get into a mess.

If the primary factor AT can adapt is intensity (not volume + intensity) then the patterns of adaptation being described make some sense to me.

Note that Time spent training is a big important variable. Time = Volume. With Plan Builder, the user inputs the Time variable so I’m assuming AT won’t be able to adapt that volume significantly. Even if it could, there is likely not data in the TR ecosystem to teach AT how much to add or subtract or at what intensity that work should be at.

Taking this a step further - how will we know if AT work?

What we need to see from AT to determine utility is first order stuff - For athletes to actually use the suite of TR tools as developed. Plan Builder to define your blocks, and then follow the plan and have good compliance over many months. Let AT run and follow the adaptation suggestions.

Example: If I wanted to give TR and AT a good chance to help me, I would let plan builder program:


LV or MV depending. I’ve done MV for all of those plans and they are not bad. (I think there is better, but that programming is OK)

Take those blocks and Follow the plan and AT recommendations verbatim for the 20 weeks (6+6+8) along with all the suggested Ramp or other FTP tests.

Then come back and decide if AT works or not and if the adaptations over a good period of time and decent blocks made sense.

When enough people do this and report back, we will have an idea of AT success or not.

If it were me…. That would have been my beta test of AT before launch. I’d want to see better improvements over a control group doing same programming but without AT. But it’s not me so I’ll just type here then consume some turkey :grimacing:


Okay, let me try a long winded approach to helping you. Bear on mind I’m on the trainer using my phone to write it since it’s a recovery ride, so…
Imagine AT like a robotic cooking assistant in your kitchen. Everyone has had experience cooking and or baking, so hopefully it’s relatable.
You normally tell AT )your cooking assistant) I want to make 2 cakes for 10 people by this date. The cooking assistant says okay, let’s help you with that, and then starts telling you how much of each ingredient to add each step along the way. You can have the cooking assistant tell you all the steps, which is like looking at the entire recipe (which in this case is seeing your workouts and levels for future weeks), but cooking is dynamic and each step will need to be adjusted slightly based on temperature, humidity, and specific ingredients you use.
Your issue is that, as you say, the recipe you are telling the cooking assistant you want to make and the recipe you are actually making are different. So you tell AT/the cooking assistant you want to make a chocolate cake for 10 people but you are actually mixing ingredients for cookies. cakes are great, and so are cookies ,but your assistant says to mix four cups of flour and a box of instant pudding for the chocolate cake, but you add four cups of wheat flower and no pudding mix. On its own AT/your assistant are like…“uh, okay, let’s keep going I guess”. And tries to work with your selected changes. But for the next step you need to add 2 eggs, but you add Greek yogurt and flax seed as a substitute for your cookies. By this point AT thinks you’re trying to make a cake for 10 people, but your ingredient list is actually for cookies, and AT/cooking assistant is only an AI, and has a hard time processing you using other ingredients. It’s supposed to help you adjust quantity as you find out more people are coming (your FTP and levels increase), but it does that as you go, which is how levels work, showing you where you’re at and where you will be based on your initial requirements (which is the plan you want and the results of an initial ramp test). But if you are trying to change cake into cookies, the best the assistant can do is try to get you to cake even though you are putting in the wrong ingredients for it. So as it recalculates and suggests what will happen in the future, it’s not smart enough to switch you to cookies entirely.
My suggestion, change your plan or just use train now. You said you want cake, but you change to cookies, then ice cream, and so on. You’ll never get cake if you add ingredients for something else (you’ll never get steady progression and levels if you are training in a different way from what your AT plan things you want). Progression levels are like the list of ingredients for the recipe, and will stay fairly consistent throughout AT, but will adjust levels (quantities of ingredients) on the fly based on what you put in and the current conditions (such as how many people are coming and the temperature). You can’t see all of that in the future until you add more ingredients (you train with your current levels). You can see the general overview of a plan, but the specific levels will adjust almost daily based on what you did in the past, so don’t be too concerned with what specific level a future W/O is, because that’s going to change and that’s fine. The issue is ingredients in general. Commit to a cake, or just wing it with train now. That will help you get to your wheat cookies you really want instead of constantly confusing your cake recipe.
Let me know if that helps… because it is Thanksgiving after all so food is on the brain :stuck_out_tongue:


In other words, in the context of OP’s wish, that means picking an LV plan and swap out one of the intensity days. Add volume as you see fit. Let AT do it’s thing and accept/reject suggestions as you see fit because “you can’t always get what you want, but if you try sometimes you get what you need” (apologies to The Rolling Stones :metal: )


I’m actually suggesting no modifications. So while LV dropping day 3 and adding Z2 added volume indeed makes a lot of sense, I don’t think AT will handle any deviations from plan well. The training sets won’t have had enough good data to inform the algorithm for an LV + Z2 use case.

For me… if I wanted to really test drive AT, I’d just give it the best chance to succeed and that means TR plan verbatim.

But…. Since I don’t think TR plan verbatim is the best way to train, I won’t be doing that experiment on myself. But I hope we eventually see that data because what I think could be wrong.

Where this eventually goes…

TR is a time constrained HIT Based training structure. That’s fine. But if riders want a different training structure, the TR plans and TR tools (AT is just a tool) are not optimal.

I don’t understand why TR has not come out with base and builds using the two day a week hard plus heavy Z2 approach. It works great and based on forum comments, customers seem to want it.

To me, TR SSB is really build with two much intensity and the issues cascade from that foundation into build and speciality.

Two cents as they say.


I’m going to sound like a broken record, so flaming me on that point is acceptable :rofl:

But this, plus other threads / sub-threads on issues with AT are why I think TR needs to build AT specific workout blocks that instead of showing specific workouts, show:

  • Workout Duration
  • Workout Zone - maybe also showing the sub-zones (e.g., for Endurance: Attacks, Continuous, With Bursts)
  • Workout Difficulty

That’s all. And then AT would replace the generic workout description with a specific workout a couple of days ahead. This way people would understand the goals for each day in a workout block - e.g., a 1:15 Endurance ride that is Achievable - instead of seeing a list of ever changing workouts that don’t make sense in terms of what is the goal for each day. This is closer to what happens with a coach: you talk through the goals for the week, and then the coach selects specific workouts that meet the goals for each day.

And then in my ideal world, this change would be the entree to TR enabling AT to work with custom plans, as people could use generic workouts (with just the above three above characteristics) to create plans. And then AT would replace the generic workout with an actual workout.

1 Like