Please add the *WHY* behind proposed adaptations

TrainerRoad Software
1

I have been using TR for about 4 months now, and am really enjoying it. It has been great to have a workout scheduled for me on each of my training days, and to see the advancement of my Progression Levels and FTP. I have also found the AI training adaptations to be very interesting.

One area where I am struggling, though, is understanding the why behind the adaptations. If I’m completing my workouts as prescribed, and not regularly rating my Hard Interval workouts as ‘Very Hard’ or ‘All Out,’ it makes sense when the AI changes me to harder workouts. But I’m left at a loss when I start getting proposed adaptations for easier workouts, so it would be nice to get some sort of (presumably AI) blurb telling me why when this happens. (It would be nice to have a blurb when the workouts are shifted upward, too - something like ‘you’re doing great, so we think you should be able to extend a bit more’ - but more important, I think, when the adaptations are scaling you back.)

Example:

Last week my workout schedule (2nd week of Specialty phase) was:

Mon: Off
Tue: Threshold 3.5 (Creise), rated Hard
Wed: Endurance 5.3 (Boarstone), rated Easy
Thu: Endurance 5.1 (James), rated Easy
Fri: Endurance 1.7 (Colosseum -3), rated Easy
Sat: Sweet Spot 7.2 (Yarbrough), rated Moderate
Sun: Endurance 5.4 (Maclure), rated Moderate

This week my schedule was:

Mon/Fri: Off
Tue/Wed/Sat: Endurance 5.1 (James), all rated Moderate
Thu: Endurance 2.4 (MacDonald), rated Easy
Sun: Outdoor Solo Ride, 1.5 hours easy, but ended up being closer to 2 hours, but was still easy (and rated Easy), and I was not given a Yellow or Red day for Monday)

However, my original schedule for the upcoming week (which called for a Threshold 4.0 ride on Tue and Endurance 5.6 rode on Wed) was showing proposed adaptations to a Threshold 3.0 ride on Tue, and an Endurance 2.4 ride on Wed (with the remainder of the week unchanged as of now) subsequent to my Sunday outdoor ride.

This, then, just leaves me wondering why??? I don’t feel tired, I’ve been completing all of my workouts, with only the Threshold workouts feeling ‘Hard’ (which I would think they are supposed to be). So it would be nice to be given some sort of insight as to what the AI is ‘thinking’ when it suggests that I decrease my workload, especially on the Threshold workout, which is where I seem to need the most work.

I really want to ‘trust the plan’, but situations like this just plant the seeds of doubt in my head, since they don’t seem to make sense. For the moment I have skipped the adaptations for Tue/Wed, but they may well show up again by the time I get to Tue. But I might be more inclined to accept the adaptations if there was some sort of explanation.

So, anyway, that’s my very long-winded explanation for why I feel it would be nice to get some insights to explain the why behind the proposed adaptations, so that we can make better decisions with regard to accepting or skipping them.

Or maybe the option to accept or skip should just be removed, and the AI trusted completely. :thinking:

28 Likes
2

Hey! My turn to reply now :grinning_face_with_smiling_eyes:! Obviously I’m still new to all this myself but I’ve watched a few videos on TR… il paste two … hopefully these will help Answer some of your questions .. unless you’ve watched them already. They certainly helped me ( trust the process)… time will tell :wink::grinning_face_with_smiling_eyes::ok_hand:t2:

2 Likes
3

Yup, I’ve previously watched both of those. And while there is definitely some interesting information in both of them, this still doesn’t change the fact that when adaptations are proposed the user doesn’t really have any idea why this is happening. They might be able to make some guesses, but the AI clearly (well, hopefully) has a reason and I think it would be nice to get some insight on this.

7 Likes
4

Agreed. There’s something really important for the team to note here. I think people find it easier to commit to things when they understand them, as well as simply trusting the authority behind them. (Think of the importance of a physio’s explanation of a rehab programme in increasing compliance.) My suspicion is that training outcomes would improve with even a rudimentary sense of why changes are being made, because compliance and commitment would increase.

12 Likes
5

The problem with AI is that we don’t understand why they do things, so what you are asking for at this point is beyond the current state of the art. AI models have millions / billions (not an exaggeration) of parameters, which they use to generate a response. These parameters don’t correspond to anything we would understand, they are just variables without meaning to us.

I get why you want the model to tell you why, it just isn’t going to happen

4 Likes
6

Why??? Just kidding.

7

I’m sure you’re right. It doesn’t think as we do, it can’t reason like we do, so it can’t explain as we’d like.

It didn’t take long to get to the limits of AI, did it?

1 Like
8

I will take all your feedback to the team!

I can definitely understand how “trusting the process” can be a little hard sometimes without that extra info you mentioned, but it’s good to note that you mentioned that you see the advancement of my Progression Levels and FTP. So this is a great indicator things are working :slight_smile:

Just some info, TrainerRoad looks beyond just Workout Levels, factoring in your fitness, recent training, and overall trends to pick the right workout for you that day.

Think of your training like climbing a staircase. The goal is to keep moving upward—but not every step needs to be taller than the last. TR adjusts the step size based on what you’ve been doing lately, so you can keep progressing in a way that supports consistency and long-term growth.

9

I should probably have been more clear in that, while I have seen the progress, the adaptations I have received to this point seem to have been in line with how I felt, and were generally advancing. So now, when I feel good but some of the workouts seem to be regressing (at a point when my training load should be at its highest), it just causes some confusion. That is why I feel that having some insight regarding the adaptations would be helpful. Sort of a ‘what are you seeing that I am not’ sort of thing.

2 Likes
10

Never mind.

11

Of course it can and will happen. You just need to implement it the AI way. For example the AI first comes up with the answer (the TR workout adaptation) and then from there can derive a plausible reasoning backwards and present that. @Caro.Gomez-Villafane

Here a paper about the reasonings inside LLM is discussed:

3 Likes
12

I think AT is sometimes a bit too quick to degrade progression levels. After a recovery week and based on how your training week is laid out, AT can degrade your threshold level, even if you hitting personal bests on long suprathreshold V02 and sweet spot. I may be wrong, but I think there are some hard wired, time based PL degradations built into AT that aren’t being overridden by AI.

1 Like
13

@Caro.Gomez-Villafane tangential to thus discussion, after today’s workout (which I rated as Very Hard), TR suggested adaptations for each of the next three days. This made some sense to me, as today’s workout was definitely taxing, so some adaptations seemed like they might be needed. FWIW, the Wed and Fri Endurance workouts were adjusted down, while the Thu Endurance workout was scaled up a bit.).

I did not immediately accept these adaptations, as I wanted to ponder it a bit. However, when I went back to the TR app a couple of hours later, it was still suggesting the same adaptations for the next 3 days, but was now also suggesting some additional adaptations: one for next week, and another for the week after. The two additional adaptations weren’t anything major, so I accepted all 5.

Then, just a bit ago, I opened the app again, and now there is yet another adaption for THIS Saturday.

What the heck? I get that I would receive adaptations after a workout, but why would there be TWO subsequent instances of additional adaptations before I have done any additional workouts? I’m aware that the program will adapt as I go along, but why is is adapting multiple times between workouts? I would think that there should be one set of adaptations presented after a workout (or after some time off) when necessary, but it seems that should be the end of changes until after the next workout (or missed workout).

Can you provide any insight as to why this happens?

3 Likes
14

I’ll add in my current example as well. I did a Vo2Max with a WL of 6 yesterday, no yellow or red day. Plan has endurance today and Anaerobic 4.6 tomorrow. Saturday would be a 4.2 WL Threshold, but an adaptation has popped up revising that down to WL 2.6.

My PL in threshold is still at 3.4, so is that the AI in the background overriding my PL?

1 Like
15

@NigelTufnel11 let me check on this with the team and get back to you. :slight_smile:

@Jesse_Vernon1, AI is not overriding your PL :slightly_smiling_face: As I mentioned, TrainerRoad now looks beyond just Workout Levels, factoring in your fitness, recent training, and overall trends to pick the right workout for you that day.

Think of your training like climbing a staircase. The goal is to keep moving upward—but not every step needs to be taller than the last. TR adjusts the step size based on what you’ve been doing lately, so you can keep progressing in a way that supports consistency and long-term growth.

That said, I took a look at your account, and it looks like your Threshold PL has decreased to “Training Zone Inactivity” for over 2 weeks, so it can make sense that your next Threshold workout is a bit lower than your current PL.

2 Likes
16

Ah, I figured it might be something like that. I think the guidance you just gave me is just what the OP is asking for. Thanks.

2 Likes
17

Yeah! I agree we could add a description of the why as well :slight_smile:

6 Likes
18

@NigelTufnel11 update!

Just heard back from the team, and all is well with your account adaptations.

A good analogy for how the adaptations workouts are the weather forecast. As you get closer to completing workouts, Adaptive Training will adjust them if needed.

2 Likes
19

The confusing thing about this is that TR has literally no additional information. When the weather forecast changes, it’s because we’ve taken additional readings and have updated the model with that information. Having TR make changes like this feels more like tossing bones than making a forecast.

3 Likes
20

@Caro.Gomez-Villafane I’d like to add another scenario for which understanding the why behind adaptations would be helpful:

Yesterday I accepted adaptations that changed today’s workout from Endurance 1.7 to Endurance 1.8, and tomorrow’s workout from Sweet Spot 8.2 to Sweet Spot 7.2. The latter change was a bit frustrating to see, because I feel like I’ve been doing well on the SS workouts, and was looking forward to the 8.2, but I’m trying to ‘trust the plan’, so I went ahead and accepted this change.

After today’s workout, though, I have an adaptation to change tomorrow’s workout from Geiger +4 to Yarbrough. Both of these workouts are 2 hour Sweet Spot 7.2, with TSS of 131 and IF of 0.81.

So, why does the AI want me to do Yarbrough over Geiger +4? This is further complicated by the fact that I just did Yarbrough a couple of weeks ago, and I’ve also done Jumba Kang, James, Slatna, and some others multiple times in the past month, so the monotony is getting to me.

I really do want to ‘trust the plan’, but when I start seeing things like this, and have no understanding as to why, I really start to question it.

4 Likes