🎉 🎉 🎉 Introducing Adaptive Training! 🎉 🎉 🎉

If a completed workout progresses any of your levels besides the primary, it’ll show you that section. It doesn’t happen on every workout.

FWIW the issue we were discussing seems to be fixed now:

image

2 Likes

Very nice post. I have been following Melanie’s work on analogies for years now. A decade ago I even met her PhD supervisor, Douglas Hofstadter during a dinner.

Working myself on AI (well, Computational Linguistics using ML) I can assure you that it will take some time before we have something meaningful from TR. It seems to me that the backbone should be fatigue modeling, but I might be wrong. In any case, self coaching will probably be with us for some time, although tools like adaptive training could somehow be of help.

Brill, thanks @ValeCyclist

new vlog, woke up today and saw pending adaptations to SSB2 HV that is starting today, just quickly walked through each week showing which workouts are changing. Last weekend of SSB2 will have me doing 9.1 and 10 level workouts, so that should be fun lol

Should note that I have a playlist of all my other vlogs, I think a lot of people watch one or two and might miss the entire context, none of these are really intended to exist as a standalone and are meant to highlight an evolving process, which AT is all about

2 Likes

Wasn’t the issue with reductions rather than increases? I’ve had multi zone increases in the past with no issues.

Apologies if this has already been mentioned in this gargantuan thread and I missed it, but one of the things I’m really liking about AT so far is that it’s leading me much deeper into the workout catalog. This keeps things a lot more interesting. Now if I could just get the workout to adapt on-the-fly…

9 Likes

I’ve thought a lot about on-the-fly adaptation as the next level of “intelligence” in workouts. I fear there are some days that I would attempt to destroy any computer that made me suffer more than expected on the fly…but built-in sandbag protection might actually be a good thing.

2 Likes

Original:

I see what you mean now. They have modified / hidden those numbers on the original ride for you.

AT has figured out my VO2max is lower on Tuesdays, and higher on Thursdays… :thinking:

image

3 Likes

Something is definitely broken. In PC desktop app survey appears and I submit the response, but outcome in website still shows Missing response. On Android app however I don’t even get to submit a response as survey doesn’t appear. When I tap “back” after the workout, app ends up with an empty screen with a spinner in the middle. Have to then kill it using recent apps screen or exit via home button.

Thank you for this paper. For someone outside the field, it is interesting and revealing reading. The question is whether training is more like chess, which machines are good at, or charades, which they are not good at.

One obvious conclusion to me is that we should not expect AT to exhibit “common sense”. In TR world, I suppose that means that AT might draw conclusions from our data that are unrealistic. One example might be progression levels that are too extreme.

Another is that developers and users need to be on the lookout for adaptations that reflect simple correlations rather than rely on an understanding of causes. In training, the possibilities for this mistake are endless, and already something that typifies our field. For example: do TiZs reflect an attempt to be polarised or do they reflect the fact that after a HIT session people need a few days easy riding?

No doubt others among you can think of more profound conclusions.

3 Likes

AT going a bit crazy with my VO2max was one of the reasons I stopped using it. It wanted me to do a 5.0 today though my level is supposedly 5.9 (was 6.9 as recently as two weeks ago).

I did a 5.8 as a compromise and to be honest it probably wasn’t as hard as it should have been (I don’t think my HR made it into VO2max levels). One bright side is that it now isn’t offering me any VO2max adaptations (it’s still insistent on dropping my sweet spot though).

Edit - I should note that I kept the types of interval the same. It was pretty consistent on wanting to drop all of it.

1 Like

Those are pretty different VO2 workout profiles though. One is short shorts and the other is more traditional intervals. From what I can tell, AT maintains an understanding of how you perform in each of the subcategories of each zone and serves you workouts that fit your abilities with respect to that information. I’ve noticed this with my own VO2Max prescriptions from AT over the past several weeks. My VO2Max level is actually lower for the longer intervals and higher for the short shorts, which I also know from experience aligns with ability to succeed in those workouts. I’m certainly better at one more than the other, so to see AT actually recognizing and adapting for that is quite cool for me (and it’s been bang on).

All that is to say that I don’t think what you are sharing will be uncommon for those using AT. It happens with a couple of the zones, but most notably happens with VO2Max and Threshold workouts (both have 12 subsets when you filter by them in the workout catalogue) as there are several ways to elicit the training adaptation for them.

3 Likes

That Tuesday swap looks like a case where the progression level might be too high for Ansel Adams / too low for Baird. Maybe AT is smarter than the progression values and is giving you a harder one :slight_smile:

Baird is a much tougher duty cycle (60/30 vs. 50/40 is a big difference), and shorter rests between sets. The only thing you can say for Ansel Adams is 135 vs. 120% target, but I think I’d still find it easier with the much more generous rests.

I disagree with the assessment that Baird is harder. Ansel Adams is more intense across the board by all metrics, despite the increase in rest time. It’s IF is still .91 with that increased rest. It even has less endurance time at the end because the rest between sets is longer too. That makes those sets just that much harder.

Never mind - tonights ramp test increased FTP a measly 2%, and VO2 progression went from 5.4 to 2.3, so I guess another round of adaptations is on the way…

Just did Ansel Adams and :cold_sweat:

1 Like

Something like that. Here’s what support told me (and probably @trpnhntr):

We did issue some fixes on the AT end (which, this display issue was a part of) yesterday, so moving forward, we shouldn’t run across this specific progression display being incorrect bug :slight_smile:

3 Likes

Did you report this to support@trainerroad.com? The survey prompts very inconsistently for me on Android 9, and support has been working on it. If it doesn’t prompt, you can press the ... button in the bottom left and choose Update survey response to fill it out.