I think Nate and Jon just avoid all the toxicity that started brewing. Probably a smart move.
I’ll lay out the welcome mat for you to chime in more often… I enjoy your posts and sense of humor! Sometimes I sit here and ponder “what would Ivy say off the record, over a beer or three?”
6 Likes
In podcast #308 Jonathan specifically mentioned (@27:12) flagging up to support for them to double-check any workout where the Level feels like it might not be right.
Although I understand your reasoning, the “tight binning” would not necessarily be an issue (purely speculating here) if the model performs well. That is, generally, when there is a large model fit compared to low error or residual variance, a model can be more precise and thus tighter binning is, statistically, permitted.
In addition to this, a small amount of input parameters do not necessarily lead to bad model fit. It all depends on the strength of the relationship or predictive power of those input parameters, and the resulting explained variance of the model. 50 input parameters may collectively explain less variance than 2 input parameters.
Having said that, I completely agree that people may have differing power profiles, which may strongly affect the usefulness or accuracy of these profiles. For example, I assume that for a typical “endurance rider”, VO2 max workout of “bin” 6 is (at least perceived to be) harder than for a “sprinter”.
I wonder whether it would be technically possible to estimate these profiles from the data they have (e.g., seeing that particular people particularly fail endurance, threshold of vo2max but not the other types).
But I guess (and I assume that this is what you are getting at) is that when AT launches in its full glory, this is taken into account by updating trainings automatically based on your individual past performances.
I actually welcome it. Gives me a little breathing room to get ramped back up and into the rhythm. That first week back after a recovery week can be really brutal if the levels were the same as the prior week. At least that’s always been my experience in that first week back.
I’ll also note that while those levels have dropped for right now, they aren’t set in stone. They’ll update after tomorrow’s workout and Thursday’s and Saturday’s and Sunday’s and next Tuesdays’ and next Wednesdays, etc. And if they don’t, then the progression is probably in the correct ballpark for where I’m at right now, and I’m fine with that.
Is it possible some gains are left on the table? Probably. I think that will be the case with any plan I do. And I’m not concerned nor do I care about that level of minutia and detail. I feel I’d benefit more from putting energy into tactics and race craft than into getting that last 1-2% from the plan. Diminishing returns and all that.
1 Like
Maybe we’re misinterpreting each other.
My position is tight binning helps with a model, but tends to reduce generalizability.
Also, the fewer parameters you can use as inputs the more general and robust a model tends to be. The issue often is finding the low dimensional input space that is sufficiently descriptive.
Nicely done!
Did your workouts get adapted after this ride?
1 Like
Is there a breakdown of what each of the survey options mean? I did a VO2 max workout tonight, one of the Gendarme variations, and I selected hard as the option because it was VO2 and it was hard. I was able to complete the workout and didn’t miss any intervals but my plan tried to adapt to make my next VO2 workout easier. I declined the adaptations. Should I be making a different choice if I complete the workouts with no interval failures?
2 Likes
Ooh, that sounds like a big bug, for my use at least. My schedule changes week to week so even within a block my workout days are all over the place. Could be M/W/F one week and T/W/Sa the next so I’m always having to swap around my workouts. Hope they can get this fixed before I get in (so they have tons of time 
).
2 Likes
Sounds like you got a success survey. Why it is reducing levels for the next workout depends entirely on the workout, where you are in your plan, what it saw in your performance, what your goals are, and so many other factors we simply do not know because it is proprietary.
Personally, I’d have probably accepted the adaptation. Assuming the next workout in that zone and of that style is in a week, it could adapt 2-3 more times before you get to it, and could even increase in workout level, so what it told you today could be vastly different from what it tells you on Sunday.
1 Like
edit: boom=6 watts per kg good!
Ok to follow up on my backwards-progression saga:
Support got back to me and said nope, this time it’s on purpose:
I just took a look at your Progression and Adaptation history, and according to your recently completed workouts, your calendar is progressing as expected. It’s worth mentioning that not every workout is going to be harder than the last, and we have some logic behind the scenes that monitors your workouts and surveys and uses that information to influence our recommendations.
I agree that the experience where future workouts appear to update, while your immediate workouts do not is a bit confusing though, and I will definitely pass this feedback onto the team for investigation.
I’m reading that as permission to not kill myself with over/unders today, can’t say I’m sad about it 
4 Likes
No. had a chat to support. the adaptions will only work on outside rides if you load a workout to your head unit and then do that workout. While the created workout does give a view on what my level may have been, because it was done after the event and matched to a ride the adaptions wont update… yet…
How is this accomplished please?
3 Likes
Thanks for the update. That is some very useful information for everyone here!
At least you know now that you have a sweet sweet spot ride as your local loop.
Thank you
1 Like
So I wonder what happens in this situation.
I shuffled my 5 workouts for this week manually to Monday to Friday. I’ve completed 2 of them. If I now move them back to the original days and then do the change as per the description above will it
- Retain the recognition as far as AT goes
- Change the days of the week for all future weeks in the plan (I assume it will)
For #1: When I have used the block annotation to modify or update my plan, I get the standard plan initially, and then a new round of adaptations to restore everything. I might not get exactly the same workouts, but I’ll get whatever AT thinks I should be doing (which is what I’m after anyway).
The workouts you shuffled manually are not part of the “Plan” anymore, so they’ll probably stay where they are, and Plan Builder will make a new set. You can delete the manually-shuffled ones.
For #2: For that block, yes.