Yes, well said. There is currently no documentation or onboarding, so it’s not clear whether a “+0” result is bad or expected
We are supposed to get support docs at some point. I will say that the results of the PL changes are subject to the Workout Level as compared to your Progression Level at the start. One higher than the other leads one way, equal another, and the reversed order leads a third way.
Much of this is essentially summarized in the “Difficulty” rating you can see in certain pages. Achievable, Productive, Stretch and Breakthrough will all be important. Then added to those ratings is your subjective feedback from the survey, and of course the actual workout data.
TR is leveraging all those data points to determine what change (if any) to apply to the related PL’s. It’s a lot of moving parts and we often oversimplify our expectations by forgetting the many factors playing in to that PL result.
Yeah, that’s what I was getting at when I commented on “positive adaptations” above… there’s a progression built in, the workouts already get harder. Just because you’re not getting adaptations to make the workouts harder after every session and all your PLs aren’t 10, that doesn’t mean you’re somehow a failure.
2 Likes
That’s not how it works though… you rank a workout hard, and your PL will match the level of the workout afterwards. So if I am threshold 4.8 and I complete a progressive 5.2 workout, rank it hard, my PL is now 5.2. My next scheduled workout is already a Threshold 5.7 as that’s the default progression, so no adaptations are needed. That is how it works in practice.
Whereas if I rank that 5.2 very hard, AT says hang on a second, this person probably isn’t ready for the 5.7 next week so we’d better give them a slightly easier workout. And vice versa, if I rank it moderate (when AT expected it to be hard) then perhaps I can do something harder than a 5.7 next time and I’ll get a 6.0 prescribed instead.
This is exactly my experience with AT. Ranking progressive workouts 3 - hard does not make AT back off - I guess if you started ranking workouts that are supposed to be achievable as hard then it probably would back you off.
4 Likes
Ah, interesting! You’re right, my PL now matches exactly the level of that workout.
Does the same thing happen if you rank it “moderate”?
Well, n=1, but I had a VO2 4.9 scheduled Tuesday. I found it easy for a VO2 workout, my HR was low threshold and it didn’t feel VO2-y. So I ranked it Moderate.
My PL is still 4.9, matchimg the PL of the completed workout, but all my future VO2 workouts were bumped up (the next one is a 5.4 - I can’t say for sure what it was before the adaptation though, I would guess 5.2).
The part we don’t know is if AT has an internal expectation for how you should rate a ride and is comparing your rating to its projected rating, and then factoring this into how it adapts your rides.
For example, if you have a stretch workout, my guess is that AT would expect you to rate it as Hard or Very Hard. So if you rated it as Hard, AT might count this as successful and increase your PL.
On the other hand, if you rated an Achievable work as Hard, AT might interpret that as a failure / need to scale back the workout intensity.
Just an intriguing thought, and would be a reason why we are being asked to rate how the workout felt in the absolute sense, and not in relation to how we thought it should feel
1 Like
Yup, that is correct AFAIK and what we’ve been pushing as one aspect that pure “feeling” is what TR wants to know from us.
1 Like
My experience so far is that is basically how it’s working - I rated a Stretch workout 3-Hard early in my last block and got a big bump in future workouts (note my PL always seems to match the PL of a successfully completed workout - but the future workouts are adjusted differently depending on how I found it).
There’s nothing wrong with ranking things 3-Hard. There’s a progression built in - the workouts get harder week on week - stands to reason that a workout you want to be Productive should feel hard.
I get the impression people are going out of their way to rank things 2 - Moderate in order to get further bumps in workout difficulty.
4 Likes
I reckon you’re right. I think people are gaming the system and then things will get too hard and they’ll blame AT. 
3 Likes
The TR crew can’t control the users ego, if someone keeps rating workouts easy and their PLs go up, when in reality it might be too much for them. So now when they blow up it’s user error to an extent. I think AT is in the right direction for personal progressive overload but yes personal ego or “toughness” the TR crew can’t and may be hard to account for.
4 Likes
A question for me is whether people gaming the system to get harder workouts somehow figures into the ML algorithms overall and could have a negative affect on the quality of AT’s recommendations for the rest of us. That would be kinda sucky if it did and all AT users are subject to the behavior of other AT users. @IvyAudrain, could the TR team comment on that?
The ML model is much more complex than lump averages for behavior+outcome. Dont worry. 
4 Likes
I frequently remind myself that it’s the hours and effort I put on my legs and heart that make me faster, not the numbers I log in TR/TP/Strava. Those SHOULD correlate well, but my focus needs to stay on working hard, training smart, and sticking with a plan.
3 Likes
Sure, but I would argue that when every Threshold or VO2max workout is hard or very hard for me, then it shouldn’t be making future ones easier after I’ve marked one as very hard.
If I find either of those sessions “Moderate” then something is likely wrong.
But that’s the whole idea of AT. Making sure that the plan is suited for you. If you’re good at Vo2/ss/TH then it should bump those workouts to make sure you get optimal training, without making workouts that suit you less, also getting harder.
1 Like
So, maybe for any workout AT expects a corresponding rating based on its difficulty. There are five levels of difficulty and five survey response ratings.
Recovery = AT expects 1 Easy
Achievable = AT expects 2 Moderate
Productive = AT expects 3 Hard
Stretch = AT expects 4 Very Hard
Breakthrough = AT expects 5 All Out
Ratings other than the expected response adapts future workouts accordingly.
Just a theory and haven’t tested.
3 Likes
This might be a dumb question, but is the progression in levels and adaptation in workouts only depended on how you score them after you’ve completed it?
Does it not take into account if you had to hit pause, cardiac drift, drift in power if not using ERG and so on?
It absolutely looks at the data within the file. It is using power for certain, and may be leveraging heart rate or cadence (now or in the future?) per some fuzzy comments I have seen.
It will know if/when you pause, backpedal, alter Workout Intensity %, etc. That data can and does trigger a “Struggle” if you don’t follow the workout closely enough, or tweak it excessively.
TR AT is looking at the planned workout, your execution, and your subjective feedback in the end, to make the future changes to PL and any adjustments to your plan.
3 Likes
But all that can be done without AI, if just considering each workout in isolation. And PL progression can be simply done, without AI, based on completed levels and survey answers. I’d like to think there’s more to it than that. Something along the lines of “well, you might of completed a stretch workout today and and only given it a ‘Hard’, but your training history shows that if you do 4 stretch workouts in a 10 day period you normally implode, fail your next over unders and drop 2% on your next ramp test. Because of that, I’m going to throttle you back, even though you think you’re nailing it”