You’re right! Doh!! I fixed my post above.
Thank you for sharing this!
Congrats - big step forward. I eventually made my way through the podcast - a question for which I haven’t heard/seen the answer. If I push a workout from my TrainingPeaks account (whether it’s a training plan workout, a coached workout, or a custom workout) is the ML able to classify that workout? If say, I have a 6 x 3 VO2 workout scheduled, but life happened and I am too tired for the workout, will AT be able to offer an alternative workout?
I’m confused…
Added to the Beta last week (awesome, and thank you). Earlier this week I had Baird -1 programmed in for Tuesday and it was rated as productive. Makes sense, I’m in base and always avoid dreaded VO2 work. Managed my way through it and post ride it’s flagged as achievable and no movement in my profile that I noticed.
I’ve just finished Tray Mountain which was again flagged as break through but sits in my history as achievable, though does appear my sweet spot score has improved (didn’t note it pre workout).
Is it the case that if you complete the productive workout as prescribed it’s reclassified it as achievable in the history but the progression level bumps up?
And also, I use iOS to run TR, but seem to be on the main release, would I be better of getting into the iOS beta to get the most out of AT?
I assume in the mainline iOS app you don’t see much related to adaptive training? I’m on the iOS beta and it seems fairly well integrated there.
I’d guess viewing workouts would show rating per current ability but would also hope the historical review / analysis would show the “then” ability…
I’ll check mine when I get a chance and see if I notice anything weird in these areas.
I checked a recent ride and in the career page it shows Productive (and when I completed it gained 0.7 in the target area). Pulling up that workout in the workout list shows it as Achievable now with the new progression. I think this all seems correct to me.
No it won’t classify that workout. But if you export it from TP and import it to Workout Builder, then as a custom workout it would be classified (in time, not today i believe)
Yes still mainline iOS app. Zero AT integration there which is to be expected. I guess I should move to the beta app to see the better functionality and be able to better provide feedback. Is that still a closed beta?
The beta is open, see instructions on first post here:
I’m in! Looks pretty slick so far.
The evil ML just gave me a harder Saturday workout compared to what I had. They are on to my sandbagging already. Damn.
Definitely looking forward to getting in. Lately been riding outside as the weather in GA is glorious. Outside of that a few meeet RCC group rides on Zwift as I don’t have a plan set yet.
As soon as I get an invite I will do TR workouts again on my Neo bike which I get next week. Replacing a N2T and Saris rocker which are solid performers and will sell at some point. I just want a simpler setup.
Future AT Feature Request: Please tie AT into my Door Dash and Uber Eats accounts. I really need to be punished and shamed for my poor decisions.
TR is using directed learning to generate the rankings. They went through and rated a subset of all of the TR workouts, pointed the ML at that set to learn, and then unleashed it on a 2nd set they had rated without exposing the ratings of this 2nd set to the ML. You compare how the ML model rated the 2nd set to the human ratings, a tweak the ML model as needed. And then when you are happy with the ML model, you let it rank the rest of the workouts.
This will allow TR AT to rank custom workouts, and feed those custom workouts into its progression ranks and workout recommendations
So I did a 2.5 hour Z2 outdoor ride today, and about 2 hours after I got home I received the beta AT invite and of course immediately activated it! It suggested some changes to my workouts this week, which I accepted. It made my Saturday ride harder than what was scheduled for my MV sweet spot plan, which is fine, and changed my Sunday ride a bit. The strange suggestion was to do a ramp test today though! I’m wondering why it didn’t recognize that I did a 2.5 hour ride already today?
One other user noted it added a Ramp Test right away to their plan and Brandon with TR said it shouldn’t do that, maybe a bug.
I had just done a Ramp Test last week, perhaps that prevented the algorithm from dropping a test on my calendar?
It makes sense that it’s a bug. Since I deleted it, I wonder if it will try to make me take a ramp test next week.
I had the same thing happen. It isn’t supposed to suggest a ramp for you in the middle of a block like that from what It sounds like. Here’s what Brandon and Nate said.
I’m not sure if it’s a bug in the system or maybe I’m just misunderstanding something/reading it wrong…but…
My Plan Builder session for today is down as being Andrews (1:30:00 - 73 TSS - 0.70 IF) as it stands, this is showing as being a Recovery ride for the difficulty level.
If I go to Train Now, it throws up an alternative session of Townsend (1:30:00 - 60 TSS - 0.63 IF).
Now, Townsend is clearly an easier ride than Andrews yet the difficulty level for this is showing ‘Stretch’
Am I missing something or should this theoretically be the other way around???
All my levels are set at 4.0 - do I need to wait for a bit for it to categorise all my previous workouts or do a workout before it starts learning?
There is another way to look at this. The early users are getting their shiny new toy now. They are having to help resolve the bugs and glitches. If you’re currently training for a confirmed event, do you want to be on a plan/platform that could still go sideways?
I’m one of those people who gets very frustrated when companies release substandard software (I’m looking at you Adobe!) and you’re forever downloading ‘updates’. Personally, I’m happy to be the last one on this Adaptive Training bus. My expectation is that the platform performs almost flawlessly by this point. I’m not interested in half-baked.
Why is that bullshit? TR will have engaged with lots of people from a broad range of demographics to test, find bugs, issues etc. some of those NEED to be new users as their experience and feedback will be totally different to that of existing users who are familiar with the platform. It just makes sense and means that when you do get your hands on AT it will be a better experience rather than potentially having bugs or issues which might derail your training. Patience is a virtue.