To me this looks like as one of stepping stones toward ultimate goal:
The updated TrainNow feels like a stepping stone and for me, fills any void caused by my unstructured outside rides not contributing to PL scores.
If I want to quickly calibrate my indoor plan, after mostly riding outside, I can throw in some TrainNow rides and if the ML has a suitable prediction for me, update those PL numbers.
This in turn adjusts the future of the plan AND in a way where I’ve proven that I can achieve what is asked indoors. As below, ML has determined that, on that day, my Sweet Spot was likely to be ~5.x and I’ve verfied that. Perfect.
6 Likes
Did TrainNow actually recommend that workout? That sort of jump in PL would be ranked as a Breakthrough workout. Pretty sure I’ve never been recommended a Breakthrough, though have manually picked them myself when I’ve been confident that I’ve done non-TR outdoor training which means my PLs aren’t a good reflection of my current capabilities.
Don’t know if this specific workout was recommended by TN but in general yes, this is how it supposed to work now, see:
4 Likes
Yes, I was cautious at first (thinking it was a bug) but was told this in another thread, went ahead and nailed the workout:
We updated TrainNow to use a machine-learning model to recommend workouts. Probably best to reference Nate’s post for details (linked above), but the TLDR is TrainNow is using a new machine-learning workout recommendation engine that the data shows will more accurately predict workouts that you’re capable of, regardless of your current PL.
TrainNow is recommending wildly different workout levels - #19 by nick
From where I’m sitting, this is as good as, if not better than the idea of WLV2 and unstructured rides contributing to the PL score. Presumably it is using unstructured rides in the determination but you have to verify the predicition to update the PL.
5 Likes
I love the idea of the updated TN and am hopeful that it’s a big improvement after Nate’s comments about it actually being better than AT! Can you explain the bit I’ve quoted above? I’m not following how you see these improvements to recommended workouts making up for not taking unstructured rides into account.
Indirectly: you do whatever work you do → your personal dataset is updated and taken into account by TN → you accept seemingly crazy TN recommended workout from official library → you’ll get your dearly desired PL update 
4 Likes
Doesn’t that mean that it IS using unstructured rides then?
I’ve been doing very very low volume sub-threshold work for four or five months and this week did a 3x10 to see where my FTP is. My V02, anaerobic and sprint PLs are 1.0 as a result.
Trainnow today recommends a level 10.0 (yes, ten) anaerobic workout, which would be +9. You think that’s a bug or maybe I might be able to do it?
1 Like
Don’t really know but if unstructured rides are taken into account for AI FTP, only conclusion can be that TN has to take it into account as well (probably with some delay, not immediately after completing ride or even next day).
They are.
Only after 10 indoor tr rides though right?
Correct
1 Like
Why ? They are two completely separate systems that do different things, I’m not sure why it can be “the only conclusion”
1 Like
True, that’s why also used words like “indirectly” and “probably with some delay”
you didn’t use the word indirectly
and the probably is about a time delay … not that the are probably (not) using outside workouts in TN
Excuse my word salad.
To clarify, I’m not fussed that unstructured don’t make the visible score go up. I’m happier that the ML does consider them and that once I’ve completed a TrainNow ride and verified that I could complete what was set out, then the score goes up.
2 Likes
Discussion started bunch of posts before that actually, with little bit more context.
Anyway, there is no point arguing one way or another. This is just us, ordinary forum users, trying to interpret what CEO or members of TR has posted in various topics.
3 Likes
Gotcha! So, if you go do an outdoor ride with some long SweetSpot intervals (for example), then TrainNow will take that outdoor ride into account, and feed you an appropriate SS workout based on those efforts. Assuming you then nail that workout, your SS PL goes up. Did I get that right? Pretty cool.
2 Likes
Exactly, that is my hot take without knowing what’s in the secret sauce.
I’ll continue to check TrainNow after rested days, try and see if I can complete other PL zones with big jumps and see if this theory holds up.
3 Likes