On top of that, I’d add. How can I rely on a system that doesn’t account for a ride I did “outside their system?”
It’s not the question of having PL 10 or 1 for Endurance. It’s the question of how this AI thing that they sell effectively adjusts your training load if it doesn’t know how to calculate it. It’s sort of nonsense.
BTW, same reason I cancelled my subscription.
No. You are confusing PL with AT. Adaptive Training is the feature that looks at your PL to choose appropriate workouts for you based on your current fitness as determined by PL. They are separate and distinct features, hence the separate marketing names, support pages, etc. Read that PL page I linked above. It clearly distinguishes between the two features.
PL is “broken” in the sense that it does not achieve its stated purpose (measure the athletes fitness, NOT TR training plan progress) for the vast majority of users. AT does the best it can with the data PL gives it.
And really, what’s the alternative?
Are you familiar with the concept of sampling? While more data is better, TR is getting pretty close by using specific tests performed under relatively controlled conditions, and the proof is in the fact that for most of us, the program consistently delivers workouts of the appropriate intensity.
Mind you, I want WL V2 yesterday, but even if 90% of my riding was outdoors, having accurate data on 10% of my riding would still get a pretty accurate analysis of where I was on any given week. Given that TR makes it possible to do workouts outdoors, it is possible to ride outdoors and still have the vast majority of your training recorded in a TR workout, which translates into solid data for evaluating your “actual fitness”.
It does matter for some training plans if your Z2 is not counted for some training plans. For example, if you’re doing polarized training or training for an all day event and AT thinks you haven’t done any / enough Z2 because your PL says so, it will keep trying to adapt your training to add more low intensity.
There are ways to work around this, I fully admit. But it’s certainly not the goal to have to do those tricks. TR product is part guidance on what workouts to do, part convenience of a system. That system is a lot less convenient when I have to game it to get good guidance. And then there’s people that don’t know they need to game it or how to do it.
Sure. But that requires a lot of futzing with the system. And how many outdoor rides fit into that kind of approach with TR?
Let’s take a very simple ride example. Say that I do a 4 hour ride with 3 hours of zone 2 and 1 hour of threshold/SS. This is the type of workout that a coach might recommend (Kolie has on his podcast) and JOIN gives.
Do I have to manually create/select two workouts, a 3 hour Z2 and a 1 hour SS? Which do I associate to the actual ride data? Then what do I do with the other workout, which will make my TSS artificially high? Or should I comb through all 4522 and growing workouts to find a close match?
What do I do if I’m racing every week? Good luck finding a decent match for that, even manually.
For many complex/variable outdoor rides, the best you can manually hack together is a set of workouts recreating a time in zone chart. But apparently PLs don’t work that way, TR must have some more complex logic looking at target vs achieved power, otherwise it wouldn’t be taking them more than two years to implement it.
And that’s where they went wrong IMO. However their way of assessing fitness through PL works, it should have been designed from the beginning with the right requirements, that being that it works fully on “unstructured” rides/workouts (i.e, TR had no data for what your power target was) just as well as it does on TR workouts. Then they wouldn’t have to manually finesse the PLs for each workout or take forever to make WLV2.
I think it was a mistake to try to assign separate scores for 7 zones. Fitness isn’t THAT fine grained. 4 or 5 would have worked better in my estimation. Endurance/Tempo, Threshold/SS, VO2Max, Anaerobic and Sprint. In the real world nobody is great at Endurance but not Tempo. Threshold overlaps Sweet Spot. Your fitness is a spectrum and there are no hard borders.
Given that this has taken so long, I’m hoping they did redo the whole PL algorithm, even for TR workouts. Otherwise they likely don’t have a robust solution that will be worth the wait.
Otherwise I think their competitors are going to start really eating their lunch by allowing people to ride outside as much as they want and actually interpreting that data well.
I used to live someplace very flat with many miles of rural roads. As boring as flat can be at times it was a dream for executing workouts.
It sounds to me like the issue is that you don’t agree with TR’s approach to training. That’s fine, there are lots of ways to train, and lots of coaches and other programs out there. But you’re making some assumptions that TR doesn’t and vice versa. You’re comparing apples and oranges.
Good luck with your training.
I hate the concept of “unstructured”. It’s lazy and wrong. Call it “unquantified”, because currently, that’s what it is.
You’re spot on there: They picked the wrong basic unit when they chose to classify structured workouts based on how difficult they are to complete. I’d imagine that was fairly low-hanging fruit for ML to take some human classified workouts to learn from and then assign PLs to a whole library.
Where they should have started was with an algorithm that took second-by-second power data and worked out how hard the workout was alongside all the other training metrics. That’s a bit more involved as a first step but makes the analysis of ‘unstructured’ rides as easy as feeding it the data, as the system isn’t looking at blocks of constant power within zones etc, it’s just calculating a number based on the power output, whether that’s proposed or recorded.
Unfortunately by going down the path they started on I think they’ve created a rod for their own backs - the system just isn’t that flexible. Imagine trying to describe a series of long outdoor rides in two words (zone and difficulty) and have a computer try to classify all of the infinite possibilities from that.
To me it’s little wonder it’s taken them so long to get where they are.
Mike
yup. Its a system built for ratcheting up difficulty of workouts. With ML assisting with the rate of ratcheting up.
Yep, I’m curious whether they are changing some of foundational structures/approaches in the current training system. If they are trying to squeeze free-form rides into the current “single” zone/level approach, that would be extremely limiting. My hope is that they could look at a 4 hour training ride and recognize a large block of Z2 while also recognizing higher intensity intervals. The current system does account for multiple zones in some cases, but I think it’s only when zones are adjacent and something can be inferred (ie - sometimes your tempo progression levels might be bumped a bit from a workout that includes a lot of sweet spot). Honestly, I just wish they would release something and then iterate to make it better. I’m probably in the minority with my tolerance of early stage alpha/beta releases, but bugs and limited functionality are welcome as long as it’s positioned that way. I assume whatever issues they are dealing with may be pretty fundamental/core to hold up release for this long, but hard to say for sure.
I would (And will/am) make the argument that, while WLV2 matters, almost everyone has missed the bigger point:
What the effect of WL is on the plan itself.
I have a video coming out in a couple weeks with a full deep dive on this, but it’s a hugely important distinction and understanding. Adaptive Training’s use of PLs to determine the workout is within the goal and scope of aligning you more closely to the pre-described plan. If you already have 100% compliance to a plan, AT via PL is borderline irrelevant. AT is not irrelevant as real life often impacts compliiance, of course, but the bigger picture is that the vast majority of us are still doing SSB/“General Base” as has been for nearly a decade.
“But Justin! Plans have been modified based on compliance level.” 1) I don’t see that as a positive, and 2) nor does it mean that the plan will be better for me. “Progression Level Go Up” does not indicate that the plan is optimal, nor does increased compliance, nor does AT adapt to me as a middle aged male vs an older male vs a younger pregnant female.
WLV2 and its integration of outdoor (or really, unstructured technically, inclusive of other indoor training platforms) does matter, and it needs to happen, but that’s still a major jump away from “This is the right workout for you.”
+1. The rides I really want it to work for are the somewhat structured ones anyway, and I suspect they’re the easiest to interpret. E.g. If I go for a 3 hour ride and spend >70% of the time in Z2 with a bit of Z1 and Z3 due to traffic, hills, etc, that’s got to be pretty easy to analyse. Same if I go do a loop with a bunch of 4-6 minute climbs done at VO2, with Z1-2 in between. Be great to get PLs from those rides without having to go through the hassle of picking/creating a similar workout and associating it.
The rides which I suspect are delaying the release of this are the totally unstructured ones where you hit all the zones. E.g. A smashy group ride. But I’m really not that bothered by getting PLs from those, because I’m not under the illusion that doing totally unstructured rides is a good predictor of my ability to complete structured workouts.
To know what TR workout you should do. That has been the primarily goal of PLs.
I am not. It literally says it is one of the goals for PLs in the documentation:
PL can help you get the right workout in multiple ways: either via AT where it progresses your workouts based on your success on completed workouts or via TrainNow or manual selection to help you decide which workout you should do. They help to decide between two workouts in the same zone.
Like I said I think PL have been transparently centered around adhering to a TR training plan.
And more essentially, TR has always been marketed primarily as a training app. Allowing a user to select a curated plan and giving them ways to measure their progress by adhering to that plan (AI FTP, PL, Plan Builder) and trying to ensure the workout is the best plan to use (AT).
I believe WLV2 is TR expanding beyond that: a new feature. Allowing you to analyze non TR workouts/rides against a proprietary system (PL) that was primarily designed for curated workout analysis.
I believe you are saying PL is ‘broken” in the sense that you can not (yet) get PLs for outdoor rides. I would push back and say that they were primarily meant to be used to rate the potential difficulty of a TW workout so their other features and systems, such as AT, know what workout to assign in AT or TrainNow.
Not sure what your question is? Is your question what alternative apps can do what TR is promising to do in WLV2?
If so I don’t know of any.
I am agreeing with you. Seems silly to quit a training platform that HAS delivered so much good but has run into a development snag (if that’s what’s going on). And if that’s not what is going then so what? Other features continue to be added and the platform advances. This is quite normal in the world of SDLC.
I have tested the waters outside TR and seem to find my way back each time.
Sorry. 100% agree. I would love to see the people who are complaining that WLV2 isn’t out yet and want to leave TR point to an alternative,
As said the last few posts, curious what you are using/doing after cancelling TR. if you really want custom, regularly updated plans, then hiring a coach is the way to go. But that is going to cost $200-$500/month. I’ve tried Join. It does adapt, but I didn’t find any magic in it. To me, TR is the best option out there outside of hiring a coach.
