Of course, a lot of TR users are going to ride outdoors. How many are regularly doing structured workouts above Z2 outdoors? Iād wager it is a small minority. But you are right, I have no data to back up my claims.
Maybe you are amongst a different cohort then. In my experience it was more mixed. Those who did not use TR tended to be less disciplined, too, e. g. theyād be the ones doing Zwift races. Or, on the other hand, they had a (human) coach. Iād just caution against āeveryone I know who is serious ā¦ā, though.
It could also be a self-reinforcing effect where TRās lack of data leads to worse predictions for athletes who want to train outdoors, which, in turn, leads to these athletes moving to a different platform.
I still think youāre way off thinking that a small minority EVER do a structured workout outdoors, but staying with the point that
[/quote]
Riding outdoors ā training outdoors
[/quote]
Do we have anything from TR explicitly stating that this ONLY happens if you do a structured workout outdoors? Did it have to be within the 30 day window? Are you susceptible to the bug if youāve EVER done a workout outdoors? What if you only do 1 outdoor workout? What about if you do any riding outdoors at all? What if you do a ride outdoors and then associate it with a workout? Does it impact every single rider who rides or trains outdoors? Etc. I know they keep saying itās a bug to do with people who train outdoors, but thatās all we have. Iād be shocked if Jonathan never trains outdoors these days. Does it impact him? If yes, why wasnāt it caught? If no, why not?
We see lots of people posting about the issue, but for what feels like months now weāve been told it impacts people who train outdoors and a fix is incoming. I donāt actually believe this is 100% correct (not saying theyāre lying, but that theyāre talking about one specific issue, not ALL users whoāve seen massive predictions followed by massive drops). Itās possible that every single post about the issue over the last few months is ONLY people who recently did an outdoor workout, and that they will ALL be fixed when the update goes live, but Iād be surprised if thatās the case.
I think you missed the word regularly in my post. Thatās very different from āeverā.
It also misses the bigger context: if the percentage of outoor workouts apart from endurance work is small, you have a much smaller data pool. It is even smaller if you filter for atheletes that do a significant share (say, more than 50 %) of workouts outdoors.
I have the impression that some are taking what I think is a statistical fact of TRās data set personally, as if it means āYou are doing it wrong.ā Iām saying neither. In fact, I hope TR is able to improve also its outdoor workout functionality. I also want those bugs fixed.
I think @Jonathan selectively tests in-development features. For a while he was doing Polarized, then a lot of outdoor training, then a triathlon. He is (being) dogfooded TRās new features. But probably he trains more outside than the average TR athlete.
which implies this isnāt something that impacts many people. I think thatās way off. Almost every single post for the past several months has been about how AIFTP is off, way too high, way too low, predicting rediculous results, dropping the prediction massively prior to detection, etc. This is a BIG problem, not some niche issue impacting a minor subset.
Is it a BIG problem because itās a small % of a very large number or a because itās a big % of a very large number. A minor subset of 30k users is still a lot of unhappy people.
Iād typically do 1-2 rides outdoors each week with the other 4-5 indoors. That said, it typically comes up with these impossible predictions straight after detection before Iāve done any workouts and then marks it down after outdoor rides and some indoor rides as I progress through the block. One of my friends who I ride with weekly (on those aforementioned 1-2 outdoor rides) is also on TR and his predictions are spot on in the sense they vary very little over the course of a month and are close to detection. Almost like our two programs are in different ML/AI models as far as how they behave. This has been interesting as we are both close/at our lifetime peaks and have the same event on our calendars which is only a few weeks away now.
I wonder if the AI assigns the outdoor workout and expects it to be nearly perfectly followed like an indoor workout, but then it sees the result of the workout and lets say it was only 95% of what was expected so it thinks the assigned work was too hard, so it makes the next workout easier and then that workout comes back only 95% of what was expected so it makes the next one easier and it turns into a vicious downward spiral. It would be nice if the AI would learn that when I assign an outdoor workout this rider consistently only does 95% of what I ask so Iām going pick a workout that asks more and then what they end up doing is actually what I intended them to do. Jonathon is such a machine that his outdoor workouts arenāt much different than his indoor workouts so this problem never popped up for him.
The context was about the depth of the data pool with regards to people who train outside. And hence, the predictive power is smaller as it gets more difficult to generate representative data sets for a given cohort (e. g. of a given gender, age bracket, performance level, training volume). I want a better product in all dimensions, including outdoor workouts, I was just trying to think through why the TR AI is not as good for outdoor workouts (and other situations where the statistical base is smaller such as people with very high FTP, the latter a bug I still try to run into ).
PS I ride outdoors 6 days per week. My commutes are part of my training. Iām not doing intervals outdoors, but I do all of my endurance work outdoors.
I think thatās an important point that cuts both ways: if something works for 95 %, it can be a tremendous improvement if it initially worked well for 85 %. Yet at TRās scale that still means it doesnāt work well for a significant number of athletes.
All of my rides are outdoors, and I complete about 95% of my structured workouts. The latest iteration of TR has been unusable for me.
Iāve manually changed roughly 90% of my intensity days because the prescribed workouts are consistently too easy. Then, whenever I swap in a harder workout, TR reduces the following weekās workouts, even when I rate the session as āmoderate.ā This has been happening for around fivemonths.
Iām sure there will be people who say Iām doing something wrong or try to find a flaw in my setup, but the reality is that I didnāt have these issues during the previous three years on TR. Something has clearly changed.
I understand bugs happen, but if the problem is this significant and takes this long to fix, why not roll the feature back until itās working properly?
I donāt think you are doing anything wrong, TR currently has a bug/bugs with outdoor stuff. They are working on it but not sure if they have any eta on fixing things as of yet. @eddie probably knows more on how that fix is progressing.
It definitely is not a good issue to have, especially with more people training outdoors as weather improves.
Yeah, @AussieRider is definitely running into this bug. I think this could be a slightly exaggerated case based on how theyāre working through their plan.
@AussieRider, while you didnāt have issues going this route in the past, this bug is whatās causing issues right now. This is our teamās top priority right now, and weāre doing everything we can to get a solid fix out.
Iām not comfortable enough yet to give a specific ETA, but weāre aware of how urgent this is.
).