So help me out here: I listened to the whole podcast. I heard they are releasing POL based plans, cool and great to have another option. What I’m unclear on is the machine learning.
So TR believes if you ride shorter you must ride harder to get benefits. I couldn’t really discern if the new AI is built on that same philosophy and how it is determining what the “best” ride for you on the given day is? Anyone able to clear that up?
I also gathered that when you’re presented with options my expectation is that many users are going to chose the tempo and above rides more often then not. From what I’ve seen most folks who are not coached go out and ride SST and above basically every ride or they don’t feel like they did anything. I’m curious in practice how that will work out with riders. I know I gain the most fitness from going out and riding in Zone 2, below a HR ceiling, for 2 hours plus and growing the time at that zone. This is within the context of a training plan with other sessions but Zone 2 is putting money in the bank vs spending it.
Will be interesting to see what people’s experiences are.
This is what Xert is doing, and if we believe that TR demands a lot over the longer term, Xert is next level. I trialed it while following my coach’s plan in the lead up to an A-event, no matter how hard or long my week of riding was on my plan, I was always in an XSS deficit with Xert. Ride 15 hours/week, Xert wants 20. Ride 20, Xert wants 22. I did tweak some of the settings, but even then, it was always very demanding. I hope if TR goes down the adaptive/ML path that they don’t build it like Xert did.
I just wanted to point out that trainer-road team addressed all of this in there very next podcast with there adaptive training machine learning. They even address the issue of if their data has a survivor bias or not.
Anyone reading this should listen to the whole Ask a Cycling Coach 298 podcast, they address all of this
I think if we look at what they’re showing here, their machine learning will show what energy systems you’re improving aka how many points you’re getting in each energy system.
The AI will probably have a whole lot more data from SST rides for sure,
But people put on a polarized plan will be able to provide data as well and that will be able to provide better conclusions as to what’s effective, what do people stick to and what makes you the “fastest”.
At least that’s how i feel/hope it’ll go down. Give people more options, study the data it gives you, adapt and overcome. Everybody is happy
The recent podcast reminded me of why I have never canceled my TR subscription, even when not using it. I appreciate the vision and the incremental improvements, I’m actually excited to see these new features in action. Great work @Nate_Pearson!
Interesting - do you know which section of the podcast this was in?
Edit: found it - turns out you can turn transcripts on in YouTube and search that in the browser, handy.
Starts around 55 minutes for anyone else interested. As one of the people raising such concerns, I’m actually pretty happy with the response here by Nate. Over 50% of people not following plans is a fairly interesting statistic! But also sounds like it’s still being taken as serious potential issue, and they’ll remain cognisant of it going forwards.
Also was great to see whilst I was scanning the transcript a genuine acknowledgement of the problems with taking 75% of the ramp test.
And today Youtube will announce that based on all the data they have from Youtubers producing content Youtube will start to create digital Youtubers with sideway hats producing cycling content 24/7
Adaptive Training going to kick ass - and im very exited about it. However, I do believe they didnt tackle the ongoing questionsmark at all - eventhough they said they would. And thats a deep dive into the science on how pol and pyramidal gains are different, and yet - the adaptive training would probably still apply more intensive rides than Dylan mentioned was way more than you should?
The TrainerRoad program is depending on the plan either threshold or pyramidical. They will now also introduce polarized plans. This will give them more data points to make even better recommendations.
For what it’s worth, they briefly said that they disagree with how they were represented and that they disagree with the interpretation of the study. Don’t quote me on the exact wording though.
I hope I now don’t make this up but I also remember that they said more is to come next week. Though again, please don’t quote me on it. I was driving while listening to the episode.
Just listened to this exact section of the Podcast.
Copied from the Youtube Closed Captions 'And we’re going to go dive into all the science between into that next week.’
‘But the short summary of, of it is, um, we don’t agree with some, of the way the study was read.’
‘And also, we don’t agree with how Trina road (how the software heard Nate say TrainerRoad, I guess…?) was represented and we’re just going to talk about the science and everyone can make their own decision on that.’
Nate then goes on to discuss TRs plans for POL Training Plans. Starts around 1:33:00
They said they’re tackling the subject more in depth in the next pod. I would withhold judgement until then. But they’re obviously going to argue for this new approach they’ve been working on for over 3 years. I do like that they are implementing beta polarised plans though, I wonder if this will also utilise the same ML.
I think that when the ML is working perfectly, it would plan your training as it fits your purpose best. So the need to discuss POL or SS would vanish. The plan would build itself to your needs and capability’s. Also it would learn along the way, something some humans don’t do so well I guess. This I say because I recently started a mid volume plan again thinking I could cope with the stress, boy was I wrong again this time. That doesn’t make mid volume a bad plan as a hole, but a bad plan for me at that time.
Also a note on the science presented by Dylan. When looking at the groups they studied, also in the meta analyses, the total group presented isn’t that big and pretty selected. Generalizing these results to the general public and saying something is the right or wrong way to train is somewhat, to say it mildly, over confident in my opinion.
What works for me doesn’t have to work for anybody, and what works for a group of people doesn’t have to work for me.
Yeah, that caught my ear as well - - kind of interesting when TR is supposed to be all about the plans/plan builder making you faster and half the people don’t use them. This new path should help bump that average more in their favour though. Maybe.
I came to TR purely for the large workout library and the ability to push workouts to Wahoo. I have TrainingPeaks too but building workouts takes time, TR has a workout library that allows me to quickly choose some workouts and create a recurring plan and do those workouts indoors or out.
I think this is mostly human nature. We are presented with choices every day. If you have the choice between and easy ride with friends, outside, with the sun shining vs an indoor interval workout - I’d be willing to bet, more often than not, folks are going to choose that easy ride. There’s nothing wrong with that choice and actually from an overall health perspective it’s probably the better choice. But, that doesn’t mean it’s going to make you a more fit athlete. I think the “Train Now” feature is going to be really good for those who are occasional or sporadic users.
