Alternative to Trainer Road

A light bike doesn't replace good fitness.

Moderator: Moderator Team

Post Reply
Andrew69
Posts: 593
Joined: Thu Feb 19, 2009 10:52 am
Location: ɹǝpunuʍop

by Andrew69

jlok wrote:
Fri Mar 12, 2021 4:09 am
cheapvega wrote:
Thu Mar 11, 2021 6:02 pm
I guess I will pose my skepticism over TR's "AI" addition as a series of questions:

- Are FTP tests better than real world power curve data for determining training intensity?
- Is there value in individualizing interval length/intensity to a rider's specific strengths and goals?
- Can adjusting training intensity more frequently yield faster adaptations?

Basically my feeling is this is still a missed opportunity. For $20/mo I want customized workouts built around my individual strengths/weaknesses, goals and rides, as well as intensity determined from training data (so I never have to do an FTP test again). From what I'm seeing this still requires regular FTP testing and only provides suggestions to canned workouts. Maybe their workout library is extensive enough to not need to build things from scratch. Maybe they are really wed to that IP. IDK. Either way I'm skeptical.
FTP = 20mins steady maximal work * 0.95. The problem of FTP is that it has an assumption of the same capability/capacity to work above FTP (i.e. FRC or HIE, or W' capacity, similar ideas) among all riders. If two riders have the same FTP, would they necessarily sprint the same?

If TR "AI" workouts are based on FTP, it is flawed fundamentally, or the "AI" is not I enough to understand different riders and prescribe more appropriate workouts based on rider's goal (to sprint better or to climb better).
A couple of things
FTP is not and never has been, 20mins steady maximal work x 0.95
That is nothing more than a crude approximation because a true FTP test (maximal steady state "for around an hour") hurts a lot if done right so the 20 minute approximation came about to encourage people to test more often
For some people, it maybe 0.96 or even 0.97 (those with low aneraerobic power contributing to FTP)
For others (like me), its more like 0.93 or 0.94
0.95 was settled on because for most people, its "close enough"
If you are already highly trained, then the outcome will be less than optimal

There is nothing wrong with basing workouts on FTP.....provided they are workouts aimed at raising FTP :mrgreen:
But yes, I agree that using FTP as a base for sprint training is flawed

A model, any model, is only as good as the data that it is based on.
If you feed the model garbage data (ie sub maximal efforts), then expect garbage to be spewed out by that model

So it is of critical importance regardless of the model used that you feed it good data, and that means testing.
Must be maximal (or very close to it) and it must be constantly tested and retested otherwise the model quickly loses its validity

This is the issue with any program that is AI or ML based.
Becuase it isnt being overseen by an experienced coach, you probably wont know that the workouts it has given you are sub-optimal (or worse, cause you to lose fitness)
If you do know that they wont work, then you probably have enough experience and knowledge to self coach anyway

There have been a lot of very intelligent people trying to crack this nut and I dont think its been done yet
If TR has come up with a half decent algorithym and can modify it based on the feedback data they receive, then they may be the first to do it

As it stands, it doesnt appear that they have done anything more than developed a model giving cookie cutter workouts based on set parameters of the "average" athlete
Obviously, no one knows yet and Im willing to give them the benefit of the doubt....but they wont be receiving any of my hard earned cash until they can prove they can do what they actually claim

Lozaen
Posts: 124
Joined: Fri Jan 08, 2021 9:44 pm
Location: Switzerland

by Lozaen

Andrew69 wrote:
Thu Mar 11, 2021 10:55 pm
Lozaen wrote:
Thu Mar 11, 2021 9:14 pm
Well, than maybe you should have a look at their podcast, where they explain what they are doing.
It seems, that they actually measure/determine your capability in 7x3 dimensions....should that be enough?

Will look into Xert as well :thumbup:
The issue is that some believe that what they say they are capable of in the podcast and what they are truly capable of delivering may be substantially different.
Have a read of this thread on slowtwitch and decide for yourself
https://forum.slowtwitch.com/forum/Slow ... ead#unread
Read the entrire thread...what should I say.
some people complain about stuff they either don't understand or have not experienced. As far as I have seen, there was not a single comment from someone who was or is part of the beta test or knows anything beyond the podcast. Seems that a couple of people actually didn't listen/watch the podcast at all and are just "preaching" what they believe anyway. And then there is a bunch of people who complain about TR not sharing their insights with the "scientific community"
Microsoft doesn't share their sourcecode with the scientific community either, but Windows and Office still work....at least that's what i have been told.... Using a Mac :mrgreen:

Again....not advertising TR in particular or any other product (XERT, AiEndurance, Enduco,...)
I truly believe, that ML can really boost the understanding of what works, and what doesn't...both...in the broad and also for the individual.
What it will definitely not be able to manage: Transform a "Joe the Plumber" into Lance Armstrong, Peter Sagan or MvdP. If this is ones expectation, than he or she will most definitely be disappointed.
If you want to receive a flexible training plan, or even better...no "plan" at all, that will lead to better than average results, I truly believe, that AI / ML / Data Science is a promising approach.

Time will tell which approach is the best, but I am really excited about the latest developments :thumbup:
Interested in buying Carbonsport Lightweight wheels with broken spokes.

BMC Teammachine SLR01 Disc Team 2018|Ridley Helium | Kuota Kross|Cannondale Scalpel 29 Hi-Mod

by Weenie


Visit starbike.com Online Retailer for HighEnd cycling components
Great Prices ✓    Broad Selection ✓    Worldwide Delivery ✓

www.starbike.com



cheapvega
Posts: 380
Joined: Sun Oct 06, 2019 1:12 pm

by cheapvega

I don't think it's about unrealistic expectations. For me there are 2 main issues. One their AI doesn't seem to lean on much more than FTP... very basic but relevant data points like FRC & power curves seem secondary at best, despite them being of great importance to a wide range of riding disciplines.

And two they seemed to choose to dumb down cycling parameters instead of using this as an opportunity to educate. For example they use "progression levels" for different... intensities?

Image

What do each of these levels mean? How are they determined? What does each unit indicate? To me it would be way more valuable to train to a power curve based on what my strengths/weaknesses seem to be and the kinds of rides/events I plan to do.

To be fair I am not sure anybody has put it all together in a way I'd like, though it seems Xert comes close. It's just strange that the most expensive and popular platform seems the furthest from usefulness (IMO). We will see how they improve on it but I will save my money.

Andrew69
Posts: 593
Joined: Thu Feb 19, 2009 10:52 am
Location: ɹǝpunuʍop

by Andrew69

Lozaen wrote:
Fri Mar 12, 2021 10:07 am
Read the entrire thread...what should I say.
some people complain about stuff they either don't understand or have not experienced. As far as I have seen, there was not a single comment from someone who was or is part of the beta test or knows anything beyond the podcast. Seems that a couple of people actually didn't listen/watch the podcast at all and are just "preaching" what they believe anyway. And then there is a bunch of people who complain about TR not sharing their insights with the "scientific community"
Microsoft doesn't share their sourcecode with the scientific community either, but Windows and Office still work....at least that's what i have been told.... Using a Mac :mrgreen:
The people commenting about what TR may or may not accomplished are very experinced exercise scientists with an intimate knowledge of the subject.
They are literally the experts in the field
Lozaen wrote:
Fri Mar 12, 2021 10:07 am
Again....not advertising TR in particular or any other product (XERT, AiEndurance, Enduco,...)
I truly believe, that ML can really boost the understanding of what works, and what doesn't...both...in the broad and also for the individual.
What it will definitely not be able to manage: Transform a "Joe the Plumber" into Lance Armstrong, Peter Sagan or MvdP. If this is ones expectation, than he or she will most definitely be disappointed.
If you want to receive a flexible training plan, or even better...no "plan" at all, that will lead to better than average results, I truly believe, that AI / ML / Data Science is a promising approach.
There is absolutely no doubt that AI/ML will be a game changer in this area.
But it does not appear that anyone is as close as they claim to be
For example, I signed up for an AiEndurance free trial period.
Linked all my data from my Strava account and got an email to say my plan was ready.
Logged in only to find.....nothing.
Emailed Marcus (owner, who was extremely helpful) who took a look to see what was going on and then came back and admitted that I was too far from the "average" at the moment for the algorithm to come up with a plan.
You see, I have had a looong time off the bike due to well...life (family,work, etc) and so while my FTP was crap, my sprint power is still pretty good.
Too far outside what they would expect to see, so I could not be helped right now.
They do have a large update planned very soon (may have already happened) and Marcus thought that would solve the issue.
But because he came straight out and said they couldnt help me at this stage rather than trying to sign me up, and they can see the limitations of their current algorithm and they are working on improving it, I will be happy to give them a go at a later date.
Lozaen wrote:
Fri Mar 12, 2021 10:07 am
Time will tell which approach is the best, but I am really excited about the latest developments :thumbup:
Exactly, however, as usual, it us poor "consumers" that are going to have to be the beta testers
I dont mind being a beta tester, but not if Im paying top dollar for something that may or may not work, and at the moment TR is the most expensive with the least "proof of concept"

Lozaen
Posts: 124
Joined: Fri Jan 08, 2021 9:44 pm
Location: Switzerland

by Lozaen

Andrew69 wrote:
Fri Mar 12, 2021 11:04 pm
The people commenting about what TR may or may not accomplished are very experinced exercise scientists with an intimate knowledge of the subject.
They are literally the experts in the field
Don't get me wrong.....I am not questioning their expertise in the Field of training science....it might simply be worthless.
What I do question is their motivation towards TR and their knowledge about Computational engineering and AI in general.
In other words....the guy that trained the google AI for the first time to search for cats on pictures in the WWW was no expert in "Cat-Science"....yet the algorithm works quite good and gets even better every single day.
Too much "cat-Science" might actually be harmful.
Interested in buying Carbonsport Lightweight wheels with broken spokes.

BMC Teammachine SLR01 Disc Team 2018|Ridley Helium | Kuota Kross|Cannondale Scalpel 29 Hi-Mod

Lina
Posts: 1055
Joined: Sat Sep 01, 2018 9:09 pm

by Lina

Lozaen wrote:
Sat Mar 13, 2021 7:54 pm
Andrew69 wrote:
Fri Mar 12, 2021 11:04 pm
The people commenting about what TR may or may not accomplished are very experinced exercise scientists with an intimate knowledge of the subject.
They are literally the experts in the field
Don't get me wrong.....I am not questioning their expertise in the Field of training science....it might simply be worthless.
What I do question is their motivation towards TR and their knowledge about Computational engineering and AI in general.
In other words....the guy that trained the google AI for the first time to search for cats on pictures in the WWW was no expert in "Cat-Science"....yet the algorithm works quite good and gets even better every single day.
Too much "cat-Science" might actually be harmful.
And at the same time the people at Google tweak their algorithms to bring up "right" search results. Not always based on what the user is searching but what Google wants to show them. It's very easy to fall to that same trap for a service like TR, and it's not exclusively just TR but every single coaching business out there. What gets the best results and what brings in clients and pays your bills are often very different. And don't think for a second TR, or any other online coaching business, will be optimizing their platform for training results but best monetary results for them.

And AI and ML often gets thrown around like it's some magical solution to everything. It's not, and 9 times out of 10 it's just buzzwords for marketing to create hype around something.

cheapvega
Posts: 380
Joined: Sun Oct 06, 2019 1:12 pm

by cheapvega

People get on TR to get faster though, no? What could they possibly be tuning the algorithm for over making people faster?

Andrew69
Posts: 593
Joined: Thu Feb 19, 2009 10:52 am
Location: ɹǝpunuʍop

by Andrew69

cheapvega wrote:
Sun Mar 14, 2021 1:01 am
People get on TR to get faster though, no? What could they possibly be tuning the algorithm for over making people faster?
I get what Lina is saying.
Its not about making people faster, its about making the "right" people faster

There is no money to be made by then by making already very highly trained atheletes faster (the 1%ers, the best of the best) as the market is so small
Better to design the system to make the other 99% faster as it will make them more money

cheapvega
Posts: 380
Joined: Sun Oct 06, 2019 1:12 pm

by cheapvega

This is the sentence I'm focused on:

"And don't think for a second TR, or any other online coaching business, will be optimizing their platform for training results but best monetary results for them."

People jumping on platforms like TR are doing so to see training results, regardless of their level. So I'm at a loss as to what they could focus on instead that would yield more monetary results. Plus if the platform works for 99% of users- which frankly is all of us amateurs- they did their job no? The top 1% are going to need more than a $20/month app to basically improve on world class performance and elite genetics.

I think TR rushed this out to respond to stuff like Xert etc. To me it seems like they didn't really think this through or make the most of it. But it is what it is.

Lina
Posts: 1055
Joined: Sat Sep 01, 2018 9:09 pm

by Lina

cheapvega wrote:
Sun Mar 14, 2021 3:48 am
This is the sentence I'm focused on:

"And don't think for a second TR, or any other online coaching business, will be optimizing their platform for training results but best monetary results for them."

People jumping on platforms like TR are doing so to see training results, regardless of their level. So I'm at a loss as to what they could focus on instead that would yield more monetary results. Plus if the platform works for 99% of users- which frankly is all of us amateurs- they did their job no? The top 1% are going to need more than a $20/month app to basically improve on world class performance and elite genetics.

I think TR rushed this out to respond to stuff like Xert etc. To me it seems like they didn't really think this through or make the most of it. But it is what it is.
Their target market is mostly untrained amateurs. Literally any structured training will get results to that group. So it becomes a quest in how to maximize retention and getting new customers. Short workouts at higher intensities is something a lot of untrained amateurs like. They want to feel like they've worked out and want to spend as little time as possible. The same group of people also want their training programs to be complex, why else would they be paying good money for the program. Their programs are also aimed at mostly increasing ramp test power. If you're mostly an untrained amateur and want higher ramp test - 20 minute power the platform definitely works. Does it work as efficiently as some other training program, most likely not. Will it have better retention than a polarized plan that is mostly Z2 with one harder ride per week, most likely yes.

The problem with TR is that like third of their plans are pretty much unusable, the high volume plans are going to overtrain most people because they have so much intensity. They even admit it themselves all the time on their podcast and forums. The best way to use TR would be to take some low volume plan and add base rides to it or take a traditional base one and add a higher intensity work to that. Otherwise you're most likely overworking yourself and aren't properly rested for the workouts. The higher volume plans just add more intensity.

You can see the same thing with personal trainers, if you go to one for a program you'll get a super complex program that contains a lot of different exercises on different machines. When in reality they'd be much better served by doing 3 - 4 compound lifts 2 - 3 times a week with increasing load for the next year. And all the PTs know that. But they all also know that they'll make a ton more money by selling complex programs every 4 - 8 weeks to the customer.

This is an inherent problem in all coaching. The top 1% are most of the time on simpler training programs than most amateurs. The same things that improve elite genetics and world class performance also work on the general majority. And until you're close to your genetic limit you can do even simpler training programs than the top 1%. But it won't sell as well as complex programs.

Andrew69
Posts: 593
Joined: Thu Feb 19, 2009 10:52 am
Location: ɹǝpunuʍop

by Andrew69

Great post Lina, totally agree

User avatar
LiKuid
Posts: 73
Joined: Thu Oct 24, 2013 1:09 pm

by LiKuid

Andrew69 wrote:
Mon Mar 15, 2021 11:54 am
Great post Lina, totally agree
I am completetly on the opposite side here...
I ve been using TR on and off (when i am not running) during the last maybe 7 years.
As a runner, the thing that always made me feel secure with TR is how much they value their base plans. If anyone studies their plans, they will understand how highly they value base training. Their build and specialty plans are only suggested for experienced athletes, or athletes that have completed a full base cycle (12 wees or more). I am talking about their recomendations... of course anybody can jump straight on their build sweet spot plans if he wishes. That said, their base plans, are not far from a polarised training regime with mostly endurance riding in place.
TR is famous for how simple and boring their workouts are and that is probably why their loyal customer base is commited hobbyists. You need to be committed to stick with their app environment. The only way to overtrain using TR is doing full cycles of build and specialty plans continuously without any breaks, but most people understand how wrong that is. The only place i felt that indoor overtraining is plausible, is ZWIFT where you can go racing every day, any time you like. In ZWIFT, even group rides get competitive at times and that is a rabbit hole one can easily find himself in. For me, that was the main reason i switched back to TR, although ZWIFT is much more fun and i can do 3 hour base rides without being bored to death.
All in all, in my understanding, TR is a fair offering. They offer structured plans to help you get faster (whatever that means) for $15-20/month. They cannot be compared to $400/month personal trainers of course, as that would not be fair. They also have set up a great and free ecosystem with the forum and the podcasts that showcases how committed they are to their scope, unlike zwift where you definitely feel like you are Mr. nobody paying a subscription.
Regarding their Adaptive training program, given that they introduced it without asking for more money, i understand they see it as an excerice. They do not claim to have all the answers, but they seem commited to try and make the most out of the data they are generating from their customer base. If they succeed, it's definelty going to be a game changer. And in mind, success is a one way road, the question is how quickly they will be able to make it happen.
I don't know where the fitness ceiling lies with TR, but as a ±4w/kg hobbyist i do not feel i am very close to that. If i wanted to get to 5w/kg fast i wouldn't be suprised if i needed more that TR (or $15/mo) to take me there, as things get much more personal when you try to fine tune the details.
Of course i am not in any way affiliated to TR, just a happy customer, maybe part of that 95% of people that TR succeeds in making faster every time.

Lina
Posts: 1055
Joined: Sat Sep 01, 2018 9:09 pm

by Lina

TR absolutely does not value traditional base training highly. The traditional base plan is buried all the way as the last one in base plans and even on them everything but the first block has tons of tempo/sweetspot instead of traditional base.

Most people will overtrain on any of their high volume plans. Even the base plans outside of traditional base has boatloads of sweetspot. So much that there are only a few people that can actually finish them without being extremely fatigued on the last few weeks. If you read the forums or listen to the podcast even they know that it's too much volume and admit that even they can't do them.

I also feel like you didn't truly understand my comparison to PTs. I never compared PTs and TR in terms of training quality. I compared them because they both fall into the same trap. They both know their training programs aren't as efficient as they could be but if they gave their customers better training programs they'd lose money because their customer retention would be worse. Spending few months subscriptions on a few books and a few weekends to read and understand them will make you faster and/or stronger than continuously giving money to TR/PTs. And most of the time those training plans will be simpler with less intensity than what TR or the PTs have you doing.

cheapvega
Posts: 380
Joined: Sun Oct 06, 2019 1:12 pm

by cheapvega

For me personally TR was a great introduction to structured training. Books are good and worth reading but for me personally I like to learn "on the job", which TR does really well. However I canceled my subscription about 3 months after I started because by then I had enough of a grasp of how to build plans and manage training load that I felt confident doing it on my own.

The problem I found immediately was that there's not really 1 great piece of software to let you do it all- import outdoor ride data, manage/analyze training load, build and run structured training etc. I have a Rube Goldberg workflow running across Training Peaks/Wahoo/GC. From the industry's point of view it makes sense- why sell 1 piece of software when you can lock users into a subscription forever? There's nothing about TR that couldn't be accomplished in a desktop app, and frankly the desktop apps I have used are more comprehensive. The main thing TR offers is the plans which we've already determined are not optimal. The whole thing kind of sucks the more I think about it.

by Weenie


Visit starbike.com Online Retailer for HighEnd cycling components
Great Prices ✓    Broad Selection ✓    Worldwide Delivery ✓

www.starbike.com



User avatar
LiKuid
Posts: 73
Joined: Thu Oct 24, 2013 1:09 pm

by LiKuid

Lina wrote:
Mon Mar 15, 2021 2:45 pm
TR absolutely does not value traditional base training highly. The traditional base plan is buried all the way as the last one in base plans and even on them everything but the first block has tons of tempo/sweetspot instead of traditional base.

Most people will overtrain on any of their high volume plans. Even the base plans outside of traditional base has boatloads of sweetspot. So much that there are only a few people that can actually finish them without being extremely fatigued on the last few weeks. If you read the forums or listen to the podcast even they know that it's too much volume and admit that even they can't do them.
If you take your time to read TR before starting a plan, most probably you ll end up selecting a base plan. If you try the plan builder tool, and dont give it a race date anytime soon, you ll definitely be given a 10-12 week base plan before doing any hard interval work. I understand that a lot of people may start hot expecting qiuck results, but each time TR directs you to start with base work.
Lina wrote:
Mon Mar 15, 2021 2:45 pm
I also feel like you didn't truly understand my comparison to PTs. I never compared PTs and TR in terms of training quality. I compared them because they both fall into the same trap. They both know their training programs aren't as efficient as they could be but if they gave their customers better training programs they'd lose money because their customer retention would be worse. Spending few months subscriptions on a few books and a few weekends to read and understand them will make you faster and/or stronger than continuously giving money to TR/PTs. And most of the time those training plans will be simpler with less intensity than what TR or the PTs have you doing.
My point on the PT comparison was to argue that from the monthly cost of TR, one cannot expect the same as from working with a PT. I mostly agree with your point on PT's training plans.
cheapvega wrote:
Mon Mar 15, 2021 4:41 pm

...The problem I found immediately was that there's not really 1 great piece of software to let you do it all- import outdoor ride data, manage/analyze training load, build and run structured training etc. I have a Rube Goldberg workflow running across Training Peaks/Wahoo/GC. From the industry's point of view it makes sense- why sell 1 piece of software when you can lock users into a subscription forever? There's nothing about TR that couldn't be accomplished in a desktop app, and frankly the desktop apps I have used are more comprehensive...
Like you, based on my experience, i am feeling pretty confident i could built my own plan with decent results, but there is no such platform to handle this end2end. However, it totally doesn't make sense to sell it as an app from a financial standpoint. I understand it's frustratig for users, but let's be realistic, a business cannot survive by selling a fixed amount of app licence once. And there wouldn't be any incentive for extra developments.

Post Reply