Can confirm that there are pros that use TR. Some use it as their coach and have it guide their training, some use it as an FTP or fatigue tracker, some as a training calendar, some as just a workout player, etc.
It always feels really cool for all of us when we hear about these scenarios .
However, we don’t actively search our user base for this kind of thing as we are strict on privacy. I’m sure there are more than we know about, but athlete sponsorship isn’t a part of our marketing strategy anyway, so it would be inappropriate to use their name/image/likeness.
I think the only exceptions to this that have occurred are if they write into our podcast to ask a question and allow us to use their name.
That said, TR is built to be individual and in doing so can meet the full spectrum of athletes where they are at. We’re super proud of that!
Thanks @Jonathan for your comments, analysis and such great content on all the podcasts.
I think the only problem with quoting statistics such as these is it’s too easy to cherry pick numbers to portray a narrative. I suspect if you look beyond the headline statistic you’ll likely see that workouts were skipped or swapped off the back of AT or RLGL recommendations. Not sure if you’re also misrepresenting ride failures when I was busy doing other outside rides instead.
What are your thoughts on my next block? What did you think of my proposal above?
I did find it entertaining to see the latest marketing taglines.
With TrainerRoad, you don’t have to choose between fun and fitness — you get both. Whether you’re chasing PRs, group rides, or epic adventures, our AI-driven system adapts to your life so you stay fast all year.
Ride how you want — indoors or outdoors Know your FTP without tests Manage fatigue so you’re always ready Add your favorite rides into your plan Get the perfect workout suggestion after every ride
Clearly I’m not getting the outcome as described above with AT. Is it a failure of the athlete or the system. It’s probably both. Maybe I’m not using the system correctly. At this point I’m choosing to take back some of the control of my fun and fitness and not solely rely upon the AI to get it right.
I tend to build my own plans currently, comprising sweet spot, threshold and VO2max progressions; staple workouts being the full SS90 library, “classic” threshold stuff such as 2x20, 2x24 (done @97-98% usually), and longer 4-5min VO2max workouts.
It does take a little time doing this, but it’s not particularly onerous, and if “full control” is required then there’s always going to be a bit of manual work to do.
The workouts I use are marked as favourites to make it easier to pick them out, and I could do with a few more to choose from to provide greater granularity on the progressions. But otherwise, the main thing that’d help me is when AT begins scoring workouts for the work completed, as I believe is in the pipeline: I use PLs as one of things to help me select workouts, and at present I don’t always get credit for the work done as I adjust the Intensity % of some workouts to pitch them at my desired level.
EDIT: the old push/pull week features, which were lost when TR moved away from the old fixed bingo card plans, would also be very handy for me
@AldridgePrior sounds great! It’d be wonderful if we had the facility to save our own custom plans. Wouldn’t be much of a stretch to then make them shareable.
The approach of analyzing every workout completed over multiple years seems like the exact opposite opposite of cherry picking? It would seem that overlooking the fact that you only followed 58.7% of the Sweet Spot, Threshold, and VO2 Max prescriptions would be more akin to cherry picking.
The problem I’m highlighting here isn’t that you made changes to the plan or didn’t follow the plan, it’s that you did those things but then have portrayed that you followed the plan and ended up plateaued, when in reality you only followed 58.7% of the most crucial parts of the plan.
To use an analogy, if you followed only 58.7% of the directions Google Maps was giving you on a drive from New York to LA, but you ended up in Seattle, it would be hard to blame the directions.
In this case, you were still impressively close to all-time fitness while only following 58.7% of the Sweet Spot / Threshold / and VO2 Max prescriptions (no doubt due to your years of hard work and TR’s AI working hard to adjust for you), so maybe a more apt analogy would be driving from New York to LA but ending up in San Diego instead.
It’s of course fine to adjust! We have tools in place for people to make adjustments to their training when they feel the need. But it’s my principle concern to make sure TrainerRoad is accurately portrayed and understood, and in this case, saying that the training plan is at fault for the plateau when only 58.7% of the key prescriptions were followed isn’t an accurate portrayal of TrainerRoad.
In regards to this approach, I think it may work for a period of time, but perhaps it’s short-lived? The SSB HV plans were originally designed with periodization in mind, implying their usage almost like a block periodization plan that would not be continuously or repeatedly used for significant periods of time.
Further, they were made for a small subset of athletes that required this significant amount of stimulus to drive adaptations.
Novelty in training stimulus is an important principle and one that governs adaptations. So it’s good to see you recognizing the need for variation and adding the VO2 Max and Threshold work, but I’d be concerned about doing this for too long and the relative lack of time at low-mid aerobic intensities.
In short, it’s not surprising to see a boost in FTP since switching to adopt this, but if you follow that approach for too long you may lack some dynamism as an athlete, and could increase the risk of plateau and burnout. Fatigue Detection will do it’s best though
I think the argument he’s making is that he wants it to adjust the directions when he chooses to detour to hit up a cool tourist stop that’s like 75 miles off route
Not saying you’re right and he’s wrong, nor that misrepresenting the training adherence isn’t relevant
I just think the expectation this user has is that AT will adjust based on these outside rides and he will improve based on these adjustments, regardless of adherence
Clearly there’s a middle ground where some level of deviation is acceptable and some level of adherence is required - that gap seems to be where expectations vs reality lead to the creation of this type of thread
It was constantly trying to adjust, but there’s only so much it can do if you keep choosing to turn left when it suggests turning right.
Outside rides were actually not the main issue here. It was mostly due to skipping, not completing, or swapping out workouts.
That’s the case with all coaching, regardless of source or platform, but that’s not directly relevant to the issue I’m pointing out of the training being blamed for a plateau but only 58.7% of the training prescription being followed.
Again, like I said in the last response, it’s okay to make adjustments! It’s just not okay to ignore those in retrospect and misrepresent reality.
Good to hear, i know several that use it as a workout player because it’s an excellent one but not aware of anyone pro who follows the ai training plans. I don’t expect you to share anyone who does for obvious privacy reasons that you stated but are there any podcast episodes you can point me to where these pros are public about following TR training plans?
Yeah, understood. I was merely trying to identify where the expectation gap was between this user’s experience and what is being delivered
ETA: he likely feels the amount of detours/wrong turns is within the acceptable range, you feel differently. Maybe some ballpark minimal amount of compliance language would help get ahead of these threads
@Jonathan - One thing I’d be curious about is where you’re pulling these metrics from. I’m guessing there’s an internal dashboard, would it be possible to share some version of it for users, along with a sense of compliance to plan/training load, etc.? After the new Zwift integration, I’m considering going back to TR for the off-season, but since everything hinges so much on plan compliance, I think it’d be a huge benefit to users to get a more detailed view into how they’re fitting into the machine’s expectations.
Something like a breakdown of (off the top of my head):
Expected Volume vs Actual Volume
Expected Time Across Zones vs. Actual Time in Zones
Compliance Across Energy System (helpful to know if you frequently skip or have trouble with, e.g. VO2 Max or Threshold)
Overall Compliance Rating (a percentage and some kind of scale like Poor/Fair/Good/Excellent)
I haven’t been subscribed to TR in a while, so some of this may be in there, but I think a dashboard that showed these values across a range of date options (week/month/60 days/90 days) would help build some guardrails around the process and provide some tools that could be useful for knowing when you should or shouldn’t go freestyle. I know the Red Light Green Light provides this to some degree, but those calculations are a black box and it seems like there’s some lag time to catch up in various situations.
If the AI model is as good as we’re claiming it to be then perhaps we need to see a better feedback loop that we are still on the right path. For example, stay within these boundaries\lane then at the end of this block you will have a 1% increase in FTP?
I don’t see anyone making claims like that and that isn’t a bar you need to get over to have an effective training system. AI is the real deal, but it’s from perfect and makes lots of mistakes just like people. Everyone wants to debate “the perfect training approach”, which doesn’t exist. Training, physiology, and performance are constantly moving targets with endless variables and outcomes in play. TR (and other) training systems are barely scratching the surface of what is possible with AI. That doesn’t make them inadequate, it just means it’s a fairly new approach with endless potential. Tesla can spend hundreds of millions figuring out how to leverage AI for self-driving, nobody is making those kind of investments in AI-based athletic training systems. AI is certainly going to play an increasing role in training as we try to wrap our heads around all the moving parts, but it’s still in it’s infancy. I don’t see TR making any specific claims about how the AI is deployed or how well it works, they are just talking in broad strokes. And remember that whatever is being said to the public is marketing. Everyone in the software business is marketing around their AI capabilities, it would be dumb not to. If users make poor assumptions that AI is magic and involved in every aspect of an application, companies are happy to let them think that. And no software company is going to give detailed information about exactly what aspects of their app uses AI and what the capabilities are. Again, I think you are throwing out a silly measuring bar. I don’t hear TR claiming their system is perfect, but that doesn’t mean it’s not a useful application at a reasonable price point for a large percentage of athletes.
I mean it is pretty common sense that if you are doing a plan that has a lot more zone 2 then it’s going to be crucial that you do the threshold/vo2 workouts in the plan as these are going to be the most crucial.
I know my feedback loop is when I realise my FTP is stagnating/declining or not going up the way that I want it to. If that’s the case I generaly think it’s good indication that I need to knuckle down and make sure I don’t skip the hard workouts as it’s generally when i’m not following the plan.
I do agree that it would be helpful to have almost a chatgpt assistant that can explain the plan and provide some coaching around plan compliance/tips etc. so that you can understand a bit more about what is happening in the background and optimise how to use the plan.
And that people should not really fall for the hype generated around the words AI and adaptive training.
As I said, the basic structure of TR’s plans does not change whatsoever regardless of your “training levels”.
So as things stand, you’ll always be repeating the same training cycles which might or might not cause to plateau. All the “ai” is doing right now is change a 3x12’ at threshold into a 3x20’ without you having to bother.
CoachCat the app from FastCat has a great AI assistant that gives you feedback on workouts, sleep etc, and guidance for your next workout. Its outstanding. Would be great if TR were able to provide something along these lines.
Yes I agree with this, and this is the point I am making in another thread here about menopausal training.
Maybe the ‘bit of everything approach’ with some sweetspot, threshold and VO2 scattered about isn’t actually the best for everyone. Maybe some people wil work/improve better on dedicated blocks of a certain zone or with a change in the zone distribution. I do not believe, at the moment, the system truly analyses that or where you are in your training.
I also hate the fact that it is always assumed that only those at the pointy end who have years and years of structured training under their belt are allowed to think outside the box with their training and zone set up. It drives me crazy actually, that is it assumed that, if you haven’t trained with power or with structure then you are nowhere near your genetic potential and you have to ‘continue your schooling’ the standard route. It is possible that people can train and work hard without structure, and get close to their metabolic potential doing so. It is also possible that people who never train with structure could benefit from a different stimulus than the standard ‘kitchen sink’ type of base-build-specialty.
I wanted to circle back on this comment as I think it’s inaccurate to suggest I’d get the same benefit from the current plan builder even if followed to the letter.
See below a comparison of what I’ve done for the last 17 weeks where my TSS has been in the range of 580 - 620. On the RHS of the calendar/plan we can see the proposed equivalent plan builder recommendation resulting in a TSS of 440 - 500. Do note there’s even a whole week of skipped workouts (no doubt included in the most recent stats quoted above).
In terms of receiving a good feedback loop and tracking my efforts it’s been great to have intervals.icu tracking RLGL equivalent and time in the optimal training zone as below. Would be amazing to see these metrics within TrainerRoad.
.
Ride how you want — indoors or outdoors 
Know your FTP without tests 
Manage fatigue so you’re always ready 
Add your favorite rides into your plan 
Get the perfect workout suggestion after every ride
