If TR AI adapts and adjusts for us, why does it not learn what I do? Is it only based on what everyone else does and aligns everyone else to us?
For example, I regularly do three day blocks - like every week. So why do I get a yellow day for day three every week? If AI was learning “me” it shouldn’t knowing what I every week.
For background, I’m self coached and use TR with interest and indoor rides when I can’t ride outside.
I got red days weekly after my big ride but now they are yellow and I expect they will be green in the future.
It takes some time and consistency but eventually starts learning what you’re capable of doing.
It’s designed to adjust a plan on the fly, not be some magic all knowing Ai that will constantly analyze you and your non-compliant rides and coach you bespoke. No such thing exists … yet.
I’m finding it quite formulaic, for example, it’s prescribing me 30 minute workouts on a Wednesday, if I ride a longer Audax at the weekend it will give me a yellow day for Tuesday and change the workout to endurance. Why not just move Tuesday’s workout to Wednesday and put the 30 minute workout on Tuesday? It does seem very, very conservative with regards to volume.
When using Plan Builder, you configure ride types and time availability for each day.
Your suggestion makes perfect sense, but TR doesn’t have visibility into your flexible time slots to automatically reshuffle workouts across days. It can only recommend reducing intensity on yellow/red days.
If you know that you always do very long rides during weekends, reshuffle week structure around yourself in PB and TR will adjust your calendar automatically.
AI is a catch all term that tells you nothing of any value.
TR doesn’t learn and adapt to you as an individual in the way you are describing. It uses a pretty simplistic (but generally effective) methodology to prescribe progressive workouts based on your chosen plan whilst ensuring you don’t do hard stuff when you are fatigued, and uses progression levels to determine how hard those workouts can be. It doesn’t take long to figure the capabilities of TR out, but be careful with terminology like AI and machine learning.
That’s useful. Thanks.
I’ve been a TR member since close to inception and defo before AI. If it hasn’t learned me now…
What @Joe and @mtbjones said is correct. TR will adjust its recommendations over time, but keep in mind that it can go both ways.
It will learn to give you a longer leash if you show it that you can remain productive and sustainably continue down the path, but if you’re not consistent, it will likely remain unconvinced..
In your case, you’ve been very quickly ramping up your training stress after a period off the bike, so TR is likely seeing your current ramp rate as unsustainable. As you start to level off and build consistent training at a specific level, it will have the data it needs to trust that you’re not flying too close to the sun.
It’s not all about what you can do based on previous seasons, it’s also about how you get there. While TR likely won’t fight you once you get back to the volume you’ve done before, ramping up too quickly, especially after some time without training, will always result in some suggestions to ensure that you’re on the right track.
Let me know if this helps, and if you have any questions about any of this. 
Thank you for your focused clarification. I do take time off the bike in the winter and come into the spring fresh and reinvigorated rather than ride all year and burn out, especially on the trainer - I know this is your product, but I cannot handle it.
What is the AI aspects behind FTP Detection, may I ask? I can into the year w a higher FTP than I expected and it hasn’t increased at all (but one point here and there). What are the process, behind increases and is there still a bias towards indoor training?
It looks like you didn’t update your FTP between August of last year and April of this year, so to see you at the same FTP 8 months later with a major training break is pretty good!
There isn’t a bias towards indoor training, but we do generally see the best/most consistent results from athletes who follow a training plan (inside or out).
If you’re looking to start getting some fitness gains, I’d recommend using TR to build out a custom training plan that you’ll be able to follow with consistency.
Let me know if you’d like a hand with that! 
For reference, I was a USAC coach but let my credentials expire - I really did it for myself anyway. So I’m self coached and do have a plan. It’s not entered within TR but I do take Train Now suggestions if I’m doing a ride inside and wish to add some variety.
As a past world class athlete competing at the highest levels, I understand my body and pull from experience, but I am getting old so I tend to listen to my body a lot and skip/shut down/adjust regularly.
You can turn off RL/YL in the settings.
As a very long term rider/racer now in my 60’s I don’t feel that TR has done enough to pinpoint AI to the extreme’s in more the more experienced. There’s generally not enough super top-end training sets and I notice this when I race and don’t have the ability to go to “11” when needed.
I think their AI does a good job for the majority but the more experienced are not benefiting as much at they could.
Why do you think the Red/Yellow/Green feature in TR is leveraging AI and “learning” over time? I’m not saying it doesn’t, but I haven’t heard any specific claims from TR that it does. I don’t have any insight beyond what I see as a user, but it appears to just be leveraging basic training principles like TSS, ramp rate, stress balance, etc.. I think it also incorporates how you are rating your workouts, but again I wouldn’t assume there is an AI/learning loop element. And that’s not a bad thing, using traditional training stress and ramp rate concepts are a well proven approach to manage training load over time. And just because you are doing 3 consecutive days every week doesn’t mean that the 3rd day isn’t a day to be cautious. There’s nothing the matter with training on a yellow day, it’s just a head’s up that you are entering the danger zone for ramp rate or stress balance. Just because you go into that zone every week doesn’t mean it’s not a risky area.
Maybe there is some aspect of AI or learning involved, but I would think TR would at least mention that in their explanation of the feature if that were the case. I didn’t look at the entire website, but this is the support link that explains what Red Light Green Light is and how it works. No mention of AI or learning over time. Sorry to beat this point to death, it’s just a pet peeve of mine. Just because a company uses AI somewhere in their solution (like AIFTP for TR) doesn’t mean it’s being leveraged for every aspect. And while AI is very real and powerful, it’s currently much more prevalent in marketing than it is in real use.
https://support.trainerroad.com/hc/en-us/articles/20753792083995-Red-Light-Green-Light
Maybe I misinterpret what “If you want to become a faster cyclist, TrainerRoad, it’s AI Cycling Coaching” means - from last podcast.
TSS, progressive overload etc is what I’m self coaching to, so I believe we are saying the same thing and asking the same question.
I think you are me: meaning that’s close to where I am
Any marketing verbiage that includes “AI” tends to pull the bugs to the light these days, so pretty much any software company that can spell “AI” is using it in their marketing (and it’s smart to do so). I don’t doubt for a second that TR is using AI in some valuable ways, but I don’t assume it’s used (or should be used) for every feature within the application. Again, the red light green light feature in TR may use AI, but I don’t see TR claiming that it does. I just scratch my head when I see people assuming AI is this magical all-encompassing thing that is looking at every piece of available data and touching every aspect of the application.
I guess we can debate the definition of “self-coaching”, but I think anyone using TR is doing a form of self coaching. The TR system is a great tool to propose and dynamically adjust training plans and take away a bunch of the heavy lifting. That’s super valuable to folks who don’t want to mess with all that stuff or don’t have the experience. And it can also be valuable for more experienced athletes who use bits and pieces of TR while taking a bigger role in managing their own training. TR is probably has the biggest bang for the buck of any coaching/training tool in the market, but that doesn’t mean it’s going to get you closer to your genetic potential compared to other approaches.
