I tried searching for the answer, but couldn’t find it, so maybe someone here happens to know?
How can I get out of TR AI and back to the regular TR? I’ve tried e-mailing support, but there seems to be at least a week lag to get a response, if one is provided at all. Right now my TR is locked in an infinite loop where it won’t find workouts. I had to manually pick one in order to ride today, but I don’t want to have to keep doing that.
You can disable some of the features if you want but I am unaware of any way to completely disable the update nor does it sound like TR plans on implementing a way to do so.
It’s sad. I enjoyed the previous tool. Now I find it over arching and mostly incorrect. I run three user accounts with separate results.
One user the tool said there was insufficient data for AI FTP detection. Then it recommended an FTP Test. They conducted the FTP Test and measured at 166 without pushing deep into the red. The AI FTP immediately calculated a 136 and set all workouts based upon this. I had to ignore this and manually enter FTP.
Yesterday the athlete completed Llala (sweetspot at 139w) in week 2 of their calendar. Their HR was average 135 in intervals. Their real LTHR is mid 150s. I dont understand what is wrong with their ML and why they released this. It needs to go back in beta as it is highly inconsistent.
My account is on the other end of the spectrum and keeps picking workouts at the hard end of the spectrum that I am barely able to complete. Every single workout is a stretch. It was never like this before, and this is not how I progress. It’s just burning me out and making me hate the TR platform.
I’m very sorry you haven’t heard back from support! It looks like your ticket was submitted yesterday; I’m not sure where you heard they’re a week behind but we try very hard to make sure that’s not the case. An agent is reaching out to you as we speak.
Can you point me to the workouts that were too hard? I want to make sure everything is working as intended with your account. Most of the workouts I see on your calendar look like your RPE response was right in line with what the AI predicted; I only see a single workout recently marked as Very Hard (the AI expected hard) and your next Threshold over under workout this week is lower intensity as a result. I do see a couple of workouts that were cut short and marked as due to “time”. Let me know and I’ll dig in deeper!
Can you DM me this athlete’s info, please? I want to take a look.
If you can talk off-line that would be great. Chat is never available when I log on.. There is a difference between a workout being hard and it and burning you out. The fatigue detection has been so bad, your chat has told me the only solution in the past was to change approach. I turned down the approach to see if it adjusted workouts and it did not. The pattern of this program is just not what it should be and doesnt even follow historical data in my history of what has worked over and over for me to come back from major FTP deficits and losses I’ve had form sickness and Injury. I would be better off self trained. I really Hate this thing now. Its very unfortunate.
I believe the other user is rickbrownatx
I’ll shoot you a DM tomorrow. As an aside, I see your Approach is currently set at Aggressive, and was previously set to Demanding in August of 2025. Those are the two settings where Fatigue detection is least likely to adjust your workouts downwards. At any rate, your experience does not reflect what we want, nor what we hear from the vast majority of athletes, so I want to make sure we get to the bottom of it.
If you have it set to aggressive and rate almost every workout as only “hard”, why would the ai back off of the intensity? There are two levels of hard above hard.
Joe
Hi Sean: I submitted a ticket on Feb 3 and just heard a response today (Feb 10), so that’s about a week. I also submitted one yesterday for a different issue. As far as this message, I did hear from support, but the proposed solution didn’t work. I hold out hope that a resolution can be found, though.
are you responding to me?
Im certainly not going to ever rate a sweet spot workout as “all out”. Its sweetspot. But there is a point where you shouldnt be doing a 2x20 with the second interval being at 94% and your HR at LTHR in your second week of base. Thats later season pain cave once you are used to that type of work. So perhaps the problem is the scaling their software uses for its strategy (algorithm/ML.”AI”) approach. The quiz never asks “was that workout too hard?” and they probably dont want to ask it that way because they dont trust that people wouldnt dumb down the plan and stop progress. Well I know when its not right. This isnt my first rodeo. I liked the previous tool far better.
The old system was much more progressive in it ramp up and worked. This one is not working for me.
Barely complete. Stretch. Burning me out. I believe you when you say these things.
So let’s say you have a coach and you’re feeling this way. But you report to them that the sweet spot is moderate or hard, not very hard or all out. What are they gonna do? They’re gonna say….”OK, this person has got it under control, we’ll keep cracking on, especially since we’re on an aggressive plan” Which is what the AI is doing. Hitting threshold heart rate on a sweet spot workout seems very hard. Maybe all out.
I’m not saying the new system is better for you but I am saying if it’s that hard (and it definitely sounds like it is)…..maybe just say it was very hard or all out?
Joe
I’m with you @Joe . A rating of “hard” means that he could have done another 20min 94% no problem. So it’s ok for the AI to assign harder workouts.
I’d keep working with support if you have an open ticket.
Also, there’s no need to open a second ticket for a separate question. If you’ve got an open ticket that is getting attention from someone, ask the questions you have there! 
It looks like that second ticket is already active, so I’ll avoid merging the two, but again, each athlete only needs one open ticket. 
If you know what’s right, communicate that to the software. I think that’s what we’re all getting at here.
Sweet spot isn’t designed to be easy or moderate, so “hard” as a survey response is a reasonable answer if the workout was right for your abilities.
If your workouts are feeling harder than that, then rate them as very hard, or however they feel.
You’ve got the first step down of being confident of how the workout felt, now you just need to tell the software.
At the time I opened the second ticket, I hadn’t received a response from the first one because of the large backlog. But good to know.
I see similar threads to this floating around the forum. I think it ultimately comes down to individuals not putting in the correct responses to how difficult the workouts are. They feel that a temp workout or a sweet spot workout shouldn’t be “very hard” or “all out”, so they rate it “hard”….which it should be. But then they come to the form to say workouts like this are too hard….so why not say it’s “very hard” when rating the workout and allow AI to adjust it?
This, don’t rate a workout how you think it should have felt, rate it how it actually felt.
No Joe,
And I answer this as a coach…
If someone says workout was “very hard”, but I see that something is happening in the interval that would indicate that it wasnt completed 100% satisfactory, I’m not going to step them up for the next workout. This is the most basic logic for practical programming. Fail once, repeat, fail twice step back and reset to push through a plateau.
Don’t claim that you have a model that is looking at the workouts, if it is not and is just looking at a Subjective quiz. “oh it will all workout once it figures out how you answer the quiz” BS. they have my quiz answers from Last year.
I also do my TR through Zwift which has a 10 point scale not a 5 point scale. the TR scale is pretty flawed and seems like a current WIP.
Lets not get into the psychological game they have created that if you give various reasons for failing one workout it takes a machete to your FTP prediction(ive tested the responses to see their outcomes. Accurate or not, you have created circular logic in the design of this tool. I am clearly not a fan and will be turning mine off and go back to just programming my workouts and using this as a calendar to do so with a library of workouts to quickly choose.
What makes you think that 10 points is any more precise than a 5 point scale? In patient experience surveys 5 point scales are better than 10 for precision (inter-respondent repeatability). More isn’t always better.
A 7 point scale would actually be ideal for the outcome the want form their post workout survey
What I am saying is that when I do my workouts, it is collected and was always collected on Zwift through a 10-point scale. I did not know for some time that this was not TRs survey. Apparently, that is mapped into zwift via their integration.
There are many statistical reasons in surveying not to use a 5 point scale, this would be one of them.
So TR has two scales? a 10 point scale that rates the effort of a ride or race, and a 5 point for a survey? Makes sense, glad we are debating this.