I presume deleting them would re-trigger all the red/yellow detection as any track on Garmin will end up in TR.
Just my two cents on this. Take it for what it’s worth or not worth. My left diaphragm became paralyzed 10 years ago. My lung will not inflate the way it is supposed to. I was told by Lung Doctors that cycling was still ok and that it has been shown that one lung will provide enough O2 that it won’t matter. In other words the fact that I only have one good lung will not be my cycling limiter. I do not include any exercise to TrainerRoad except cycling. The system does not know that I only have one functioning lung nor does it care.
1 Like
I think the answer is just not to carry breathing exercises and dog walks over to TR. It’s just not going to have a big impact and most people don’t post these so the model doesn’t have a whole lot to go with.
3 Likes
That’s for an old model that estimates TSS based only on HR for rides that don’t have power. IE you have a mountain bike without a power meter.
It’s used to help fatigue management understand stress when there’s no power. Definitely not as good as a power meter but better than guessing or ignoring it.
If you have power on all your bike rides it won’t be used.
I know all this. Just wanted to know if it’s also relevant for the new AI model? Since you only mentioned “old model” I assume it’s no input to the new AI model.
Have you found anything yet? It really does look like this is the cause.
It is caused by the high number of Garmin activities. Your system interprets this as me having an overly stressful day, simply because it cannot properly handle such a large number of low-intensity activities.
These are 5-minute breathing exercises or simple dog walks three times a day, which I’ve been doing consistently for the last 12 years.
As a result, your system keeps reducing my training plan. Garmin correctly recognizes my training load and recovery time — unfortunately, your system does not.
The only workaround would be to disable the Garmin connection and delete all Garmin workouts, but that surely can’t be the intended solution, right?
I get what you’re saying, but realistically, I would just delete those things from your TR calendar rather than let them degrade your training plan while you wait for them to fix it. This has been an issue ever since RLGL came out a few years ago. I doubt it will be fixed quickly, if at all.
I’d also think more about reducing the number of “sets to failure” that you log. Your number is quite high and I’d bet that is having an impact. Think about how many sets of “he really went very hard on that one” you want the system to see.
4 Likes
I think, I understand the strength training thing wrong 
Anyway I disconnect Garmin an delete all imports when it make 0 Sense
Curious - if this is something you do daily, what value do you find in even recording these for Garmin? It just becomes part of your daily load. Much like someone who has a job where they are on their feet all day. They don’t need to log their workday if this is part of their routine and they are adapted to this load.
7 Likes
I document these activities for multiple reasons.
First, I record my dog walks, including photographs, for personal documentation.
Second, this documentation supports adherence to daily routines and serves as a motivational tool.
In addition, I am participating in a dedicated medical study. I have a history of lung cancer and am currently considered functionally recovered following surgical removal of approximately half of one lung.
As part of the study protocol, continuous 24-hour data collected via Garmin devices is being analyzed.
3 Likes
Personally I’d keep the Garmin stuff in there if you want and just turn off TR fatigue detection 
1 Like
Your reasons are absolutely fine. The issue is, however, that the data TR AI model was trained on most likely look very different, and therefore the predictions can be way off in your particular case. Most people do not log such activities. E.g., I do not log most walks under two hours (and even many longer ones) - I’m not that disciplined and do not use Strava as a social network. More importantly, I do not consider them a part of a training load. Some days I walk around the city running errands for hour or more, at a brisk pace (> 6km/h) - these activities are also not recorded. For many people, this is just a regular background activity not worth recording. If most people do this, this is what the model would consider the norm. People with many short activities are likely so much underrepresented in the dataset, that the model is unable to make any reasonable predictions. Heck, if I record a three-hour swift but easy walk (with or without a HR strap), which I usually do not, the AI model thinks I must be fatigued even though I barely notice.
If I were you, I would go for one of the solutions mentioned above:
- Do not import these activities to TR, or
- turn off fatigue detection.
5 Likes
We are working on this now.
I believe we are tweaking some things on the backend that should result in fewer red/yellow days from those less stressful activities.
I wouldn’t turn fatigue detection off or adjust your training approach, as that creates other issues. This fix we’re working on should get things sorted out. 
5 Likes
Thanks, that sounds good