Adaptive Training Closed Beta Update

WHere do you find the FTP projections with AT?

There is no necessity to do workouts along the entire curve, you can still pick you plan according to your needs. In my case the static plans worked fine in certain zones but were too hard in others. The progression levels made that problem go away. I am not sure if a strong improvement in one zone affects the PLs for other zones in the TR model, but don’t think that is a major issue anyway. If you don’t want to use workouts in those other zones it is irrelevant, and if you do the PL is corrected pretty quickly.

Why even use default workout screen? Just create the second screen as suggested by tr and go with that. Works great for me.


FTP projection is not available yet, but the feature is being tested on TR internal builds (Nate posted an instagram story with it the other day).

1 Like

Not sure why I’ve not seen an increase in my Endurance PL with this workout - It was meant to be 90mins but with it being fine weather here in the UK I extended it by almost double. The hill at the end was a choice of one of three to get home but admittedly the hardest one. The end of workout questionnaire was the “failure” one, I chose the “I didn’t fail” option and set the success questionnaire as “moderate” - apart from the fourteen minutes of the final climb it was very much easy going.

In my calendar the workout is characterised as “stretch” but there’s no change to my PL. Seems a bit hit and miss as to which workouts contribute, on Tuesday I did which is characterised as “Breakthrough” and four of the progression levels moved up.

Don’t see how machine learning added value here. Progression levels are defined as the “relative difficulty of workouts within each energy system”. Sounds similar to the IF factor. So took 10 sweet spot workouts at different progression levels and plotted them against their IF factor. It shows a linear correlation and for this cross section 97% (R squared) of the variance in progression levels is predicted by the IF factor. The remaining 3% is already in the range of power meter accuracy and precision so not noticeable most by users.

Creating the plot took 10 minutes. If you would dive a little deeper, eliminated the warm-ups and looked at interval lengths and %FTP in the intervals I am pretty sure you will find the exact analytical formula to calculate PLs.


duration is also a factor, not just IF

Because its a sub-par user experience. TrainerRoad talks about Cognitive Load and Deep Work and Flow, and generally sets a high bar for usability. But TR-on-Garmin breaks the mold, in my humble opinion. I’ve tried a lot of different data screens and having been doing outside workouts in 2016 (before joining TR in December 2017).

And the sub-par user experience continues after the outside workout is completed. In TrainingPeaks I can see the target power, in TrainerRoad I cannot.

1 Like

I’m sort of mumbling general agreement, but are you sure? I do have a target power on my Garmin Edge for TR workouts.

I’m comparing TR flow vs TP flow. When the workout is created in TrainingPeaks, you can see target power in both TP and Garmin after the ride. The TP workout can sync to TrainerRoad and you can only see target power in TR if performed indoors.

FWIW the difference between TR and TP structured workouts on Garmin: in the workout’s .FIT file the workout steps are defined, and the field target_type is defined as:

  • “power_lap” by TR
  • “power_3s” by TP (and by Garmin too)
1 Like

Just been signed up to the Beta

Is there an FAQ link anyone can point me to that will deal with the main ‘basic’ newbie adopter queries to save asking the forum and clogging up threads?

TIA :+1:t2:

1 Like

Brilliant one @mattonabike, sums it up perfectly…


Strange theory: is this talk of AT actually just a very elaborate hoax to get us to read 3000+ posts? Now where did I put that tinfoil hat again? :sweat_smile:


There is not. We’re happy to answer questions you have! You can also use the search function and limit it to this thread to find answers as well.


Just want to say thanks TR for the automatic TSS estimate from heart rate function! A pleasant surprise from the beta I wasn’t aware of.


It’d be good to prompt the survey manually, I’ve had it not prompt me. History on progression levels would be cool.

1 Like

I thought that but with levels being reset/adjusted whenever your FTP is modified, either via ramp test or manually or even by AT itself, then it is only useful over a short period of time.


If it were that simple, then TR would have had adaptive training years ago.

IF just correlates with progression, but it is not the same. Like others said, duration and workout structure (4 x 10 minutes = 40 minutes at FTP is very different from 1 x 40 minutes = 40 minutes at 100 % FTP) are key. Plus, there are different algorithms for the different energy systems.


I can imagine that it’s pretty linear IF to level relationship for sweat spot and endurance workouts but I don’t think it is the case for workouts made up of many shorter intervals - Machine learning was used here according to the podcast.

Machine learning is also used when it comes to prescribing future workouts. I.e deciding what workout would give the best performance improvement.

I have not used TR for a month or so (because the sun is shining!) but my inclusion into the beta has me back on the wagon.

I set up a new plan, didn’t do a ramp test because I’ve recently done a 20min test, and adaptive training immediately wants to downgrade everything. My plan is to ignore the adaptions for the first couple of weeks until TR gets a hang of my current performance and then let it start doing its thing after that.

Is this a good plan?