🎉🎉🎉 Introducing Machine Learning-Based TSS Estimation! 🎉🎉🎉

Is the other way around possible? Override all previous RPE scores with the new ML Prediction without having to go through them one by one?

Currently we don’t have that option available.

The model was trained across different riding styles and disciplines, but doesn’t differentiate what type of ride it’s looking at individually for each workout. I wouldn’t worry about any margin of error specifically introduced by this variable, since it’s essentially “baked in” to the system already.

1 Like

What is net effect of TSS Estimation on your training plan if, like me, you’re still NOT :frowning_face: yet in AT Beta?

TSS does not proactively impact your training plan, whether you’re on Adaptive Training or not. We like to think of TSS as a way to track broad trends over time, and a tool for you to monitor your training load as you progress through the season— especially if you’re adding unstructured or unplanned rides.


I still can’t change max heart rate, stuck at 196, change to what it is (189) and change is lost when leave page …

The fix is in progress but things still aren’t back to normal across the board. It does sound like things are working better for most athletes, but I’ll let you know when I hear that things are fully sorted out.

The TSS graph only has an option of viewing the past 4 years of TSS. Is there a way to view progression spanning a much longer career?

Hi Sean, its generally working and giving me roughly the same hrTSS as TrainingPeaks. So I’m happy to finally see hrTSS on TR.

Randomly pulled up February 2016:

and TR:

However Saturday the 21st wasn’t updated automatically:

That was a random spot check. Hovering mouse over the tile and I was able to click “Add TSS” and tell it to estimate using HR. Pulling up Past Rides there are 51 w/o TSS, but the first 5 had no HR data.

Overall looks good to me - TR’s hrTSS closely aligns with TP hrTSS, and it appears to have automatically updated past rides (with 1 known exception).

1 Like

Date of Birth has to be in American MM/DD/YYY format but I doubt it will make any difference if you put 7/7/1990 rather than 7/21/1990 or 21/7/1990.

I did suggest having individual select drop down boxes for each value then it’s immaterial what the underlying format is.

1 Like

Hello! I was just trying this new feature, and I have the correct max heart rate set (183) and DOB, and today I did an 4 hour ride but the TSS seems off, it’s giving a 34 TSS, but if I use RPE, it’s showing me more or less accurate based on how I felt. Is there something that I’m doing wrong?

Same here - FWIW I use the TR website using Chrome on iPad. But it also reproduces easily with Chrome on Windows 10 as well. I click the ‘all past and future rides’ option, get a confirmation dialog, click OK, then see a further ‘option saved’ confirmation in the UI - then going back later to revisit the option it shows ‘I will decide per ride’ again.


Shoot an email over to support@trainerroad.com and they’ll take a look. As of end of day Friday there were still some residual bugs popping up so I suspect this may be one.

1 Like

There still seem to be some bugs happening for a few athletes, mostly with the “all past and future rides” option. Our engineers will be back on it tomorrow morning, I’ll let you know when it’s totally resolved.

I’ve had the same. I’ve raised with support but just adding to the thread :+1:

1 Like

Seems a bit off for me as well. 2.5 hour MTB ride, 45km, 1000m climbing during which I completed 3x 12min threshold intervals (based on HR / RPE), roughly tempo for most of the rest of the ride (apart from downhills). Estimated ~60 TSS puts it effectively as a recovery ride, which it definitely was not.

I’ll ask support about it as well.

Yep, still the same for me at the moment.

1 Like

As others have commented, I did a multi-hour ride and the ML estimate of the TSS was equivalent to a 1-hr very light recovery ride. I also had issues trying to set my Max HR and DOB. Will stick with the RPE estimate for the time being.


I did four races so far; all were XC marathon competitions 50-70 km long events; 2-3,30 hr to complete depending on the course.
I’m 45 y.o, my max cycling heart rate is 182 b.p.m, lactate threshold is 168-172 b.p.m.

I did all my races at an average heart rate of 168-170 b.pm, based on the RPE scale Very Hard or All Out efforts.

For these efforts, the system offers me approx 50-60% from the manually set TSS (based on RPE).

I cannot agree that my last 68 km marathon, which I did for 3,30 hr, is estimated as a 150 TSS effort. For comparison: RPE-based effort (All OUT option) is weighted 314 TSS, and this is a pretty accurate score.

At the moment, the automatically calculated TSS is too way off for me.

1 Like

Our engineers found the problem, which relates to how some rides store and save HR data. I’m told the fix is being implemented and that we’ll be automatically reprocessing rides with the mistakenly low estimates once it’s in place. I’ll let you all know when it’s done for you to verify your rides are fixed.