Basically I rode ~10 minutes there, hit lap button, sat around for an hour, then rode back same ~10 minutes. No hero sprints or anything like that, in fact I was keeping it quiet because I had a TR workout scheduled for later that day. Looked at TSS and it is showing as 55, huuuuh??? Goddard is 57 and that’s a much, much, much harder effort!

Maybe numbers are skewed because of a long pause in the beginning of lap 2 and somehow the way TSS is calculated isn’t handling such a thing well? Power zones correctly shows over an hour of “coasting” so that’s there. Calories are at 170 so reasonable. Duration in the interval table is also showing correctly but start/end is revealing the “sit in the coffee shop” time.

I doubt he’s built this ride into his training schedule, he probably noticed this being odd and thought he would ask the question of our highly esteemed peers.

This is weird. The calculation for TSS is (time in seconds x NP x IF / FTP*3600)*100. For some reason this ride got a normalized power of 199, so that’s throwing the whole equation off I think. I don’t know why it didn’t count all the zeros in the middle. My completely naive opinion is that because of the very short nature of the ride with the long break in the middle, the system just wasn’t able to resolve the averages and trends. Did you upload to Strava or Training Peaks, did it give a TSS score?

did you recalibrate the powermeter before the trip back?

lap 2 makes no sense.

anyway, I"m going to double down. If this PM works fine when you calibrate and then train, then so what if it read wonky on a coffee run. don’t use a PM on a coffee run.

Yeah, it’s like the NP figured only on moving time, but the TSS was figured with that NP on total activity time.

[EDIT] NP of 199 seems plausible given the considerable number of surges on the back. But the moving time was only 19.3 minutes. So the TSS should be 100*( (1.07 x 1.07 x 19.3)/60) or 37

I was just making sure power meter still works good, haven’t used it fora few weeks. The sync happens automatically so yeah, I’ll delete it so my weekly TSS is correct. The point is I shouldn’t have to do that.

Power meter is Powertap G3 so very unlikely it’s off.

TSS is based on your NP and NP is based on your FTP. The issue is that short duration efforts have little correlation with FTP. This ride looks like you smashed it for 10 mins to the coffee shop and then smashed it back. So you’re calculating a NP number from basically 2 * 10 min sprints, which isn’t what NP was designed to accurately measure.

TSS is just an extrapolation of that since TSS = (IF)2 x Ride Time (hours) x 100. So your sprint efforts artificially inflate the IF of the ride and then it just multiplies out from there.

I split that workout in Strava into 2 without the gap, and they evaluate into 19 and 13 TSS. Can’t import them both since they claim to be duplicate workout so had to do import/delete on those. Also TrainingPeaks TSS for that same ride (coming from Wahoo Android app) is 33 without any modifications. So TSS of 55 is a fluke.

I didn’t hammer the coffee shop run, in TrainingPeaks overall NP shows as 153 which is well below my FTP.

Send email to support@trainerroad.com with a link to your ride - I’ve had a few rides with bad TSS in TR, but good TSS in TrainingPeaks/WKO. They should be able to help.

Thanks for the super useful link @julianoliver
I started the thread thinking I’m just not understanding something about TSS, and yea maybe this wan’t what I wasn’t understanding but I was not understanding a bunch of other stuff, so no regrets!

Consider how small the amount data is, we should expect a very small delta. If this was an hour long ride with a two hour break then another hour long ride, the burrito and beer effect would be greater.

Not sure if that’s it in this case, I’m comparing TSS as calculated by TrainingPeaks to 2 figures calculated by TrainerRoad, which we now know don’t match exactly. Also such small samples, could be just rounding error. But yeah just by pure math alone if I understand it right, averaging the (whole sample set)^4 and then taking ^1/4 of that vs. doing it to splits and then summing those up would yield such results. It is some pretty simple math and there’s nothing magic about it, one could use ^5 instead of ^4 and get more “reward” for hard efforts in TSS calculation for example. This all strikes me as pretty un-scientific but good enough to make some sense of my training.

Here’s a simple demonstration of this math. Say you had a 2 minute ride with 30-second average power values of 2, 3, 4, 5. Let’s compare, calculate NP once for the whole ride, or calculate each minute separately and add those up. We get 7.9 for the single span and 7.22 for the sum of two spans.