Bannister's TRIMP

I’ve communicated to @david that displaying Bannister’s TRIMP will be a good addition to s-SPE based load metrics previously implemented.

Bannister’s TRIMP = A.B.C
A = Duration (minutes)
B = Delta HR ratio
= (HRexercise - HRrest)/(HRmax - HRrest)
C = (0.64.e^1.92.B) for males
or (0.86.e^1.67.B) for females
e = 2.712

C is a pre-evaluated co-efficient of blood lactate rise for males and females with increase in intensity.

Bannister’s TRIMP maybe useful as a load metric and adds potential value over the current non-power load metric available, which may not generalize well to other sports outside of cycling.

BASIS:
Modeling human performance in running
October 1990
Journal of Applied Physiology: Respiratory, Environmental and Exercise Physiology
https://www.researchgate.net/publication/20910238_Modeling_human_performance_in_running

3 Likes

I have implemented this. Tx for the detailed info! It will take a while to add it to everyones activities.

You can plot it on the /fitness page with a custom chart:

Interesting that it matches my TSS based fitness line so well.

You can also add TRIMP as a column on the activity list view and filter for activities by TRIMP.

10 Likes

Thanks @david . I now see the metric in the header to a running activity but do not see it there for a cycling activity. Or is it just me?

TRIMP should be applicable for all activities, so one should be able to compare Load vs TRIMP.

I can see it for my cycling activities (although I have no other activity type)

1 Like

Sorry, I was looking at an activity earlier that didn’t record heart rate. It works.

I cross-checked the TRIMPS calculation with my own calculation, and they match.

We now have an arsenal of metrics here for our data science projects :slight_smile:

4 Likes

Thanks a lot for adding Bannister’s TRIMP.

I just played around with the Load and Bannister’s TRIMP numbers of a HR-based cycling session.
I would expect that both numbers to approx. match, but the Load number was way lower (250 Vs 380). It was set the normalized TRIMP as the HR load model (default). If I set time in zones or average HR the load value increases, average HR almost matches the TRIMP value (384).
So how should I assess all numbers. What is your impression about the most realistic load estimation?

I wouldn’t get too hung up on comparing the precision of Load and Bannister’s TRIMP. Understand the differences in the math.

Bannister’s TRIMP is a pretty well studied metric . Dependencies are exponents where co-efficients for males and females come out of a paper I referenced earlier, plus its also dependent on the resting heart rate in your settings.

Is it sensitive to individual cases?

Probably somewhat, probably not at all. If you need more sensitive metric, you may need an individualized TRIMP solution based on your own blood lactate rise with intensity. Something like that can be programmed using Golden Cheetah. For most people, this is already taking things too far.

Btw, I don’t really know where the HR Load estimation methodology comes from. I expect David to enlighten us.

In my personal case, I use TRIMP & s-RPE scores which form two subjective measures of training load. Then I combine it with some objective measure. For running, if I have pace, that’s fine. If its cycling, then power works.

The HRSS load model is from Elevate. This is TRIMP normalized to more closely match power based training load. Normal TRIMP numbers are a lot higher than TSS from power which is what Intervals.icu uses for power based load. As far as I know HRSS is TRIMP but on the same scale as TSS.

Sorry for the delayed answer.

Thanks for the link, David! This make things clear.

To me, it looks like the Load value is the normalised TRIMP:
hrPSS = Exercise TRIMP/ 1hr FTP TRIMP *100

"
Note that an hrPSS is always going to be less accurate than a calculation based on a power meter, something that is also made clear in TrainingPeaks. But I do believe that an hrTSS is better than a TRIMP, as it is comparable between different athletes, conceptually easier to understand and takes account of changes in fitness levels that alter the FTP heart rate.
"

I was though confused about his comment in the tooltip of the HRSS:
"
Either the time in each HR zone or the average HR for each activity is used. This works well for cycling but does not generalize well to other sports. For running and other sports HRSS (from Elevate) works better.
"

I find it confusing why HRSS is chosen as default for cycling.

Mostly because it gives good results with very little data (avg HR for the ride, resting HR, LTHR, max HR). Time in zones needs a lot of rides with power and HR to give good results.

1 Like