Over on the " Graphical Representation of Training Load & Adaptation" thread, a few people have been critical of Dr. Steve McGregor's normalized graded pace (NGP)/rTSS calculation (http://home.trainingpeaks.com/...rmalized-graded-pace). As it so happens, he and I chatted for about an hour yesterday, during which he made the point that rolling NGP should be essentially equivalent to Stryd's estimated running power. Since I'm doing a free webinar tonight on the use of WKO4 to analyze running data (https://attendee.gotowebinar.com/.../6892692532430862082), I thought I'd throw together a new chart to test Steve's hypothesis.
The pic below shows some typical results. As you can see, generally speaking rolling normalized graded speed (i.e., the inverse of rolling NGP, shown as a purple line) and power-to-weight ratio (shown as a yellow line) parallel each other rather nicely. The biggest difference would seem to be during the long, gradual downhill between ~27 and ~38 min, during which NGP seems to give more credit to the effects of gravity than does Stryd's algorithm. Based on the heart rate response, you could argue that the directly-measured power may be closer to the truth, but I'm not sure that is really a valid conclusion, given the influence of cardiac drift, etc.
So what does this mean for rTSS? As a first approximation/corollary to TSS, I think these data suggest that Steve's approach is actually rather good (note that rTSS also includes a correction or adjustment for VO2 drift, i.e., the formula is a bit more complicated than for plain ol' TSS). IOW, for a pure runner, using rTSS as an input to the PMC definitely has merit (as Steve has shown: https://www.ncbi.nlm.nih.gov/pubmed/19910822). However, this still doesn't mean you can or should add TSS and rTSS, since the stresses created by cycling and running are not equivalent.
ETA: Note that I smoothed the power/weight data using an exponentially-weighted moving average. I don't know if NGP uses a rolling or exponentially-weight average, but for visual comparison at least it doesn't really matter.
The pic below shows some typical results. As you can see, generally speaking rolling normalized graded speed (i.e., the inverse of rolling NGP, shown as a purple line) and power-to-weight ratio (shown as a yellow line) parallel each other rather nicely. The biggest difference would seem to be during the long, gradual downhill between ~27 and ~38 min, during which NGP seems to give more credit to the effects of gravity than does Stryd's algorithm. Based on the heart rate response, you could argue that the directly-measured power may be closer to the truth, but I'm not sure that is really a valid conclusion, given the influence of cardiac drift, etc.
So what does this mean for rTSS? As a first approximation/corollary to TSS, I think these data suggest that Steve's approach is actually rather good (note that rTSS also includes a correction or adjustment for VO2 drift, i.e., the formula is a bit more complicated than for plain ol' TSS). IOW, for a pure runner, using rTSS as an input to the PMC definitely has merit (as Steve has shown: https://www.ncbi.nlm.nih.gov/pubmed/19910822). However, this still doesn't mean you can or should add TSS and rTSS, since the stresses created by cycling and running are not equivalent.
ETA: Note that I smoothed the power/weight data using an exponentially-weighted moving average. I don't know if NGP uses a rolling or exponentially-weight average, but for visual comparison at least it doesn't really matter.
Last edited by:
Andrew Coggan: Apr 28, 16 12:49