Hello all,

I have been lurking over the last few days and reading the comments from
both Will Hopkins and Andy Coggan on the power normalization approach
for non-steady state cycling that Dr. Coggan has developed. One key
word in the discussion has been "arbitrary". Hopkins says the choices
for the model parameters (4th power weighting and 30 sec averaging) are
arbitrary. Coggan says they are absolutely not arbitrary, even if not
formally tested in a controlled study. Here I would side with Coggan,
if only because I think I know him well enough to say that he is not a
guy who picks numbers out of the air.

But, the bottom line is that essentially ALL of the methods we use for
training load estimation are based on more or less arbitrary weighting
scales that are not derived directly from experimental data. Experience
suggests that simple is superior to complex (and there have been some
complex mathematical models proposed for quantifying training load,
believe me). For example, Carl Foster's session RPE scale, which I have
used to quantify training load both in published studies and practical
work with Olympic level athletes, is based on a 10 point perceptual
rating scale with verbal anchors. What is the actual "distance" between
a 3 and 4 on the scale, a 7 and 8? Don't know. Why collect the data at
exactly 30 minutes after the end of the training session and not 15
minutes or 60? Not sure. But I am convinced that the method gives
meaningful information and shows good validity when compared with
physiological measures. On the other hand, in my mind, any factor
weighted "time-in-zone" approach using HR is VERY arbitrary. Straight HR
based time-in-zone will always underestimate the stress of interval or
intermittant exercise. And, linear weightings like 1 for 50-60%, 2 for
60-70% and so on STILL probably underestimate it, for several reasons.
One published study we did where we quantified training load in XC
skiers suggested that straight "time-in-zone" estimation of training
load fit very poorly with the perceptual ratings of training stress
provided by the athletes (session RPE). The problem is particularly
apparent when working with elite athletes who may warm up for 20-30
minutes and cool down similarly, such that even their hard interval
sessions come out looking "easy" when HR based time-in-zone is
calculated for the entire session.

Which brings me back to Coggan's power normalization approach. I confess
that I am just now checking out. I'm a rower, not a cyclist! After a
couple days of ruminating and then simulating some data in EXCEL, I have
to say I like it. The numbers behave in a manner that is consistent
with my experience from 2 studies on interval training responses and
other data. For example, I used the method to normalize power from an
imagined, but representative, ergometer interval session:

Athlete had power at VO2 max of 450 watts
He performed six, 4 minute intervals at 89% of this power (400 watts)
separated by 4 minutes active recovery periods at 35% of PVO2max (160
watts). His average power for the 48 minute period was 280 watts.
After 4th power normalization (and assuming each of my cell values was a
30 second average for power), the normalized power for this session was
338 watts, or 21% higher. Now, what if I go to the extreme case and
model total rest during the recovery periods (0 watts), and still 400
watts during the work intervals. The actual average power drops to 200
watts (quite easy for this subject) The normalized power for the session
is almost unchanged (336 watts). I believe the picture the normalized
power comparison paints is quite consistent with the underlying
physiology in the sense that zero intensity during the recovery periods
would not be expected to dramatically reduce the stress of the session.
Indeed, keeping the legs spinning a bit is thought to improve lactate
elimination as we all know. 3rd and 2nd power normalization is more
effected by the low power values. 4th power modelling clearly heavily
weights the impact of high power outputs as it should. After this and
some other simulations, I see why Andy chose 4th power. What I am less
clear over is the statement that the difference is only about 5% for 2nd
to 4th power. Using my extreme example of intermittent exercise, the
differences were bigger (19% differene in normalized power for 4th vs
2nd order normalization in my extreme example of 400watt/0 watt interval
exercise. However, when I modelled an athlete doing 4minute bouts at
350w and 250w, the actual average of 300 watts was normalized to 312,
291, and 304 watts, using 4th, 3rd, and 2nd power normalization
repectively. And, yes, I checked this. 3rd power normalization
actually resulted in a normalized power that was slightly below the
actual mean for this condition. Numbers behave in curious ways
sometimes. But, the point is that for this less extreme condition of
stochastic loading, the differences associated with the choice of
exponent were indeed very small, just as Andy suggested. Still though,
4th power would win my vote.

Now, obviously the average power of a training session only has meaning
when combined with some measure of the duration. Does the stress impact
increase linearly with time or non-linearly? When we measured RPE
during two studies employing similar 6 x 4 minute interval sessions, RPE
increased linearly with each work bout. So, that data supports the
TTS approach of just multiplying normalized power x exercise duration.
However, we recently examined autonomic recovery (using Heart rate
variability measures) following controlled time exercise bouts performed
under VT1 (1mM lactate, 60% VO2 max), between VT1 and VT2 (3mM lactate,
86% VO2 max) and over VT2 (intervals at 95% VO2 max and 7mM lactate).
We found that in these highly trained runners 1) increasing work
duration from 60 to 120 minutes at 60% VO2 max had essentially no impact
on session RPE or the rapidity of autonomic recovery after the bout, but
2) as soon as intensity increased above VT1, autonomic recovery was
delayed significantly but similarly ("identical" time course) after
lactate threshold and hard interval training sessions. THIS data
suggests to me that maybe the time factor used to calculate Total
Training Stress (TTS) should perhaps be different for low intensity
exercise in highly trained athletes. But what should the weighting
factor be? Arbitrary.

My 2 cents worth, back to work.


Stephen Seiler PhD FACSM
Associate Professor
Faculty of Health and Sport Sciences
Agder University College

Service box 422
4604 Kristiansand
Norway

tel: 47 3814 1347
fax: 47 3814 1301