Forgive me if this seems insulting, but I think the answer to your
question is, or should be, self-evident: you need to account for all the
periods when your power is zero, because they influence how much power
you can generate when you actually are pedaling.


To resurrect a previously-discussed topic as an example: it was shown
back in the 1970's that as long as the exercise (and rest) period was
kept very short, say 10 seconds or less, then the physiological and
metabolic responses to intermittent exercise is nearly identical to
continuous exercise at the same average power. That is, an hour of
exercise performed at 400 W for 10 seconds, rest 10 seconds, ad nauseum
will elicit essentially the same responses as continuous exercise at 200
W. This makes sense when you consider both the pedaling and non-pedaling
periods (as well as the half-lives of various physiological responses),
but doesn't make sense at all if you use only the power that was
generated during the non-zero periods (i.e., 400 W vs. 200 W).
While power output during "free range" cycling is rarely so structured,
the same principle applies...hence, you need to include the zero values
into the calculation of the overall average power output. As was
discussed at the Philly seminar, and previously in detail on this list,
this still doesn't fully account for the "stochastic" nature of cycling
power output, but nonetheless the average including all zero values
still says much more about the actual demands than using the average of
just the non-zero values.