In the paper he refers to something called Meyers pacing model for the marathon. Apparently an engineer, George Meyers and Hal Higdon developed a model for the typical pacing seen in elite runners and also non-elite runners.
The Meyers model breaks the marathon down into 4 sections as shown below.
I wrote a routine that lets me enter my marathon time and creates a 'fake' set of run data with the same finishing time and Meyers paces for each distance range.
This is what the unsmoothed data would look like for a 3:54:54 marathon:
Of course, we don't really slow down in jumps like that so I do some smoothing to get the same net time but more evenly. Here I have plotted the smoothed Meyers pacing vs my Eugene performance:
I find that a darn good match to what I actually ran for the last miles. The two bumps in miles 1-12 are due to the hill at mile 4-5 and mile 11.
(Coming up: Mr J.B. Elliots paper is actually about hill corrections to the Meyers pacing model..I plan to implement those next and see what that looks like. 8)
I think that for an optimal marathon you probably have to slow down a bit in the last miles, i.e. have slightly positive splits to run your theoretical fastest.
My conjecture is: it's better to go a teeny bit faster when your running economy is good (e.g muscles are springier ) the day is cool, etc and then accept that you will slow down. (even though you fight it as much as possible).
Right now I don't think we understand the system well enough to figure this out for sure. But I know my 'springs' (legs) aren't very springy past mile 20!
I know it feels a lot better to run even or negative splits. But I think we are leaving performance on the table when that happens. I'm talking about very modest positive splits ..less than 1%.
On the other hand, these model was created from people out there doing their thing, and their thing might just be sub-optimal.
What do you think?
Assuming a flat course, what kind of splits would you try for? For your best marathon times what did you do?