The goal is to find out the formula to calculate the appropriate zoom level,
or equivalent: to calculate the size of the world viewport depending on, well,
dunno.

Data collected using bea9a124317, on 2015-08-25 *iirc*. (-> stats.pickle.xz)

Data analyzed using ea4515d0516, with:

	python analyze.py stats.pickle > temp

then manually, create win.{size,mass}.{1,2,3,4,5,6} by cutting the
analyze.py output into pieces (depending on number of cells).

view with

	gnuplot win.{size,mass}.gnuplot

the gray lines in win.size.gnuplot denote the filtering conditions of
filter_data.py

Then merge and filter the data using

	cat win.size.[123456] | python filter_data.py > win.size.all.filtered

win.size.gnuplot will automatically fit a "a * size**b" - like function.

Result: visible diagonal = 369.399 * sum(sizes)**0.431776

Also, analyze.py tells us that the viewport ratio is 1.7 : 1

Additionally, if size<70, then the value 70 instead of "size" is used (found
out by tinkering around manually)