Weigh the boxplot dots.

[a-t-0]

Initial concern:

• Instead of having 1 dot that represents: 40 right 1 wrong, and another dot 200 right 3 wrong, make 41 dots at position 1/40 and 203 dots at position 3/200. (Or make the dot size a function of the number of resuts.)
• Verify all snns have the same number of dots per boxplot.

[a-t-0]

Instead of weighing the box plots, run on graph sizes that support an equal nr of graphs.

• [x] E.g 10 graphs per graph size, 10 graph sizes per seed, 100 random seeds.
• [ ] Write an assert that verfies there were indeed 10 unique input graphs.
• [ ] Write an assert that verifies that the 100 random seeds indeed lead to 100 different orderings of the random values over the nodes, per unique graph. (Otherwise, multiple seeds are ran on the exact same "random graph initialisation". ) That is not bad, because the radiation is also random with that same seed (which differs over that same graph, yielding different radiation effects, hence a different simulation).