I would have appreciated that blog post more if the author had explicitly stated that those charts weren't based on any actual data. It feels scummy to plot random points just to make a conversational point.
It's already explicit that it's a computational model. I can still edit it to emphasize that more, and did.
A visual graph is the clearest way to show that correlation usually doesn't hold as a threshold is raised. Much clearer than equations or handwave-y English.