Apparently, Cactus plots can be confusing to audiences.
To see just how confusing, take this talk of Armin Biere at the 2014 BIRS workshop, in front of a partially non-SAT audience.
The moment he shows a Cactus plot (min 25:55), chaos errupts. People are yelling, questioning, shouting – it’s really quite intense. They can just not make sense of the plot and demand a thorough explanation. It takes Armin three (!!) minutes and multiple rounds of explantions to calm the crowd sufficiently to be able to somehow continue. Later Karem Sakalla even apologizes in advance (min 5:08) for the upcoming plots, and begs for no further questions regarding them.
I assume seasoned SAT researchers can recall more stories of this kind. Something about the cactus plot can clash with peoples minds. This is one of the advantages of CDFs. Eventhough they present exactly the same data, just flipped, my impression is that CDFs are better received. Maybe due to the more intuitive higher = better relationship.
Does this detail even matter? Maybe not so much at your average workshop. Maybe a bit more in front of a funding agency.
Psychological damage aside, cactus plots also have their advantages: They make it easier to spot performance differences between different curves, as x-axes tend to be longer than y-axes. So eventhough I will probably mostly go with CDFs, it is a small bummer that they squish the data bit more than is in principle necessary, unless I make my plots taller than wide.
