In the first part of this column, I explain Simpson’s Paradox. This is the situation when Software System A is worse in every area of comparison than Software System B, yet System A can still be shown to be the better system. In the second part, I demonstrate Braess’s Paradox, which is when adding a perfectly good load-balancing server actually reduces network performance in a most surprising way. I finish up with a look at Parrondo’s Paradox, where two completely independent but faulty systems can, incredibly, produce a correct system. Simpson’s Paradox occurs quite frequently. Although the chances of you actually encountering Braess’s or Parrondo’s Paradoxes are slim, I think you’ll find reading about them quite entertaining.
It is quite an enjoyable and interesting read.