just as there are three (yeah, I know, four) states of matter: solid, liquid, and gas, we can classify domains based on the type of randomness they exhibit.
Benoit Mandelbrot organizes randomness in seven states but, at the core, there are three: mild, slow, and wild.
At one side sits mild randomness. This is the normal distribution. The stuff we learn in basic stats and looks so tractable that we use its computation techniques on everything we do.
Somewhere in the middle is slow randomness. It has extreme variations, but not infinite (or near infinite) variation. Floods for example. Severe floods make future flooding more likely, but it's unlikely that a single flood will exceed the cumulative of all previous floods combined.
On the other extreme is wild randomness. Here's an excerpt from Mark Buchanan's Ubiquity: Why Catastrophes Happen:
"Imagine wandering into the street, anticipating how tall the next person might be. If people’s heights worked like these avalanches, then the next person might be less than an inch tall, or over a mile high. You might crush the next person like an insect before seeing him or her. Or imagine that the duration of your trips home from work went this way; you’d be unable to plan your life, since tomorrow evening’s journey might take anything from a few seconds to a few years. This is a rather dramatic kind of unpredictability, to say the least."
It did a good job driving the point home for me. Individuals' net worth is another example, but we are so accustomed to its wild variation that we consider it "normal".