Suppose that we observe an event X, and would like to characterize its causes. In some cases, it may be that there was a single cause on which X was contingent, and in other cases there may have been many small causes. For instance, compare the death of a suicide bomber, which has a single, unambiguous cause, versus the death of an elderly person on life support, which usually has a myriad of small causes which add up. These are two extremes; in complicated systems, there will on average be a small amount of large causes, a moderate amount of moderate causes, and a large amount of small causes, though the ratio between these may vary significantly.

I call the frequency vs. power distribution the explanatory power distribution of X. Consider for instance the following picture (excuse the shittiness):

Note that the x-axis represents inverse power, so that moving probability mass to the left replaces some moderate causes with a few large causes.

Each colored line here represents the explanatory power distribution of a certain event.

• The red line represents an event predicated on a small amount of large causes. For instance, death by suicide bombing, which kills you by blowing you to pieces. Such an event has a focused explanatory power distribution.
• The orange line represents an event predicated on a few large causes, but some moderate causes as well. A bullet to the chest, for instance, may kill you primarily through disrupting the circulatory system (destroying vital blood vessels, or causing exsanguination), but death is contingent on a few other factors as well, such as collapse of a lung (pneumothorax), organ failure, and so on.
• The yellow line represents an event predicated on maybe a couple of large causes, but primarily on moderate and small causes. Gavrilo Princip's assassinating the Archduke Franz Ferdinand, which was contingent on a lot of factors, is an amazing example. (See for instance this video, 3:12 to 7:18).
• The green line represents an event predicated on some moderate causes and a lot of small causes. The outcome of a rolled fair die, for instance, is not totally random — you can learn to marginally control it — but is still contingent on the exact motion of the hand and fingers, the contours of the table, and the physical properties of the die itself.