It’s a famous example of survivorship bias. The idea is, adding armor to planes make them heavier, so you want to minimize where you put armor. After some flights, you take note of where the bullet holes are in the planes that come back.

Where do you put more armor? Do you put it where there are the most bullet holes? That seems to be where the planes are being shot the most.

The problem is, your sample isn’t representative of your underlying population. These are the planes that came back. If they get shot it the cockpit, they die.

So, where should you put the armor? Well if they can get shot and come back, it’s not all that important, so put it everywhere else.