When you making predictions about wide ranges of possibilities along many axes, it's very reasonable to choose the single most likely option on each of these, b

When you making predictions about wide ranges of possibilities along many axes, it's very reasonable to choose the single most likely option on each of these, but you end up multiplying some 10% chance by a 20% chance by another 10% chance N layers deep, and the net result still ends up vanishingly unlikely. We are almost certainly not in the most likely possible world. By example, most users on this forum are men from the US born in the 80s or 90s. The most common first name in the US those decades was Mike. The most common surname in the US is Smith. The most populous county in the US is LA County. The single most likely combination of these is that you are Mike Smith from LA County, but you are probably not.