Risk You Can Estimate, And Risk You Cannot

Confusing the two is the most expensive mistake in the whole subject.

Some uncertainties have known shape. You may not know which face a die will show, but you know the possible faces and that nothing else can happen. Others have no known shape at all. You do not know the range of outcomes, you do not know what could be involved, and you may not even know which questions to ask. These are not two points on one scale. They are different kinds of not knowing, and the tools that work on the first are close to useless on the second. Yet we habitually smuggle the second into the first by inventing a range for it and then reasoning as though the range were solid.

The tell is usually confidence about the edges. If someone can tell you exactly how bad the worst case could be for something genuinely novel, they are not estimating, they are decorating. Real deep uncertainty shows itself as an inability to bound the outcome, and the honest response is to say so, then to build in slack rather than precision. Slack is what you have instead of knowledge.

Practically, the split changes what you do. For shaped uncertainty, you can plan against the range and accept the variation. For unshaped uncertainty, you make yourself hard to break: keep reserves, avoid arrangements that fail all at once, prefer moves you can reverse, and keep the size of any single exposure small enough that being wrong is survivable. That is not pessimism. It is the recognition that you cannot compute your way out of a question whose shape you do not know.