Mediocristan vs. Extremistan

  • Mediocristan refers to domains where deviations are small and predictable (example: human height, weight).
  • Extremistan refers to areas where a single event can have a massive impact (example: financial markets, wars, technological revolutions).

Most of modern life is influenced by Extremistan, meaning Black Swans are more common than we assume. Reminder, a Black Swan event:

  • It is highly improbable.
  • It has an extreme impact.
  • It is unpredictable, but in hindsight, people try to rationalize it as if it was predictable.

Weak-link vs strong-link problems

Weak-link problems are problems where the overall quality depends on how good the worst stuff is. You fix weak-link problems by making the weakest links stronger, or by eliminating them entirely (example: food safety, nuclear proliferation, space missions safety, football).

In strong-link problems, the overall quality depends on how good the best stuff is, and the bad stuff barely matters (example: music, science, research, basketball).

Strong link: improve the best, ignore the worst. Weak link: improve the worst, ignore the best. Image source: Experimental History (according to the author, “science is a strong-link problem, but most people treat it like it’s a weak-link problem”).

Making the link

Weak-link problems are more likely to happen in Mediocristan:

  • The system’s performance depends on improving the weakest part rather than relying on one exceptional element.
  • Small, incremental improvements matter more than single outliers.
  • Black Swans do not play a major role—consistency is more important.

Strong-link problems are more likely to happen in Extremistan:

  • The system is shaped by a few extreme outliers rather than the overall average.
  • A single individual or event can dominate and change everything (a few star investors, a group of scientists making a major breakthrough, etc).
  • Black Swans are crucial—a breakthrough or disaster can redefine the entire landscape.
Visual done in napkin.ai