Mediocristan vs Extremistan

Taleb's distinction between two fundamentally different types of domains. In Mediocristan, outcomes are bounded and bell-curve distributed—no single observation can dramatically change the aggregate (human height, calorie consumption, exam scores). In Extremistan, outcomes are unbounded and follow power laws—a single observation can dominate the total (wealth distribution, book sales, casualty counts, startup valuations). The critical error is applying Mediocristan tools (averages, standard deviations, bell curves) to Extremistan domains where they're dangerously misleading.

When to use it

When choosing analytical tools—bell curves vs. power laws; when evaluating whether averages are meaningful for a particular metric; when designing strategy in winner-take-most markets; when assessing risk in domains where extreme events dominate outcomes.

How it can help

For any domain you're analyzing, determine: Mediocristan or Extremistan? If a single observation can dominate the total, you're in Extremistan and standard statistics will mislead you. Most business metrics (revenue, returns, virality, impact) are Extremistan: they're dominated by outliers, not averages. In Extremistan, focus on positioning for positive outliers and protecting against negative ones—not on optimizing the average.