Power-Law Distribution

Overview

A power-law distribution is a statistical distribution where probability of size x is proportional to x^(−α): large events are rare but far more probable than a Gaussian model predicts, and the largest events dominate the total — there is no "typical" case.

First quantified by Pareto (1896) in wealth; formalized by Mandelbrot (1963) for financial returns; surveyed universally by Newman (2005) across cities, earthquakes, citations, and web traffic.

Composes with pareto-principle (80-20 is the most famous application; this skill provides the math foundation), black-swan (black swans are the extreme upper-tail events power laws make far more probable), expected-value-and-kelly (Kelly sizing breaks under infinite-variance power laws), and antifragile (antifragile strategies exploit the upper tail).

When to Use

• Allocating capital or resources across a portfolio — power-law returns mean design must prioritize outliers
• Prioritizing customers, channels, content, or features where a small number account for most value
• Assessing business risk — Gaussian risk models (VaR, std dev) systematically underestimate extreme risk
• Any domain where "average" is the planning assumption and extreme outcomes are possible
• Evaluating AI/compute concentration — AI capex, chip export controls, frontier-lab funding, or "AI bubble" questions where value is capturing into a thin tail of companies

Not when: distribution is demonstrably Gaussian; stakes are low enough that shape doesn't affect the decision; audience will misuse power-law framing as nihilism.