Extreme Value Theory

Extreme Value Theory (EVT) is a branch of statistics concerned with the behavior of the tails of probability distributions — the rare events that lie far from the mean. While the law of large numbers and the central limit theorem describe what happens at the center of distributions, EVT describes what happens at the extremes. It answers questions like: what is the probability of a flood exceeding the historical maximum? What is the expected loss in the worst 1% of trading days?

The theory was developed in the early 20th century by statisticians including Ronald Fisher, Leonard Tippett, and later Emil Gumbel. The foundational result is the Extreme Value Theorem, which states that the maximum of a large number of independent random variables, properly normalized, converges to one of three possible distributions: the Gumbel, Fréchet, or Weibull families. Collectively known as the Generalized Extreme Value (GEV) distribution, these three types capture the asymptotic behavior of extremes across a wide range of underlying distributions.

EVT is essential in domains where rare events dominate outcomes: insurance catastrophe modeling, financial risk management, structural engineering, and climate science. The theory is particularly valuable because it allows extrapolation beyond observed data — estimating the probability of events that have never occurred but could. This extrapolation is dangerous when misapplied, but indispensable when grounded in sound theory.

Extreme value theory is statistics for the apocalypse. It does not tell us what usually happens; it tells us what could happen, and how bad it could get. The central limit theorem comforts us with averages; extreme value theory warns us that the average is not the enemy — the outlier is.