The Law of Large Numbers
Jacob Bernoulli's proof that observed frequencies converge to true probabilities as sample size grows — the mathematical foundation of insurance and statistics.
Bernstein covers Jacob Bernoulli's law of large numbers, published posthumously in 1713, as the next major conceptual leap after Pascal and Fermat: a rigorous mathematical proof that as the number of observed trials of a repeatable event grows large, the observed frequency of an outcome converges toward its true underlying probability. This sounds close to obvious stated today, but Bernoulli's contribution was proving it rigorously rather than simply assuming or observing it informally — establishing that probability calculated from theory and frequency observed from real-world data are, given enough observations, the same thing, a bridge between abstract mathematics and empirical measurement that the whole rest of statistics depends on.
The book connects this directly to the rise of the insurance industry — insurers need exactly this bridge to function at all, since pricing a policy correctly requires trusting that the frequency of past claims (deaths, fires, shipwrecks) observed across a large enough pool of prior cases is a reliable guide to future probability, not just a coincidental pattern in one particular batch of historical data. Bernstein traces the growth of actuarial science and early life-insurance pricing directly to practitioners applying Bernoulli's theorem, often without full awareness of its formal mathematical justification.
The larger the sample of past observations, the more reliably the observed frequency reflects the true underlying probability — the mathematical basis for using historical claims data to price insurance.
Bernstein flags a common popular misunderstanding worth being precise about: the law of large numbers describes convergence over a large number of observations, not a correction that kicks in over a small handful of them — it provides no guarantee that a short run of results will look anything like the true underlying probability, the opposite of the popular gambler's-fallacy intuition that a coin "owes" a tails after several heads in a row. This distinction matters directly for insurance and any other application built on the theorem: the pool of past observations has to be genuinely large before the convergence the theorem promises can be trusted, and applying it to too small a sample produces exactly the kind of false confidence Bernstein warns about throughout the book.
- Jacob Bernoulli's law of large numbers (1713) proved that observed frequency converges to true probability as the number of observations grows.
- This provided the rigorous mathematical bridge between abstract probability theory and empirically observed real-world data.
- The insurance industry's entire pricing model depends on this bridge — trusting that historical claims frequency across a large pool reliably predicts future risk.