Cabinet · Case IX — The arithmetic of luck

The Pioneers of Probability

For four thousand years humanity played with chance without measuring it. Then, within a single century, gamblers’ questions became a science.

The scandal of ancient gaming is that nobody could count. Romans, Indians, Chinese and medieval Europeans all wagered constantly on dice and cards, yet before the Renaissance there is no known correct calculation of odds in any literature on Earth — not the number of ways three dice can fall, not the fair division of stakes in an interrupted game. The mathematics of chance is the youngest branch of ancient mathematics, and it was invented, embarrassingly late, by people who wanted to gamble better. Insurance, statistics, polling, genetics, thermodynamics, machine learning: all are descendants of that unpromising maternity.

Cardano: the gambler’s book

First comes Girolamo Cardano (1501–1576), the Milanese physician, algebraist and compulsive gambler whose posthumous Liber de ludo aleae (“Book on Games of Chance”), written around 1564 and published only in 1663, contains the first reasoned treatment of dice probabilities. Cardano grasped the central idea — count equally likely outcomes — and computed correctly for two and three dice, along with frank chapters on cheating and his own losses. His rule of proportion for stakes anticipates expectation. It is a strange, brilliant, morally queasy book: its author believed simultaneously that fortune was ruled by the stars and that it could be counted like coins, which is precisely the historical hinge this cabinet documents.

1654: the problem of points

The founding document is a set of letters. In 1654 Blaise Pascal, mathematician and religious obsessive, wrote to Pierre de Fermat, lawyer in Toulouse, about the problem of points: two players quit a game early — how should the pot be divided? The question, posed to Pascal by the Chevalier de Méré (a salon gambler whose instincts about the odds of dice problems were better than his arguments), is at least four centuries old; Italian abacists had fumbled it since the 1400s. Pascal and Fermat, corresponding through the summer, solved it two different ways — Fermat by listing outcomes, Pascal by his arithmetic triangle of combinations — and in doing so created expectation, the concept on which all risk-bearing thought since has rested. The physicist Christiaan Huygens, visiting Paris and hearing of the letters, wrote the first printed probability text in 1657, the De ratiociniis in ludo aleae, which taught Europe the new arithmetic.

Bernoulli and the long run

The next leap answered a philosophical objection: an odds calculation is a statement about idealized chances — what do they have to do with the actual world? Jacob Bernoulli spent twenty years on the question; his Ars Conjectandi, published posthumously in 1713, proved the law of large numbers: as the number of trials grows, the observed frequency of an event converges toward its true probability. The dice of Ur, thrown long enough, confess their own geometry. Bernoulli’s theorem is why the house always wins in the long run — the arithmetic behind the green zero and the payout table of every lottery — and why statistics can learn anything at all. Abraham de Moivre then found the shape of chance itself, the normal curve, and estimated how fast frequencies settle; Thomas Bayes (1763) inverted the reasoning, learning causes from effects; Laplace bundled the whole into the first great treatise. The gamblers’ tool had become the engine of actuarial science, astronomy, jurisprudence and the state.

What the theory actually says

Two consequences deserve their museum labels. First: expectation is an average over a long run, not a prophecy. The wheel owes the player nothing this evening; it pays its debts only to patience or to the house — which is why this archive documents games rather than promoting them. Second: probability does not abolish the oldest questions. Pascal himself, having founded the calculus of chances, spun it into his famous wager on God’s existence — the first time probability was aimed at the same questions the astragalus-throwers had asked their bones. The tools changed; the questions stayed human, as the arts drawer keeps demonstrating.

From dice to Monte-Carlo methods

The 20th century closed the circle in style. Physicists at Los Alamos, needing to compute what equations could not, simulated their problems with random numbers and named the technique after the world’s casino capital: Monte-Carlo methods now run everything from nuclear reactors to the rendering of light in films. Randomness, once the property of oracles, had become a tool you apply. Meanwhile game theory — von Neumann and Morgenstern, 1944 — gave strategic play its own mathematics, and computers learned backgammon and chess. Every one of those developments quotes the founding century: count outcomes, weight by probability, sum.

Reading the mathematics historically

This drawer keeps its mathematics narrative, not technical: the point is that a body of knowledge usually credited to physics and insurance was commissioned by dice. For the instruments that posed the questions, see dice and cards; for the institutions that industrialized the answers, roulette and lotteries; for the century-by-century order of events, the timeline; for the terms — expectation, law of large numbers, problem of points — the glossary. The whole arc, from bone to theorem, is told in the short history of games of chance.