In the winter of 1946, the Polish-born Jewish mathematician Stanislaw Ulam was recovering from a serious illness and passing the long convalescent hours with Canfield solitaire. Somewhere in the dealing, a question occurred to him: what are the chances that a layout of 52 cards will come out successfully? He attacked it the way a mathematician is trained to, through pure combinatorics, and got nowhere. Then came the thought he later described as more practical than abstract thinking: lay the game out a hundred times, and simply observe and count.
The idea followed him back to Los Alamos, where John von Neumann saw immediately what it was worth. The physics problems of the day, neutron diffusion above all, were too tangled for closed solutions, but they could be imitated: run thousands of randomized trials and read the probabilities off the results. The technique needed a name, and their colleague Nicholas Metropolis supplied one, a suggestion, he wrote later, “not unrelated to the fact” that Ulam had an uncle who would borrow money from relatives because he “just had to go to Monte Carlo.” The name stuck.
There was precedent for the borrowing. Formal probability entered mathematics through play in the first place: in 1654 the Chevalier de Méré, a French nobleman with an expensive interest in games of chance, brought Blaise Pascal a puzzle about dividing the stakes of an interrupted game fairly, and the letters Pascal exchanged with Pierre de Fermat that summer laid the foundations of probability theory. Long before anyone could calculate the odds, structured games gave people a place to think about risk: fixed rules, unknowable outcomes, and a running negotiation between what a player controls and what no one does. Modern play still sits on that line.
Psychology has added a caution about the human half of the arrangement. In 1985, Thomas Gilovich, Robert Vallone and Amos Tversky examined a belief common among basketball players and fans, that a shooter on a streak is more likely to hit the next one, and found that the shooting records of the Philadelphia 76ers showed no such correlation. The conviction, they argued, comes from a general misconception of chance, our expectation that even short random sequences will resemble the process that produced them, so that ordinary clusters read as meaningful runs. The finding transfers directly to the card table, where a run of strong hands feels like information about the hand to come. Knowing the statistics does not turn that feeling off. It changes what the feeling is allowed to decide.
When classical card games were rebuilt as software, their designers inherited Ulam’s problem. He reached for repeated random trials because the probability was beyond computing; a digital card game starts from settled probability and has to manufacture the randomness. The dealing in a game of online blackjack is driven by random number algorithms, audited against batteries of statistical tests, because once there is no physical deck the entire game rests on the quality of that manufactured chance. It is the convalescent’s experiment running the other way around, at industrial scale: Ulam dealt a hundred games to approximate a probability he could not reach, while the software holds the probabilities fixed and deals the games.
Whether the arithmetic deserves belief is itself a statistical question, and it has a public answer. The National Institute of Standards and Technology publishes a suite of fifteen statistical tests for random number generators, probing their output for every family of pattern a sequence might betray. The verdict such testing can deliver is a modest one: it speaks to long-run behavior and stays silent about the next draw, because randomness can only be tested across many outcomes, never observed in one. Transparency here takes a particular form: published rules, and published evidence of how the mechanism was judged.
The computer scientist Judea Pearl, interviewed about The Book of Why, has spent a career on the neighboring question of what evidence can and cannot establish, and why correlation is so easily mistaken for causation by machines and people alike. The card table teaches a species of the same rigor: sorting what the evidence supports from what it merely suggests. Ulam never presented his solitaire question as profound. He wanted to know whether a game would come out, conceded that cleverness had failed him, and counted. Most of what has been built on his method since amounts to faster ways of doing exactly that, which leaves the patience, as ever, to be supplied by the person doing the asking.