Backgammon
The board has not changed in fifteen centuries, but a small numbered cube invented in a New York card room turned an ancient race into a game of nerve.

A backgammon board set up to start looks like a quarrel waiting to happen: thirty checkers, fifteen to a side, planted across twenty-four narrow points in a pattern that throws each army straight into the other's path. Two players roll two dice and run their checkers around the horseshoe in opposite directions, from the far corner home to their own inner board, and the first to lift all fifteen clear of the board wins. That much a child can learn in five minutes. What keeps grandmasters at it is the dice, because every roll opens a fan of legal moves, and behind each one waits the question of what the other player might roll back.
Running is only half of it. Land two or more checkers on a point and you hold it, and your opponent cannot touch down there at all. Leave a single checker alone on a point, a blot, and a passing opponent can hit it, knocking it onto the bar that splits the board, from which it must re-enter on the far side and run the whole length again. So the game pulls two ways at once, toward the dash for home and toward the wall that strands the other man behind it, and reading a position means weighing the one against the other. Get all fifteen of your checkers home and you begin bearing them off, lifting them clear; fall behind badly and you can be gammoned, beaten before you bear off a single checker, for double the stakes.
The board itself is ancient, older than almost anything else still played for pleasure. Backgammon belongs to a wide clan known as the tables games, race games played across a two-row board, and its relatives run back some five thousand years into the Near East. The Romans bet on a version they called the game of twelve lines; a Byzantine epigram pins a famously ruinous run of dice on the emperor Zeno around the year 480, in a game already played on twenty-four points with fifteen pieces a side. The English name first surfaces in a letter of 1635, built most likely from back and the old word gamen, meaning game, for the way a struck checker is sent back to begin again. Edmond Hoyle, whose surname still stands for the rulebook itself, brought out his short treatise on the game in 1753.

For all that age, the move that made the modern game is barely a century old. In the card rooms of New York's Lower East Side in the 1920s, someone whose name is now lost added a small six-sided block marked 2, 4, 8, 16, 32, and 64: the doubling cube. At any moment before he rolls, a player who likes his position may offer to double the stakes, and his opponent must either accept the larger game or concede the point on the spot. The cube sounds like a gambler's gimmick and turns out to be the soul of the game. It forces a judgment a pure race never demands, an honest estimate of your own chance of winning from exactly where you sit. A sound rule of thumb says take the double whenever you would still win at least one game in four, and fold below that to save your points. Skill in backgammon became, above all, skill at reading the odds.

A grandmaster walks through when to offer the cube, when to take, and when to fold, the decision that separates the strong player from the merely competent.
The cube also made the game worth taking seriously, and serious players arrived. Backgammon's modern boom came in the mid-1960s around Prince Alexis Obolensky, a Russian émigré who ran the first major international tournament in 1964 and was crowned the Father of Modern Backgammon; within a few years the boards were out on college lawns, in country clubs, and at Hugh Hefner's parties. In 1976 a young mathematician named Paul Magriel published Backgammon, six hundred diagrams of patient, building-block analysis that players still call the Bible. He signed himself X-22, after the anonymous winner of a sixty-four-player bracket he had once played out entirely in his own head, and he took the world title himself two years later.
The last turn in the story is that the machines reached backgammon early, and for once the result outran the game. Its dice and its roughly eighteen quintillion legal positions made it too unruly for the deep lookahead that would later crack chess. In 1992 a researcher at IBM named Gerald Tesauro built TD-Gammon, a neural network that learned by playing itself hundreds of thousands of times and adjusting a little after every game, until it played even with the world's best and handed them opening ideas they had dismissed for years. It was among the first convincing wins for a machine that learns from its own experience rather than from a rulebook, the same principle that two decades later would conquer Go. The oldest game on the table turned out to be a door into the newest one.