Backgammon
Backgammon is a two-player race game played on a board of twenty-four narrow triangles, called points, with fifteen checkers per player and a pair of dice.
5 frameworks, from 800 to today · Tabletop Games · Games
Overview
Backgammon is a two-player race game played on a board of twenty-four narrow triangles, called points, with fifteen checkers per player and a pair of dice1. Each player moves checkers in opposite directions around the board toward their own home quadrant and then bears them off; the first to remove all fifteen wins1. What makes the game worth a history is the tension at its center: it is a game of luck, since every move is dictated by a dice roll, yet it is also a game of skill, because the player chooses among the legal uses of each roll, and over many games the better chooser wins1. Almost everything that has happened in the game's long history—new rule sets, new ways of scoring, new theories of how to move, new machines that play better than any human—has been an attempt to manage, exploit, or master that tension.
The ancient race games and the emergence of nard
Backgammon belongs to a family of "tables" games—race games in which pieces travel around a track and are removed at the end—that is older than backgammon itself12. Board games of this shape appear in the archaeological record of the ancient Near East and Mediterranean: a game played on a grid of twenty squares, often called the Royal Game of Ur, is attested in Mesopotamia from around 2600 BCE, and a game on a similar track was played in Egypt. These are not backgammon, and their rules are only partly recoverable, but they establish the deep ancestry of the race-game idea.
The decisive step toward backgammon was the introduction of dice-driven movement around a track of twenty-four points with the possibility of hitting and re-entering pieces. A game called nard is documented in the Persian and Islamic world by the early medieval period; Persian tradition associates it with the Sasanian court, and it spread with Islamic culture across the Middle East, North Africa, and into Europe. Nard already had the essential backgammon structure: two sets of fifteen pieces, movement in opposite directions, and the use of two dice. Its exact rules varied, and different regions played different versions, but the family of games it founded—known in Europe as "tables"—was the direct ancestor of the modern game.
European tables and the long standardization
In medieval and early modern Europe, tables games were enormously popular and remarkably diverse1. Players used the same board for several games with different starting arrangements and different rules for hitting, re-entering, and bearing off. The most widespread was a game the English called "tables" and the French and Spanish played under related names; a version in which all pieces began off the board and entered on the roll of the dice—the ancestor of the modern opening—became dominant. The game appears in medieval manuscripts, in Chaucer (who mentions "tables" among the games of the tavern), and in the household accounts of nobles and merchants. It was a gambling game above all, played for stakes, and its literature was largely practical: how to play, how to bet, how to cheat.
The rules were not uniform. Different towns and countries used different conventions for the doubling of stakes, the treatment of "blots" (a single checker exposed to being hit), and the direction of play. Standardization came slowly, through the printing of rule books and the migration of players. By the seventeenth and eighteenth centuries a recognizable modern game—played on a board with the home and outer tables, with hitting and re-entry, and with bearing off—was established in western Europe, though regional variations persisted. The game's name in English, "backgammon," is recorded from the seventeenth century, and its distinctive feature, the "back game" of holding checkers deep in the opponent's home board, was already a recognized strategy.
Hoylean codification and the rule-book tradition
The great turning point in the game's public history was the codification of rules in printed manuals. Edmond Hoyle, an English writer on card and board games, published treatises on whist and other games in the 1740s; his name became attached to a whole genre of "Hoyle's Games" compendia that standardized the rules of backgammon for English-speaking players. These books did two things. First, they fixed the rules: the direction of play, the use of the doubling cube's ancestor (the practice of doubling stakes), the rules for bearing off, and the handling of ties and gammons. Second, they began to teach strategy in a systematic way, offering advice on opening moves, on when to take risks, and on the management of the "back game."
The Hoylean tradition was not a single book but a publishing phenomenon: successive editions, expanded and corrected, carried the game's rules to a wide readership and made backgammon a game with a written standard. It also reflected the game's social character—played in coffee houses, clubs, and private homes, for money and for amusement—and its literature remained practical rather than theoretical. The codification made the game portable and teachable, and it set the stage for the modern era in which the game would be analyzed mathematically.
The doubling cube and the birth of match play
The single most important rule innovation in the modern history of backgammon was the doubling cube, introduced in the 1920s in New York1. The cube is a die marked with the powers of two (2, 4, 8, 16, 32, 64) that a player may turn to double the stakes before rolling; the opponent may accept the new stake or decline and forfeit the current one1. The cube transformed backgammon from a game of pure position into a game of position plus risk management: a player must decide not only how to move but whether the position is strong enough to double, and the opponent must decide whether to take or pass. It made match play possible, because a match could be played to a fixed number of points rather than for cash, and it created a new layer of strategy—the "cube decision"—that had no counterpart in the older game.
The cube spread rapidly through the clubs of New York, London, and the Mediterranean, and it became the defining feature of modern competitive backgammon. Its introduction also changed the game's literature: strategy books now had to teach doubling theory, and the concept of "equity"—the expected value of a position in points—became central. The cube is the reason backgammon is today a game of both position and probability, and it is the single rule that most distinguishes the modern game from its ancestors.
The professional gambling era
From the 1920s through the 1970s, backgammon was above all a gambling game, and its culture was organized around money play. The game's popularity surged in the 1960s and 1970s, when it became a fashionable pastime in New York, London, and the international jet set; clubs such as the Mayfair in London and the Cavendish in New York were centers of high-stakes play. Professional players made their living from the game, and the era produced a body of literature on money management, psychological play, and the exploitation of weaker opponents. The game's image as a glamorous gambling pursuit was reinforced by its association with celebrities and by the growth of tournament play, which offered prizes and prestige alongside cash games1.
This era also saw the first serious attempts to analyze the game mathematically. The publication of strategy books that used probability and expected value, and the growth of tournament circuits, began to shift the game's center of gravity from pure gambling toward competitive skill. The professional gambling era was the crucible in which modern backgammon's culture—its clubs, its tournaments, its literature, and its self-image as a game of skill disguised as a game of chance—was formed.
The computer revolution and neural-network analysis
The deepest change in the game's history came with the computer. In the 1970s and 1980s, programmers began to build backgammon-playing programs, and the game proved unusually amenable to machine analysis because its dice rolls make it a game of probability rather than perfect information. Early programs used hand-crafted evaluation functions—rules of thumb about the value of a position—and they played at a respectable but not expert level. The breakthrough came with the application of neural networks, computational systems that learn patterns from large numbers of examples rather than being programmed with explicit rules.
In the 1990s, a program called TD-Gammon, developed by Gerald Tesauro at IBM, learned to play backgammon by playing against itself and adjusting the weights of a neural network through a method called temporal-difference learning13. TD-Gammon reached a level of play comparable to the best human players, and its analysis of positions revealed strategies that human experts had not fully understood—particularly about the value of aggressive play and the timing of doubling. The program's success was a landmark not only for backgammon but for machine learning generally, and it changed how the game was studied: players began to use computer analysis to check their moves, and the concept of "the computer's evaluation" became part of the game's vocabulary.
Later programs, including commercial and open-source engines, refined the approach, and today neural-network engines are the standard tool of serious analysis. They have made the game's theory more precise, exposed errors in traditional wisdom, and created a new kind of player—one who studies the machine's evaluations as earlier generations studied books. The engines have not solved the game (backgammon remains unsolved, and the dice ensure that no perfect strategy exists), but they have transformed how it is understood and taught.
The present
Backgammon today is played at every level, from casual games in cafés to international tournaments with substantial prizes. The doubling cube is universal in competitive play, and match play is organized around equity and cube decisions that the professional era first formalized. The game's literature is now dominated by computer-assisted analysis, and neural-network engines are used by players of all strengths to study positions and openings. The old gambling culture persists in money games and in the game's popular image, while the tournament circuit and online play have made the game more accessible and more competitive than ever. The open questions are the ones the game has always posed: how to balance risk and position, how to value a double, and how to play the probabilities that the dice deal—questions that the machines have clarified but not closed.
References
- Backgammon (introduction, Setup, Bearing off, Tables games, Western Europe, Modern backgammon, Doubling cube, Software, Studies and analysis), Wikipedia
- Tables game, Wikipedia
- TD-Gammon, Wikipedia
The frameworks of Backgammon
Each framework is a way of working the field has used: when it held, and the question it answers.
- 800-1600
Nard and the Tables Family
How did the tables family of dice race games work?
- 1740-1900
Hoylean Rule-Book Codification
What are the standard rules of backgammon?
- 1920-Present
- 1920-1980
Professional Gambling Money Play
How does a professional make a living from the game?
- 1992-Present
Neural-Network Engine Analysis
Can a machine learn to play better than humans?
How the ideas connect
Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.
Hoylean Rule-Book Codification grew out of Nard and the Tables Family
The Hoylean rule books codified the European tables game that had descended from nard, fixing the rules of backgammon and its doubling of stakes for English-speaking players. The printed manuals standardized a game whose structure and conventions had been carried forward from the medieval tables family.
Carried by Edmond Hoyle.
Doubling-Cube Match Play influenced by Hoylean Rule-Book Codification
The Hoylean manuals had already codified the practice of doubling stakes, which the New York players of the 1920s formalized into the doubling cube. The written standard made the older stake-doubling convention portable enough to be turned into a fixed match-play device.
Professional Gambling Money Play influenced by Doubling-Cube Match Play
The doubling cube, introduced in 1920s New York, gave money players a way to manage risk and made high-stakes professional gambling the organizing culture of the game through the 1970s. Cube decisions became part of the professional gambler's stock in trade.
Neural-Network Engine Analysis influenced by Professional Gambling Money Play
The professional gambling era produced a body of practical strategy literature on position, doubling and equity that later served as the benchmark against which neural-network engines were measured. TD-Gammon's analysis of aggressive play and cube timing was framed in the vocabulary of expert money play.
Carried by Gerald Tesauro.
Beyond the field
Where Backgammon's ideas came from in other fields, and what they led to there.
Nard and the Tables Family led to Hunt and Race Games in Abstract Strategy Games
The hunt and race games of medieval Europe share descent with the nard and tables family, whose dice-and-track mechanics spread from Persia and the Islamic world into European tables games.
Neural-Network Engine Analysis influenced Computer shogi in Shogi
Neural-network evaluation methods demonstrated in backgammon (notably TD-Gammon's self-play learning) influenced the machine-learning approach later applied to shogi engines, where neural evaluation functions were trained on game records. The technique of learned position evaluation passed from backgammon computing to shogi computing.
Carried by TD-Gammon.
Neural-Network Engine Analysis influenced Neural-network policy and value learning in Computer Go
TD-Gammon showed that a neural network could be trained by self-play with value estimates to play backgammon at strong level, providing the template for neural value and policy learning in games that AlphaGo's makers built on.
Carried by Gerald Tesauro.
Neural-Network Engine Analysis influenced AlphaGo in Life And Death
AlphaGo's use of a neural network to evaluate positions descends from TD-Gammon, which learned backgammon evaluation by neural network and self-play. The AlphaGo team cited this line of neural game evaluation as background.
Carried by Gerald Tesauro.
Neural-Network Engine Analysis influenced Engine-assisted play and analysis in Abstract Strategy Games
Neural-network engine analysis pioneered in backgammon by TD-Gammon's self-play learning informed later neural engines such as AlphaZero, which applied similar reinforcement learning to chess and shogi.
Carried by Gerald Tesauro; TD-Gammon.
Hoylean Rule-Book Codification influenced Gambling Lore and Oral Tradition in Poker Strategy
Hoyle's rule books and their American successors codified the rules and odds of card and table games, and poker lore drew on this printed authority for hand rankings and procedure. The oral tradition transmitted and embellished material that the rule-book tradition had fixed in print.
Carried by Edmond Hoyle.
Neural-Network Engine Analysis influenced Open-source engine development in Computer Chess
Neural-network engine analysis in backgammon, notably TD-Gammon, demonstrated learned evaluation and inspired later open-source chess engines to experiment with learned evaluation functions.
Carried by Gerald Tesauro.
Neural-Network Engine Analysis influenced Neural self-play in Chess
Neural-network engine analysis, as developed in backgammon programs such as TD-Gammon, showed that a trained network could evaluate positions without handcrafted features. This demonstrated the viability of learned evaluation that neural self-play later adopted for chess and Go.
Carried by Gerald Tesauro.

Backgammon is a planet in Nousfarer
Its city has a tower for each of these 5 frameworks. Land, walk in, and climb one concept at a time.
Land on BackgammonRelated fields
Fields its ideas are linked to
- Abstract Strategy Games 2 links
- Poker Strategy 1 link
- Chess 1 link
- Computer Chess 1 link
- Computer Go 1 link
- Life And Death 1 link
- Shogi 1 link