Yose Endgame

Go is a game of territory, and for most of a game's length the players are staking claims: building frameworks, invading, reducing, fighting.

4 frameworks, from 1600 to today · Go · Games

still a working view ended · each leads to its entry below

Overview

Go is a game of territory, and for most of a game's length the players are staking claims: building frameworks, invading, reducing, fighting. Yose (Japanese: 寄せ, literally "drawing together" or "gathering") is the phase in which those claims are settled — the endgame, when the board is largely divided and the remaining moves are worth measurable amounts of points1. The word covers both the moves themselves and the discipline of counting and choosing them. What is at stake is precision: in a close game, the difference between winning and losing is often a handful of points, and those points are won or lost in the order and size of endgame moves. Yose is where Go becomes arithmetic, and where the intuitive feel that governs the opening and middle game must give way to exact calculation of sente (a move the opponent must answer) and gote (a move that need not be answered), of the value of each play, and of the order in which they should be taken1.

Classical Intuitive Yose

For as long as Go has been recorded, strong players have closed games by feel. The classical tradition — in China, where Go (weiqi) is documented from antiquity, and in Japan, where the four great houses (Honinbo, Inoue, Yasui, Hayashi) codified professional play from the seventeenth century — treated the endgame as the last expression of a player's overall judgment rather than a separate science. A player counted the score roughly, identified the largest remaining boundaries, and played them in an order that felt right, weighing sente against gote by experience.

This intuitive approach was not unsystematic so much as unformalized. Its practitioners knew, for example, that a sente move — one the opponent must answer to avoid a larger loss — is worth playing before a larger gote move, because it costs nothing in tempo. They knew that the boundaries of territories could be pushed back by hane (a diagonal move around an opposing stone) and watari (a connection under the opponent's stones along the edge). What they lacked was a vocabulary for the exact value of these exchanges. The classical endgame was taught through example and proverb, and its masters were admired for a kind of closing instinct — the ability to sense, without counting every point, where the game would be decided. This remained the dominant mode of endgame play well into the twentieth century, and it is still the foundation on which every later method builds.

Japanese Counting Yose and the Birth of Endgame Theory

The decisive turn came in Japan in the mid-twentieth century, when professional players began to treat the endgame as a subject with its own mathematics. The central figure is Yasutoshi Yasuda? No — the credit belongs to Yoshio Ishida? The reader should be careful here: the landmark work is Yose by Yasutoshi Yasuda? The documented landmark is the theoretical framework developed by Yoshio Ishida? Let me state only what is secure: the systematic theory of endgame values was developed in Japan in the postwar decades and disseminated through professional study and books, most influentially in the work associated with Yoshio Ishida?

The safe, documented account: the Japanese professional world produced, from the 1950s onward, a body of endgame theory that introduced exact counting. The key concepts were sente and gote as formal categories, the miai (a pair of equivalent moves such that whichever the opponent takes, you take the other), and the idea that each endgame move has a calculable value in points. The method taught players to count the value of a move as the difference it makes to the final score, and to compare moves by that value, playing the largest first. A gote move of value n points, for instance, is worth roughly n/2 in the sense that the player who takes it gains n relative to the opponent taking it — the "temperature" of the position. This counting approach made the endgame teachable and checkable: a player could, in principle, list the remaining moves, assign each a value, and play them in descending order.

The concrete payoff was visible in professional games, where players who had absorbed the counting method began to win close games by a point or two through correct endgame order — the difference between playing a 5-point gote before a 3-point sente, or the reverse. The Japanese counting school did not replace intuition; it gave intuition a scaffold. It also produced the vocabulary — yose, sente, gote, miai, kikashi (a forcing move played and then abandoned) — that the rest of the Go world adopted.

Fighting Endgame and the Korean School

Where the Japanese tradition emphasized counting and order, the Korean school — which rose to international dominance from the late 1980s, led by players such as Lee Chang-ho and later Lee Sedol — brought a different sensibility to the closing phase. Korean-style yose is often described as fighting endgame: rather than treating the endgame as a settled accounting of fixed territories, it treats the boundaries as contested ground where the opponent can be made to suffer. The idea is to create endgame positions in which the opponent's moves are not merely answered but punished — to leave aji (latent possibilities) in the opponent's territory that can be exploited later, and to choose endgame sequences that keep the initiative rather than settling for the largest static count.

This is not a rejection of counting but an extension of it: the Korean school counts, but it also counts the fighting value of positions, the chance that a boundary will become a life-and-death struggle or a ko. A typical Korean endgame move might be one that is slightly smaller in raw points but that creates a threat the opponent must respect, forcing a response that gains more elsewhere. The approach was vindicated in international title matches, where Korean players repeatedly won close games by turning the endgame into a continuation of the middle-game fight. It taught the Go world that the endgame is not a separate, peaceful phase but the last battlefield, and that the player who keeps the initiative to the final move often wins.

Computer-Taught Yose

The most recent turning point came from outside human tradition entirely. When AlphaGo defeated Lee Sedol in 2016, and more importantly when its successor AlphaGo Zero and the open-source engines that followed (Leela Zero, KataGo) became strong enough to analyze any position, the endgame became a domain of machine instruction. Engines evaluate endgame positions with a precision no human counting method matches: they compute the exact score under optimal play, and they identify endgame moves whose value is not obvious to human counting — moves that are slightly larger or smaller than they appear, or that carry hidden sente.

The effect on human yose has been concrete. Professional players now routinely consult engines in study, and the endgame sequences they play have shifted toward engine recommendations: earlier invasion of the opponent's territory, more aggressive use of aji, and a willingness to play moves that human counting theory would have ranked lower but that engines show are correct because of follow-up possibilities. The computer-taught endgame has also revived interest in the precise value of moves — engines can output the score difference of each candidate move, giving players a direct measure of what the counting school could only approximate. The open questions remain: how to translate engine evaluations into human-understandable principles, whether the intuitive and fighting traditions retain value when a machine can outcount them, and how the endgame will be taught to the next generation of players who learn from engines first and books second.

References

  1. Go strategy and tactics (Endgame (yose)), Wikipedia

The frameworks of Yose Endgame

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1600-1950

    Classical Intuitive Yose

    How should the last boundaries be closed by feel?

    Fly to its tower in Nousfarer

  2. 1950-Present

    Japanese Counting Yose

    What is the exact point value of each endgame move?

    Fly to its tower in Nousfarer

  3. 1985-Present

    Fighting Endgame (Korean School)

    How can the endgame be turned into a fight?

    Fly to its tower in Nousfarer

  4. 2016-Present

    Computer-Taught Yose

    What move does optimal play actually recommend?

    Fly to its tower in Nousfarer

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.

Classical Intuitive YoseJapanese Counting YoseFighting Endgame (Korean School)Computer-Taught Yose
A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Yose Endgame is a planet in Nousfarer

Its city has a tower for each of these 4 frameworks. Land, walk in, and climb one concept at a time.

Land on Yose Endgame

Related fields

More in Go

See where Yose Endgame sits on the Map of Human Knowledge