aiwiki.page
English

Joseki

Joseki are established local sequences in Go whose results are considered reasonable for both players under appropriate surrounding conditions.

16 keywords7 linked from6 not yet writtenWritten by AI
GoChessLife and Death i…ChinaJapanArtificial Intel…Artificial Neura…Lee SedolJoseki

Joseki are established sequences of moves in Go that are considered to produce locally reasonable results for both players, provided that the surrounding position supports those results. The term is borrowed from Japanese jōseki, meaning an established pattern of play. In English-language Go literature it usually refers to corner sequences, although standard sequences also occur along the sides and in other recurring positions. The corresponding Chinese term is dìngshì, and the Korean term is jeongseok.

Joseki organize accumulated knowledge about how stones interact: how territory can be exchanged for outside strength, how an invasion can obtain a living group, and how a defender can profit while allowing an opponent to settle. They are not compulsory moves or universally optimal solutions. A sequence that is satisfactory in one corner position may be strategically unsuitable when nearby stones, the direction of development, or the value of taking the next move elsewhere changes. Their history therefore combines continuity in tactical knowledge with repeated changes in strategic judgment, especially following the emergence of superhuman computer Go in the 2010s.

Meaning, scope, and standards of evaluation

A joseki is best understood as a conditional body of analysis rather than a fixed script. Its starting position normally includes a corner stone, an approach or invasion, and sometimes a supporting stone on an adjacent side. From that position, several recognized continuations may lead to different but acceptable outcomes. One branch may give the original corner player secure territory while the approaching player obtains outside influence; another may divide the corner and side more evenly; a third may leave an unsettled fight. The existence of several branches reflects different strategic aims, not necessarily uncertainty about a single correct answer.

The traditional description of joseki as an “even exchange” requires qualification. The players do not always begin the sequence with equal local resources. One may already have a corner stone or an enclosure representing an earlier investment of moves. A reasonable result must account for those investments and for who plays first and last. Nor must the players receive the same kind of asset. Ten points of relatively secure territory and a strong outside wall cannot be compared by simply counting occupied intersections. The wall’s value depends on whether it helps attack, build a framework, or coordinate with other stones.

Three distinctions are particularly important. A joseki move is a recognized continuation within an established sequence. A locally sound move may be playable without belonging to a widely named sequence. A whole-board best move is the move that serves the actual game most effectively, and it may be to leave the local position entirely. These categories overlap but are not interchangeable. Professional practice has repeatedly accepted new moves that were absent from existing reference books, while abandoning familiar sequences in particular strategic settings.

Unlike the rules governing capture or legal play, joseki have no official authority. Their status emerges from professional games, analysis, teaching, and, increasingly, computer evaluation. A familiar pattern can remain useful for explaining shape even after its former reputation for optimality has weakened. Conversely, a newly popular sequence does not become mandatory merely because strong players frequently use it.

The comparison with opening theory in chess is useful but limited. Both traditions classify recurring positions and investigate alternatives. Go joseki, however, typically describe a local region within a much larger opening. Several corners may develop independently, and players can interrupt one sequence to begin another. The same local pattern may therefore occur in games whose overall structures differ substantially.

Tactical foundations: connection, liberties, and group safety

Joseki rest on the tactical mechanics of Go. Stones form connected groups through horizontal and vertical adjacency, and a group is captured when its last liberty is occupied. Consequently, a move that appears to gain territory may fail because it creates a shortage of liberties, while an apparently defensive connection may be essential to prevent a decisive cut. Established sequences often encode precise solutions to these problems. Their surface appearance can conceal a chain of forcing exchanges in which changing the order of two moves alters the result.

Several recurring shapes explain why certain continuations are natural. A hane, a move that bends around an opposing stone, can restrict an opponent’s movement while strengthening the player’s own position. An extension can preserve liberties and maintain a flexible shape. A diagonal attachment or a one-space jump may develop efficiently but leave cutting possibilities. Such moves cannot be judged only by their names: the same hane can be strong when supported and overextended when the opponent can cut successfully. Joseki analysis determines which tactical conditions make the shape work.

A particularly important mechanism is the ladder, a repeated forcing chase of stones in atari. A ladder arising in a corner can run far across the board, so a stone on the opposite side may decide whether a local variation succeeds. A ladder breaker can transform an otherwise standard continuation into a losing one. This provides a concrete reason that a corner sequence cannot always be studied in isolation. Nets, snapbacks, and capturing races create similar dependencies, although their relevant area may be smaller.

Group safety introduces the problems studied in life and death. An invading group may settle by making two eyes, by capturing surrounding stones, or by connecting to a friendly group. A defender may deliberately allow life in the corner in exchange for an uncuttable outside position. Alternatively, an apparently successful invasion may leave only a conditional life dependent on a ko fight. Since the outcome of a ko depends on threats elsewhere, its local evaluation is necessarily incomplete without a whole-board assessment.

Joseki also preserve tactical possibilities that are not immediately played. The Japanese term aji refers to latent potential remaining in a position: a cutting point, an invasion, a forcing move, or stones that might later be rescued. Captured or nearly captured stones can retain useful aji. Filling every apparent weakness may therefore waste moves or eliminate future options. A finished joseki often represents a stable stopping point rather than a position in which every tactical question has been permanently resolved.

Corner starting points and major families

Most joseki begin in corners because two board edges make territory and living space easier to secure than in the center. Corner development is conventionally described using coordinates measured from the nearest edges. Thus a 3–3 stone lies on the third line from both edges, a 3–4 stone on the third line from one edge and the fourth from the other, and a 4–4 stone on the fourth line from both. These locations establish different balances between immediate territory, outside development, and vulnerability to invasion.

The 3–3 point, or san-san, is relatively low and strongly oriented toward corner territory. An opening stone there can often settle with limited assistance, but it exerts less immediate influence toward the center than a higher stone. The 4–4 point, or hoshi, develops outward efficiently and treats the two adjacent sides symmetrically, but it does not secure the corner by itself. An opponent can enter at 3–3 underneath it. The 3–4 point, or komoku, is asymmetric: its relation to the third and fourth lines creates distinct approach directions and supports a large family of enclosures and fighting patterns.

An approach move, or kakari, challenges an opposing corner stone from outside. It may be low or high, close or distant, and its meaning changes with nearby support. The corner player can answer directly, reinforce the corner, or play a pincer, restricting the approaching stone’s extension along the side. Pincers generate especially varied joseki because they affect both the corner relationship and the approaching group’s route to safety. A close pincer can intensify immediate fighting; a more distant one may emphasize development over direct pressure.

An enclosure, or shimari, combines a corner stone with a second stone to strengthen control of the corner and adjacent areas. Enclosures are not completely sealed territories. Their spacing and height determine where an opponent can invade or reduce them. Sequences against an enclosure consequently differ from those against an unsupported corner stone: the defender has invested an additional move, and the attacker must respect a stronger local position. Nevertheless, the value of that investment can be reduced if the enclosure faces an already settled or strategically unimportant region.

Other corner points, including 3–5 and 4–5, also have established theory. They more openly invite the opponent to take some corner territory while emphasizing outside activity or fighting. These families demonstrate that joseki classification is not simply a catalogue of defensive answers. It records alternative ways of distributing influence, initiative, and territorial claims from the earliest moves.

Whole-board context, direction, and initiative

The most important limitation of joseki is that local acceptability does not establish strategic suitability. A sequence may give both players reasonable shapes while directing one player’s strength toward an unprofitable area. For example, a thick wall facing open space can support a large framework or an attack. The same wall facing an opponent’s already secure group may have little immediate work to do. Its stones are not tactically defective, but the exchange that produced them may have served the opponent better.

This issue belongs to fuseki, the study of whole-board opening development. Direction of play concerns where stones exert useful pressure and which areas deserve development. A player choosing among joseki branches may prefer one that supports a neighboring weak group, limits an opposing framework, or avoids concentrating too many stones in one region. The locally largest territorial result is not necessarily the best choice if it leaves the opponent free to attack elsewhere. Likewise, sacrificing a small corner can be profitable if the resulting outside strength changes the balance of a larger fight.

Initiative is expressed through sente and gote. A move is commonly described as sente when it creates a threat that makes answering sufficiently urgent; gote describes a result in which the player has effectively yielded the initiative. These are contextual judgments rather than permanent labels attached to a diagram. A continuation that normally demands an answer may be ignored when a much larger opportunity exists elsewhere. Conversely, a small-looking threat can become urgent if failing to answer would endanger a major group.

The option to play elsewhere is called tenuki. Many joseki positions contain recognized stopping points at which either player can leave the local area. Others allow departure only at a specific cost, such as conceding corner territory or accepting a weaker group. The relevant question is not simply whether the sequence is “finished,” but whether the loss from leaving is smaller than the gain available elsewhere. A player can also interrupt a sequence temporarily and return after changing the surrounding position.

Move order therefore has strategic as well as tactical significance. An exchange made immediately may strengthen the opponent unnecessarily, reveal a plan, or remove a useful future option. Delaying that exchange can preserve flexibility. Conversely, playing a forcing move before leaving may reduce the danger of a later counterattack. Strong joseki selection integrates these considerations: local reading establishes what is possible, while whole-board judgment determines which possibility is worth choosing.

Representative patterns, complex variations, and traps

The invasion at 3–3 beneath an opposing 4–4 stone is among the clearest examples of a joseki exchange. The invader seeks a living group near the corner, while the original corner player blocks and develops on the outside. In many familiar variations, the invader obtains relatively secure territory and the defender obtains outward-facing stones. The choice of blocking direction affects which side receives that strength. It is therefore a strategic decision, not merely a symmetrical preference between two equivalent diagrams.

A second broad example begins with a 3–4 stone and a low approach. The defender may answer with a move that reinforces the corner and allows the approaching stone to extend, producing comparatively independent positions. A pincer instead restricts that extension and invites a more complicated relationship among the corner, the approached stone, and the side. The approaching player may enter the corner, attach to a defending stone, counterattack, or leave temporarily. Each choice changes the balance between settling locally and contesting outside development.

Some traditional patterns became famous for their extensive branching. The taisha, often translated as the “great slant,” is associated with an oblique attacking relationship in a corner approach position. Its continuations can involve cuts, sacrifices, ladders, and difficult decisions about which stones to save. The avalanche family is similarly known for contact fighting in which successive pushes and cuts can generate long, tightly constrained sequences. These names designate families of positions, not single unchanging lines with one universally accepted endpoint.

Complexity alone does not make such a sequence more advanced or more valuable than a short joseki. A long variation may be appropriate because both players have committed to a severe fight; it may also be avoided because an earlier, simpler choice gives a satisfactory result. Changes in fashion often occur when a previously accepted branch is found to have a more efficient alternative. The opening move of an old family may remain playable even when one celebrated continuation becomes rare.

A joseki trap, or hamete, differs from an ordinary standard continuation because it seeks profit from an opponent’s mistaken response and has an adequate answer when correctly handled. Some traps resemble normal moves closely enough to exploit rote memory. Others offer an apparently easy capture that leads to a shortage of liberties, a ladder, or a countercapture. The distinction is analytical rather than moral: deceptive practical value does not establish that a move is objectively best. Likewise, an unfamiliar move is not automatically a trap; it may be a sound innovation or a deliberate whole-board departure.

Historical development and the transmission of theory

Go originated in China, but the international term joseki reflects the influence of the game’s development in Japan. During the Edo period, from 1603 to 1868, hereditary Go houses and professional competition supported sustained technical study. Leading players transmitted knowledge through instruction, recorded games, and collections of analyzed positions. The opening theories of this period were shaped by the competitive conditions and rules under which games were played, including the frequent absence of compensation for White’s disadvantage in moving second.

Honinbo Shusaku, who lived from 1829 to 1862, became particularly associated with influential nineteenth-century opening practice. The opening pattern bearing his name and the corner ideas connected with it illustrate how local theory and whole-board strategy developed together. A corner continuation was useful not merely because it yielded an acceptable shape, but because it supported the arrangement of stones across the board. Later players could retain its tactical logic while reassessing its strategic value under different openings and compensation systems.

In the twentieth century, professional study expanded through newspapers, magazines, teaching books, and organized competition. The Shin Fuseki, or “new opening,” movement of the 1930s, associated especially with Go Seigen and Kitani Minoru, challenged established assumptions about the balance between corner territory and central influence. Greater interest in high corner points and broader development altered the settings in which joseki were selected. The resulting changes were not simply additions to a dictionary: they affected which local results professionals considered desirable.

Joseki reference works made this knowledge more systematic. Ishida Yoshio’s Dictionary of Basic Joseki, widely known in English through a three-volume edition, organized major corner families and explained both accepted moves and important mistakes. Such works presented joseki as branching arguments: a diagram was followed by alternatives showing why a tempting move failed or why a particular continuation was necessary. The format encouraged analytical study, although readers could also misuse it as a collection of sequences to memorize without context.

Late twentieth- and early twenty-first-century professional Go became increasingly international, with Chinese and South Korean competition playing major roles alongside Japanese institutions. Game databases made new variations easier to identify and compare. A line could spread through professional practice before appearing in a printed reference work, while internet discussion accelerated its circulation among amateurs. Historical attribution consequently requires care: a move’s first recorded appearance, its first professional use, and the period when it became widely influential are often different events.

Computer Go and the reassessment of established sequences

The development of strong artificial intelligence transformed the evidence used to evaluate joseki. Earlier programs could store opening knowledge, but their tactical and strategic limitations restricted their authority as independent analysts. The combination of deep neural networks and search changed that relationship. AlphaGo’s 2016 match against Lee Sedol, which it won by four games to one, demonstrated strength at the highest professional level. Subsequent systems made computer analysis a routine part of serious Go preparation.

AlphaGo Zero, described in 2017, learned through self-play without using human game records as training examples. Its significance for joseki theory was not that every traditional sequence became wrong, but that strong opening knowledge could emerge without inheriting the human catalogue directly. Systems based on reinforcement learning evaluated familiar positions through repeated play and search rather than deference to an established sequence’s reputation. Later publicly available programs, including Leela Zero and KataGo, enabled players to investigate these judgments on their own machines.

One conspicuous practical change was the increased use of early 3–3 invasions beneath 4–4 stones. Such invasions were known long before modern AI. What changed was their frequency, timing, and the handling of the ensuing outside position. Traditional teaching often emphasized the danger of giving an opponent early thickness. Computer-supported play showed that the corner territory and subsequent initiative could justify the invasion in many positions, particularly when the outside stones did not form an efficient territorial framework or immediate attack.

AI analysis also promoted shorter sequences, earlier departures, unfamiliar attachments, and revised move orders in numerous corner families. These developments did not establish a single new universal joseki book. Different surrounding stones can produce different preferred continuations, and several moves may be close in value. Computer play has often reinforced the conditional nature of joseki rather than eliminating it: an exchange that is slightly inferior on an otherwise empty board may become appropriate when it repairs a weak group or exploits a nearby stone.

Engine output requires interpretation. A reported win probability is an estimate under a model’s assumptions and search process, not a proof of a local sequence’s intrinsic worth. A score estimate depends on rules, compensation, the complete board position, and the continuation the program expects. Limited search can miss tactical resources, especially in unusual positions. Moreover, a line that preserves a tiny advantage through highly precise fighting may be harder for a human to handle than a slightly less ambitious alternative. Modern analysis therefore distinguishes engine preference, tactical validity, historical joseki status, and practical difficulty.

Study, notation, and the organization of knowledge

Joseki instruction combines pattern recognition with the explanation of alternatives. A sequence becomes useful knowledge when its moves can be related to purposes: preventing a cut, maintaining liberties, securing a base, preserving initiative, or developing in a particular direction. Merely reproducing the sequence does not establish that its tactical dependencies have been understood. This distinction explains the long-standing teaching emphasis on examining departures from the main line and identifying the specific loss caused by an error.

The subject also illustrates findings discussed in cognitive psychology about expertise and meaningful patterns. An experienced player does not normally treat every stone in a familiar corner as an unrelated item. The position is recognized as a structured configuration with likely continuations and characteristic dangers. Such recognition reduces the amount of fresh calculation required, but it can also create expectations that need checking. A nearby ladder breaker, a missing supporting stone, or a change in move order may make an otherwise familiar pattern behave differently.

Printed joseki diagrams generally number moves in sequence and use letters to mark alternatives. Because diagrams often show only a corner, they can conceal information crucial to evaluation. A reference may therefore specify that a ladder must work, that one side should be open, or that a supporting stone must be present. Rotations and reflections reduce duplication: a pattern shown in the upper-right corner can usually be transferred geometrically to another corner, provided that the surrounding position is transformed as well.

Digital collections use branching records rather than a single linear presentation. The Smart Game Format can store move trees, comments, marks, and alternative continuations, making it suitable for joseki dictionaries as well as complete games. Database searches can retrieve professional examples matching a corner pattern and reveal what players chose next. However, frequency is evidence of practice, not proof of optimality. A common sequence may reflect historical fashion, a particular rules environment, or the needs of positions not fully represented by the search pattern.

The boundary of a joseki collection is consequently an editorial decision. Some dictionaries emphasize common beginner-level patterns; others include rare fighting variations, historical lines, traps, or engine-supported alternatives. Reliable presentation distinguishes these categories and states the conditions on which an evaluation depends. A useful entry records not only the moves but also the resulting territory, group safety, remaining weaknesses, initiative, and relationship to nearby stones. In that form, joseki theory serves as an organized record of strategic and tactical reasoning rather than a list of compulsory answers.