Loading field map
Endgame theory in xiangqi (Chinese chess) is the systematic study of positions with few pieces remaining on the board, where the objective is to determine correct play—whether a position is a win, loss, or draw for the side to move, and how to convert or hold it. Unlike the opening or middlegame, where principles are probabilistic and strategic, the endgame is the domain of exactness: many positions can be solved completely, and the value of a material advantage often depends on precise piece placement rather than mere count. The subfield thus sits at the intersection of practical technique, combinatorial analysis, and, in the modern era, computational verification.
The foundational question of xiangqi endgame theory is simple to state but often hard to answer: given a position with few pieces, what is the outcome with best play? From this flow several subordinate questions. Which material configurations are winning, and which are only drawing? How does the position of the kings (called generals in xiangqi) alter the evaluation? What are the shortest or most reliable winning procedures? And when a position is drawn, what defensive resources must the weaker side exploit?
The stakes are practical and theoretical. For the player, endgame knowledge converts half-points into full points and saves lost positions. For the theorist, the endgame offers a rare arena in xiangqi where complete knowledge is attainable. Because the board is small (9 by 10 lines) and the piece movement rules are fixed, a position with, say, two or three pieces per side can be exhaustively analyzed. This makes endgame theory the most rigorous branch of xiangqi knowledge, closer in spirit to a solved puzzle than to the open-ended complexity of the middlegame.
A distinctive feature of xiangqi endgames is the role of the generals. Each general is confined to a 3-by-3 palace, and the two generals may not face each other along an open file—the "flying general" rule. This rule is not a mere formality; it is a powerful attacking and defensive resource. A general can "check" the opposing general by moving onto the same file with no pieces between them, and the defender must interpose or move off the file. Endgame theory therefore treats the generals not as passive kings to be sheltered, but as active pieces whose positions and mutual visibility often decide the outcome. Many theoretical positions hinge on whether the attacker can use the flying general rule to gain a tempo or force a decisive interposition.
Endgame knowledge in xiangqi is old, but its systematic study is comparatively recent. Traditional Chinese chess manuals, such as the Ming-dynasty Jipu (collected games) and the Qing-dynasty Xiangqi Shixian (xiangqi masterpieces), contained endgame positions, often as composed problems or as instructive examples from play. These were typically presented as "residual games" (canju), a term that already suggests a distinct body of positions worth studying apart from full games. The authors of these manuals were concerned with practical correctness: they showed a winning method or a drawing resource, often with commentary on why alternative moves fail.
However, these early treatments were not theoretical in the modern sense. They did not attempt to classify all positions of a given material balance, nor did they distinguish rigorously between positions that are won, lost, or drawn under all circumstances. A manual might show a winning line against a particular defense, but it would not necessarily prove that the defense was the best, or that no other defense could hold. The emphasis was on transmitting technique, not on establishing exhaustive truth.
The modern subfield emerged in the twentieth century, driven by two forces. The first was the growth of organized xiangqi competition, especially after the mid-century, which created a demand for reliable endgame knowledge. Players needed to know, for example, whether a chariot (rook) against two advisors was a win or a draw, and how to play it. The second force was the influence of Western chess endgame theory, which had long distinguished between theoretical positions (known results) and practical play. Chinese authors began to adopt this distinction, producing systematic treatises that classified endgames by material and by result.
A landmark in this development was the work of the Chinese player and theorist Tu Jingming, whose Xiangqi Endgame Studies (published in the 1950s) attempted a comprehensive classification. Tu's work, and that of his contemporaries, established the basic vocabulary of the field: terms like "winning position" (shengju), "drawing position" (heju), and "lost position" (baiju), as well as the distinction between "theoretical" and "practical" endgames. These authors also introduced the idea of the "endgame library"—a body of known positions that a serious player should memorize, much as a Western chess player learns king-and-pawn endgames.
The second half of the twentieth century saw the field mature through the accumulation of composed studies. Composers created artificial positions that isolate a single idea—a subtle waiting move, a sacrifice to gain a tempo, a deflection of the defending general—and these studies became the laboratory of endgame theory. They tested the limits of known principles and sometimes overturned them. A composed study might show that a material configuration previously thought drawn is actually won, or that a "standard" winning method fails against a clever defense. This interplay between composition and analysis remains central to the field.
Endgame theory in xiangqi is not divided into rival schools in the way that, say, economic theory or literary criticism is. There is no "formalist school" opposed to a "materialist school." Instead, the field is organized by method and by object of study, and these approaches coexist and reinforce one another.
The oldest and most basic approach is classification by material. The theorist takes a fixed set of pieces—say, one chariot against one chariot and two advisors—and asks: what is the result? The answer is not a single value but a set of values depending on position. The task is to map the space of positions into winning, drawing, and losing regions, and to describe the boundary between them.
This approach is inherently taxonomic. It produces tables and rules: "chariot against chariot and two advisors is a draw if the defending advisors are on the same side of the palace, but a loss if they are split." Such rules are the bread and butter of practical endgame play. The classificatory approach has a natural limit, however. The number of material configurations is large, and the position space within each configuration is vast. Complete classification is feasible only for the simplest endgames—those with two or three pieces per side. For richer endgames, the best that can be achieved is a set of heuristics and typical positions, not an exhaustive map.
The compositive approach works from the opposite direction. Instead of starting with a material configuration and asking for its general result, the composer starts with an idea—a tactical motif, a zugzwang, a stalemate resource—and constructs a position that embodies it. The result is a study: a position, usually with a stipulation such as "red to move and win," that is solved by a unique or narrow sequence of moves.
Composed studies serve two functions in endgame theory. First, they are a testing ground for principles. A principle such as "the chariot should be placed behind the pawn" is not a theorem; it is a generalization from many positions. A composed study can confirm it, refine it, or show its limits. Second, studies are a pedagogical tool. They train the player to see hidden resources and to calculate precisely, which is the core skill of endgame play. The compositive approach is not opposed to the classificatory one; rather, it feeds into it. A study that reveals a new winning method may lead to a reclassification of a material configuration.
The most recent approach is computational. Since the late twentieth century, researchers have used retrograde analysis to solve xiangqi endgames exhaustively. The method is the same as that used for Western chess tablebases: starting from all possible terminal positions (checkmate, stalemate, or a draw by rule), the computer works backward, assigning a value (win, loss, draw) to every position with a given material balance, along with the distance to the result.
For xiangqi, this is complicated by the flying general rule and by the fact that the game has no stalemate—a player with no legal move loses, but this is rare. The first complete tablebases for xiangqi endgames appeared in the early 2000s, covering configurations with up to a handful of pieces. These tablebases have confirmed some traditional analyses and overturned others. For example, some endgames that classical theory held to be draws have been shown to be wins, but only with precise play that may require dozens of moves. The computational approach has also revealed the existence of "cursed wins"—positions that are theoretically winning but require so many moves that the win is impractical under tournament time controls.
The computational approach differs from the earlier ones in a fundamental way. It does not produce human-understandable rules; it produces a database of facts. A tablebase can tell you that a position is a win in 47 moves, but it cannot tell you why, in terms a human can grasp. The challenge of converting tablebase knowledge into comprehensible theory is an open problem in the field. Some researchers have attempted to extract "patterns" from tablebases—recurring motifs that explain the result—but this work is still in its infancy.
These three approaches are not rivals but layers of a single enterprise. The classificatory approach provides the map of the territory; the compositive approach provides the detailed case studies that refine the map; the computational approach provides the ground truth against which both are checked. In practice, a modern endgame theorist might use all three: consult a tablebase to verify a position, study a composed problem to understand a motif, and then formulate a classificatory rule that captures the general principle.
There are tensions, however. The classificatory approach aims for simple rules, but tablebases often reveal that the truth is complex—that a position is won only by a single move, or that the boundary between win and draw is a razor's edge. The compositive approach values aesthetic economy, but a tablebase win may require ugly, repetitive maneuvering. These tensions are productive. They force the theorist to ask what "understanding" means in the endgame: is it the ability to state a rule, or the ability to play correctly? The field has not resolved this question, and it is unlikely to.
The current state of xiangqi endgame theory is characterized by a division of labor. For simple endgames—those with up to about five pieces per side—tablebases provide definitive answers, and the theoretical results are largely settled. The classical rules have been confirmed or corrected, and the remaining work is pedagogical: how to present these results so that players can use them.
For richer endgames, the field remains open. No tablebase can cover positions with many pieces, and the classificatory approach reaches its limits. Here, theory shades into practice. The "endgame" in a real game is rarely a pure theoretical position; it is a messy configuration with multiple pieces, where the principles of the simple endgames must be combined and adapted. The durable contribution of endgame theory is not a set of memorized positions but a set of transferable skills: the ability to calculate precisely, to recognize the key features of a position (which pieces are active, which are passive, where the generals stand), and to know which material advantages are worth pursuing.
The field also retains a strong composed-study tradition. New studies continue to be published, and they remain the most reliable way to discover new ideas. The relationship between composition and computation is now symbiotic: composers use tablebases to check their creations, and tablebase researchers use composed studies as a source of interesting positions to analyze.
For the educated newcomer, the practical takeaway is this. Endgame theory in xiangqi is not a closed body of knowledge but a living discipline with a clear hierarchy of certainty. At the base are the solved simple endgames, which every serious player should know. Above that is a layer of heuristics and typical positions, which are reliable but not infallible. At the top is the open frontier, where the player's own calculation and judgment must carry the day. The field's central lesson is that material advantage is not enough; the position of the generals, the activity of the pieces, and the precise order of moves decide the outcome. Mastery of the endgame is mastery of these details.