Loading field map
Endgame theory in xiangqi has always faced a fundamental challenge: how to compress the enormous variety of simplified positions into knowledge that a human can learn and apply reliably. The board's unique features—the river dividing the two sides, the palace restricting the generals and advisors, and the cannon's need for a screen to attack—create endgame patterns that differ sharply from Western chess. Over five centuries, three distinct methodological schools have each offered a different answer to that challenge, building on and reacting against one another.
The earliest systematic approach to xiangqi endgames was a tradition of mnemonic pattern cataloging. Classical masters compiled manuals filled with named positions—"Single Horse vs. Single Pawn", "Cannon and Advisor vs. Two Pawns"—each accompanied by a sequence of moves to memorize. The goal was to equip a player with a mental library of standard wins, draws, and defensive setups. This school treated endgame knowledge as a collection of concrete recipes rather than a set of general principles.
Classical manuals were often organized by piece type, but they offered little explanation of why a particular sequence worked. A player who encountered a position not in the manual had to rely on trial and error or on the advice of a teacher. The school's strength was its practicality: a diligent student could learn dozens of common endgames by heart. Its weakness was that the catalog could never be exhaustive, and the reasoning behind the recipes remained opaque. By the early twentieth century, tournament play had grown more intense, and players began to demand a more systematic understanding that could handle unfamiliar positions.
The Modern Analytical School emerged in the mid-twentieth century as a direct response to the Classical School's limitations. Instead of memorizing isolated patterns, modern analysts began to classify endgames by typological categories—for example, "cannon endgames", "horse endgames", "pawn endgames", and mixed-piece endgames—and to derive general conversion principles that applied across many positions. Key principles included king activity, pawn promotion threats, piece coordination, and the importance of controlling the palace entrance.
This school did not discard the Classical patterns; it absorbed them by reclassifying them under its typological headings and explaining why the classical sequences worked. A Classical recipe for "Cannon and Advisor vs. Two Pawns" was now understood as a special case of the broader principle that a cannon needs a screen and that the advisor can serve that role while also protecting the king. The Modern Analytical School thus preserved the Classical lore while giving it a theoretical framework.
By the 1970s and 1980s, this approach had become the dominant way to teach and study endgames. Books and articles organized endgames by piece composition and presented general rules alongside illustrative examples. The school's main limitation was that its principles, while powerful, could not always handle complex multi-piece endgames where the interaction of several pieces created too many variables for human reasoning. Players still had to rely on memorization for the most intricate positions, and the boundary between principle and recipe remained fuzzy.
The arrival of powerful computers in the 1990s brought a third approach: exhaustive computation. Instead of cataloging patterns or deriving principles, the Computer-Assisted Endgame Analysis School uses tablebases—databases that store the game-theoretic outcome of every possible position with a given set of pieces. For a position with, say, three pieces per side, a tablebase can instantly report whether it is a win, loss, or draw with perfect play, and can provide the shortest winning sequence.
This school has transformed endgame study in xiangqi, though progress has been slower than in Western chess because of the larger board and the unique movement rules of the cannon and the palace. By the 2020s, tablebases for positions with up to six or seven pieces had become available, and they continue to expand. The computer does not need to understand why a position is winning; it simply calculates all possibilities.
The Computer-Assisted School has not replaced the Modern Analytical School; instead, the two coexist in a symbiotic tension. They agree that systematic classification is valuable—tablebases are themselves organized by piece count and type. But they disagree on the role of human understanding. Modern analysts argue that principles and heuristics are essential for practical play, because a human cannot consult a tablebase during a game and must rely on general knowledge. Computer-assisted analysts counter that for positions within the tablebase, direct memorization of the perfect line is superior to any principle, and that principles can even be misleading if they conflict with the tablebase result.
Today, a serious endgame student draws on all three schools. The Classical patterns remain pedagogically useful for beginners, who learn the basic wins and draws by rote. The Modern Analytical School provides the conceptual vocabulary—king activity, piece coordination, promotion threats—that allows a player to reason about unfamiliar positions. The Computer-Assisted School offers the final arbiter: when a position is within the tablebase, its verdict is definitive, and it has corrected many long-held beliefs from the earlier schools.
The leading frameworks today—the Modern Analytical School and the Computer-Assisted Endgame Analysis School—agree that endgame knowledge should be systematic and verifiable. They disagree on whether the ultimate goal is to internalize principles that generalize beyond the tablebase or to memorize the tablebase lines themselves for the positions that matter most. This disagreement is not a weakness; it drives ongoing research. Modern analysts refine their principles by testing them against tablebases, while computer-assisted analysts push the frontier of solved positions, knowing that each new tablebase will challenge some cherished rule of thumb. The Classical School, though no longer a living research tradition, survives as the foundation on which the later schools built.