Chinese Chess (Xiangqi): Rules, Opening Moves, and an Honest Look at the Complexity Claim

Xiangqi (象棋) is not a Chinese variant of chess. It is a different game that happens to use a similar board and the same starting ambition — trap the opponent’s general. About 183 players from 25 countries played it at the 2025 World Championship in Shanghai, and you can learn the whole rule set in one sitting. This guide covers the moves, the five rules beginners always get wrong, an opening you can actually use tonight, what a decent set costs, and — because a lot of articles get this wrong — what the complexity numbers really say.

The board

Nine vertical files and ten horizontal ranks cross at 90 intersections, and pieces sit on the intersections, not inside the squares. The blank band between the two armies is the 河界 (river). At each end, a 3×3 grid of diagonal lines marks the 九宫 — the palace — where the general and his advisors are confined.

Official competition equipment: pieces 3.0–4.0 cm in diameter and 1.0–2.0 cm thick; board squares 4.0–5.0 cm, board 40–50 cm long by 36–45 cm wide.

How every piece moves

• 帅/将 (General). One step orthogonally, inside the palace only.
• 仕/士 (Advisor). One step diagonally, inside the palace only. Five palace points are reachable.
• 相/象 (Elephant). Exactly two points diagonally — and it cannot cross the river. If a piece sits on the intermediate point, the move is blocked (“塞象眼”). It is a defensive piece, permanently.
• 傌/马 (Horse). The knight’s L-shape: one orthogonal step then one diagonal. If a piece occupies that first orthogonal point, the leg is blocked (“蹩马腿”). Most beginner blunders come from here.
• 俥/车 (Chariot). Any distance orthogonally, exactly like a rook. The strongest piece on the board — a chariot is worth roughly two horses.
• 炮/砲 (Cannon). Moves exactly like the chariot, but to capture it must jump over exactly one piece — the 炮架 (screen). No screen, no capture. It is the single biggest tactical difference from western chess.
• 兵/卒 (Soldier). One step forward. After crossing the river it may also step sideways. It never moves backwards, ever.

Five rules beginners miss

• 飞将 / flying general. You may never make a move that leaves the two generals facing each other on an open file. This is not a clever tactic — it is an illegal move, and it is also the finisher: forcing that file open can win outright.
• Stalemate is a loss. In western chess, a player with no legal move and no check is stalemated and draws. In xiangqi, 困毙 — having no legal move — loses the game. So does perpetual check (长将) and perpetual chase (长捉).
• The cannon needs a screen to capture. A cannon alone on a file hits nothing. Beginners push cannons at empty files and wonder why nothing happens.
• Horse legs and elephant eyes. Two blocking rules, both easy to forget, both cost pieces. Before you move a horse, look at the square one step orthogonally toward the target.
• Soldiers never retreat. A soldier that has crossed the river is a permanent asset. Trading it away casually is a slow leak.

Does “more strategic than chess” hold up? Not the way it’s usually claimed

You will read that Chinese chess is deeper, more strategic, more complex than international chess, “and here’s the data.” We went looking for that data. There is essentially one source, and it does not say what people claim it says.

The figures come from Yen, Chen, Yang and Hsu, Computer Chinese Chess, ICGA Journal 27(1):3–18, published March 2004. Their Table 1 gives state-space complexity and game-tree complexity:

• Chess: ~10⁵⁰ positions, ~10¹²³ games
• Xiangqi: ~10⁴⁸ positions, ~10¹⁵⁰ games
• Shogi: ~10⁷¹ positions, ~10²²⁶ games
• Go: ~10¹⁶⁰ positions, ~10⁴⁰⁰ games

Read that carefully. Xiangqi’s state space is about a hundred times smaller than chess’s. Its larger number is the game tree, derived from a branching factor of 38 and an average game length of 95 moves (Hsu, 1990). A bigger game tree with a smaller state space tells you the games run longer and branch wider — it does not establish that the game is “more strategic.” And these are single estimates from roughly two decades ago, using parameters we could not trace back to a primary paper. We found no newer peer-reviewed complexity measurement specific to xiangqi.

Here is the honest version: xiangqi gives you a cannon, an elephant that cannot cross a river, a river that changes what soldiers can do, and a rule that turns stalemate into loss. That produces a different kind of game — more open, more tactical early, more reliant on piece coordination than pawn structure. Whether it is “better” is taste. Anyone who tells you a complexity table settles it is overselling.

An opening you can actually use

The most common first move for Red is 当头炮 (central cannon): cannon to the central file, aiming straight at Black’s central soldier. Black’s standard reply is 屏风马 (screen horses) — developing both horses to protect the centre — and the resulting 中炮对屏风马 family is the backbone of xiangqi opening theory.

• Bring out both horses, then the chariots. A chariot on an open file is worth more than an early attack.
• Don’t move the same piece twice in the opening unless it captures or escapes a real threat.
• Control the two central files. Everything in xiangqi runs through the centre.
• Develop the cannon with a screen in mind. An unsupported cannon on an empty file is decoration.
• Keep your advisor structure intact. Breaking your own 士 formation to chase a pawn is how beginners lose to a single chariot.

How to start playing this week

• Learn the cannon before anything else. It is the piece with no western equivalent. Play ten games where you deliberately look for screens; the rest of the tactics follow from it.
• Play slow games first. Blitz teaches you nothing except how to lose quickly. Twenty minutes a side is enough to actually see a threat.
• Review every loss with an engine. Pikafish is a free, open-source xiangqi engine derived from Stockfish and is the strongest publicly available one; run your games through it and look only at the move where your evaluation collapsed.
• Find opponents online. The main international platform reports hundreds of thousands of active players across nearly a hundred countries, so you will get a game at any hour.
• Then find a club. National championships draw players from two dozen countries and more; local associations exist in most major cities with a Chinese community.

Five dates that explain the game

• c. 3rd century BCE — the phrase 象棋 appears in 《楚辞·招魂》, but it refers to an ivory liubo board game, not the modern game.
• 569 CE — Emperor Wu of Northern Zhou promulgates 《象经》 and convenes officials to discuss it. The earliest substantive textual anchor.
• 762 CE — the 宝应象棋 episode recorded in 牛僧孺’s 《玄怪录》.
• Song dynasty — the board standardises. A 1983 hoard at Jiangyou, Sichuan yielded a set of 32 bronze pieces identical in form to the modern set; 李清照’s 《打马图经序》 shows a board diagram matching today’s.
• 1956 — xiangqi becomes an official national sport with a first national championship (won by Yang Guanlin). The Chinese Xiangqi Association was founded in Hefei on 11 November 1962.

Xiangqi entered China’s second national intangible-heritage batch in 2008, and the English name was standardised to “Xiangqi” in 2009.

Buying a set: materials, sizes, honest prices

Common retail diameters are 25 / 28 / 33 / 38 mm; competition standard is 30–40 mm. Materials run from plastic and acrylic through birch, beech and maple, up to rosewood (红花梨), padauk, 小叶紫檀, 黄花梨 and 金丝楠 at the collector end.

• Plastic folding starter sets: roughly ¥10–35, and around US$27 for a 14-inch wood folding set on US marketplaces.
• Wooden folding sets: wholesale listings cluster at ¥50–198, though many claim “小叶紫檀/黄花梨” at those prices — treat such claims as unverified at that level.
• Genuine mid-to-high wood sets: around ¥300–800.
• Collector grade: ¥3,800–6,800, with pieces up to 58 mm.

If you are buying for a child, official junior competition specs are 30 mm diameter ±0.5 and 12 mm high ±0.5.

The scandal the game is still dealing with

If you follow xiangqi news, you have probably seen headlines about cheating. The accurate version: in 2023 recordings circulated alleging match-fixing and rating manipulation. On 19 September 2024 the Chinese Xiangqi Association issued a lifetime ban on Wang Tianyi and Wang Yuefei and stripped all their technical titles, including Grandmaster. On 12 January 2025 the national chess administration announced penalties covering 41 more people — including lifetime bans for Zhao Xinxin, Wang Yang and Zheng Weitong — for violations spanning 2012 to 2023, bringing the total to 43.

One correction worth making: what the official announcements establish is match-fixing and result manipulation. Allegations of engine use appear in the leaked recordings, but we found no official finding that separately confirms an AI-cheating case. Reporting the two as the same thing is how the story got distorted.

What we could not verify

• Total number of xiangqi players in China or worldwide. “Hundreds of millions” circulates but we found no primary count.
• The number of registered grandmasters (特级大师). Figures of 62, 76 and 237 all circulate; none traces to an association announcement.
• Any quantitative comparison of xiangqi engine strength versus chess engine strength.
• Any complexity measurement for xiangqi newer than the 2004 ICGA figures.

Sources

• Chinese Xiangqi Association 《象棋竞赛规则》 (2011 edition, approved 2014-04) — piece movement, flying general, illegal moves.
• 《少年儿童象棋竞赛棋具通用技术要求》, sport.gov.cn — board and piece dimensions, junior specs.
• Yen, Chen, Yang & Hsu, “Computer Chinese Chess”, ICGA Journal 27(1):3–18, 2004-03 — complexity table.
• CCTV, 2006-08-09/10 and China News, 2006-08-10 — first man-versus-machine match results.
• Chinese Xiangqi Association announcement, 2024-09-19; chess administration announcement, 2025-01-15 (via Workers’ Daily and Xinhua).
• Shanghai municipal government, 2025-09-25 — 19th World Xiangqi Championship participation figures.
• olympic.cn (2003-09-17) and chiculture.org.hk — historical sources; 1983 Jiangyou bronze hoard.