The history of xiangqi: when Chinese chess was invented
The historical claims on this page are not machine-verifiable, unlike the rules content elsewhere on this site. Where the scholarship is contested — the origin of the game, the dating of the earliest manuals — the page says so rather than choosing the tidier account.
Direct answer
Xiangqi in the form played today is about a thousand years old: the cannon, the river and the palace are all documented from the Song dynasty (960–1279). The name 象棋 is considerably older and referred to earlier, different games. Most scholars treat xiangqi and western chess as separate descendants of the Indian game chaturanga, though the route and direction of that transmission are still argued over.
How old is xiangqi?
Two answers, and the difference between them is where almost every confident claim about this game goes wrong.
- The game as it is played now — roughly 1,000 years. A board of 9×10 intersections, a river across the middle, a palace at each end, and a cannon that captures by jumping. All of it is documented from the Song dynasty (960–1279), and the earliest surviving records of complete games come from the Southern Song encyclopedia 《事林廣記》 (Shilin Guangji, 13th century).
- The name — considerably older, but attached to other games. 象棋 appears in much earlier texts, and the Xiang Jing 《象經》 written under Emperor Wu of Northern Zhou in 569 CE describes a game of that name which is not this one.
So the popular line that xiangqi is “more than two thousand years old” is doing something specific: it dates the word and reports the number as the age of the game. The honest version is that xiangqi has a documented thousand-year history in its present form, and a longer, murkier prehistory shared with every other game in the chess family.
There is a Tang-dynasty (618–907) ancestor as well. The ninth-century collection 《玄怪錄》 (Xuanguai Lu, “Record of the Mysterious and Strange”) contains a story describing a game with generals, chariots and horses — recognisably related, and conspicuously missing the piece that defines the modern game.
The cannon dates the game
Of all the arguments about the age of xiangqi, the cannon is the one piece of hard evidence, and it is hard because it is a physical object as well as a game piece.
A game cannot contain a cannon before the culture playing it has one. Gunpowder artillery is a Song-dynasty development in China, and the cannon does not appear in any earlier description of the game. The character itself carries the date: 砲 is written with the “stone” radical, for the stone-throwing trebuchet that came first, and 炮 with the “fire” radical, for what replaced it. Both spellings survive on modern boards — Red conventionally writes 炮 and Black 砲.
This is why the Tang-era ancestor and the Song-era game are treated as different games rather than as one game with a long history. The cannon is not a refinement; it produces almost everything characteristic about xiangqi — the screen, the batteries down the central file, the fact that a shielding piece is simultaneously a firing platform, and the way endgames change character as the screens disappear.
Chaturanga, and the argument about where it came from
The mainstream scholarly position is that xiangqi and western chess are cousins rather than ancestor and descendant: both descend from chaturanga, played in India by around the sixth century CE, which spread west through Persia — where it became shatranj, and then European chess — and east along the trade routes into China.
It is not unanimous. A minority of writers, most prominently David H. Li and Sam Sloan, have argued for a Chinese origin with the transmission running the other way. The evidence on both sides is thin in the same place: there is a gap of several centuries between the games we can document, and the earliest material is literary rather than archaeological — stories and poems that mention a game, rather than boards and pieces in the ground.
What can be said without hedging is narrower, and more useful:
- The games are related. A chariot that is a rook in all but name is not a coincidence.
- Neither is a variant of the other. They had already diverged before either took its modern form.
- The direction and route of transmission are not settled, and any source that states it flatly in either direction is reporting a position rather than a consensus.
Our editorial standard is to say so where a claim rests on judgement. This whole section is such a claim: the rest of this site is verified against a rules engine, and history cannot be.
How xiangqi differs from chaturanga
The four divisions of chaturanga — infantry, cavalry, elephants, chariots — survive intact in xiangqi as the soldier, horse, elephant and chariot. What xiangqi added, and what it did with the rest, is where the games part company.
| Chaturanga | Xiangqi | |
|---|---|---|
| Playing surface | 8×8 squares; pieces sit inside them | 9×10 grid; pieces sit on the 90 intersections |
| Terrain | None | A river across the middle and a palace at each end |
| The king | Moves freely over the board | Confined to nine points for the whole game |
| Elephant | Jumps two points diagonally | Two points diagonally, blockable, and may never cross the river |
| Horse / knight | Jumps | Cannot jump — a piece on the first step blocks it (蹩馬腿) |
| Cannon | No such piece | Captures only by jumping exactly one screen — the defining addition |
| Counsellor / advisor | One, moving one point diagonally | Two, one point diagonally, confined to the palace |
| Facing kings | No such rule | The flying general: the two generals may never face each other down an open file |
The two structural inventions are the palace and the cannon. The palace fixes the target, which is why xiangqi mating patterns are geometric and repeatable rather than a chase across open ground. The cannon supplies a way to attack that target which needs the defender's own pieces in order to work.
For the comparison most English speakers are actually after — the modern games rather than the ancestor — see xiangqi vs chess.
What “xiangqi” means, and how to say it
It is pronounced “shyahng-chee” — Mandarin xiàngqí, IPA [ɕi̯âŋ tɕʰǐ]. Two things trip English speakers up, and both are the fault of the spelling rather than the speaker:
- x is not “ks”, and not quite “sh”. It is a sh made with the tip of the tongue down behind the lower teeth — closer to the sh in “sheep” than to anything containing a k.
- q is the ch of “cheese”. Not a k, and not a kw.
The tones fall and then rise: xiàng drops, qí climbs. “Zi-ang-kwee” and “eks-ee-ang-kee” are the two common misreadings, and neither will be understood.
The name is written 象棋. 棋 is the general word for a board game of this kind — the same character sits in 圍棋 (go). 象 is where it gets interesting: read literally it is “elephant”, which fits a game descended from an army with an elephant corps, and 象棋 is often glossed as “the elephant game”. But 象 also means “figure” or “image”, and the older 《象經》 used the name for a game of astronomical symbolism with no elephants in it at all. Both readings have a scholarly following, and the ambiguity is genuinely in the word rather than in the translation.
The classical manuals
Xiangqi has a written tradition older than any opening book in western chess, and the classical manuals (古譜) are still studied — the endgame compositions in particular are part of the standard training material, not museum pieces.
| Manual | Date | What it is |
|---|---|---|
| 《事林廣記》 Shilin Guangji | Southern Song, 13th c. | A general encyclopedia; contains the earliest surviving records of complete xiangqi games |
| 《百變象棋譜》 Baibian Xiangqi Pu | Traditionally 1522 | The earliest complete endgame manual to survive — though the dating is disputed, see below |
| 《橘中秘》 Juzhong Mi | 1632, by Zhu Jinzhen | “Secrets in the Tangerine”. The classic treatment of the Central Cannon, and still a standard reference for the sharpest lines |
| 《梅花譜》 Meihua Pu | Qing, by Wang Zaiyue | “The Plum Blossom Manual”. Made the case for Screen Horses against the Central Cannon — the cannon-versus-horse argument that shaped opening theory for centuries |
The 1522 date on 《百變象棋譜》 is the kind of claim worth flagging. It rests on a Ming-dynasty reprint attribution, and Professor Zhang Ru-an has argued that the manual is in fact a Qing-dynasty work presented as a Ming one by booksellers, for whom age was a selling point. The date is repeated as settled fact almost everywhere. It is not settled.
Timeline
| When | What |
|---|---|
| c. 6th century CE | Chaturanga played in India — the common ancestor of the chess family, on the mainstream account |
| 569 CE | 《象經》 written under Emperor Wu of Northern Zhou. The name 象棋 in use, for a different game |
| 618–907 | Tang dynasty. 《玄怪錄》 describes a recognisable ancestor — generals, chariots, horses, and no cannon |
| 960–1279 | Song dynasty. The cannon, river and palace appear; the game takes essentially its modern form |
| 13th century | 《事林廣記》 records the earliest surviving complete games |
| 1632 | 《橘中秘》 — the first great opening manual |
| Qing dynasty | 《梅花譜》 answers it with the Screen Horses defence |
| 1993 | The World Xiangqi Federation is founded and international rules follow — the basis of the rules this site implements |
The rules have been remarkably stable across that span. What changed in the modern era is the adjudication of repetition, which was settled by custom and by the judgement of onlookers for most of the history of the game, and only became a written, decidable procedure in the twentieth century.
Related questions
When was xiangqi invented?
Xiangqi took its modern form during the Song dynasty (960–1279), which makes it about a thousand years old as it is played today. No single inventor is recorded, and none is likely: the game accumulated its pieces over centuries. Ancestors under the same name existed in the Tang dynasty and earlier, but without the cannon.
How old is xiangqi?
About 1,000 years in its present form. The claim that it is 2,000 or 3,000 years old dates the name 象棋 rather than the game — the name is attested far earlier, attached to games that were not this one.
Who invented xiangqi?
Nobody, in the sense the question implies. Legends attribute it to Han Xin, a Han-dynasty general, among others; there is no evidence for any of them. The game as played now is the product of gradual change across the Tang and Song dynasties.
How is chaturanga different from xiangqi?
Chaturanga was played inside the squares of an 8×8 board and had no river, no palace and no cannon; its king moved freely, and its elephant and horse both jumped. Xiangqi moved play onto the 90 intersections of a 9×10 grid, confined the general to a nine-point palace, made the horse and elephant blockable, barred the elephant from crossing the river, and added the cannon — which has no counterpart in chaturanga or in any of its western descendants.
Is xiangqi older than chess?
Both reached their modern forms at roughly the same time — xiangqi under the Song, European chess in the fifteenth century once the queen and bishop gained their present moves — so on that measure xiangqi is the older of the two. Their shared ancestor is older than either, and which descendant diverged first is not established.
Did xiangqi come from India or from China?
The mainstream view is that it descends from the Indian game chaturanga, transmitted eastward. A minority of writers argue for a Chinese origin and a westward transmission. The documentary gap between the games is several centuries wide in both directions, and the question is genuinely open rather than merely contested.
How do you pronounce xiangqi?
“Shyahng-chee” — Mandarin xiàngqí. The x is a soft sh rather than a ks sound, and the q is the ch of “cheese” rather than a k. The first syllable falls in pitch and the second rises.
