Endgame verdicts, proven
Every position, move and legality claim on this page is verified programmatically against our open rules engine, which implements the WXF laws for movement, checkmate and both repetition rulings: a side that forces a repetition by perpetual check or perpetual chase loses rather than drawing. The chase half has not been validated against the official casebook.
Direct answer
For small material the verdict is not a matter of opinion: the positions can all be counted. A chariot beats a bare general in every one of 3,834 legal positions, and still wins all 125,118 positions against a general with both advisors. A horse and a soldier win every one of 276,390. A cannon and a soldier do not — they win 86.8% and draw the rest, because a cannon needs a screen and the defender can refuse to provide one.
What “proven” means here
Every endgame verdict on this page was produced the same way, by our own engine, and none of it is quoted from a manual:
- Enumerate every legal position for the material — every square for the attacker, every palace point for each general, every advisor and elephant point — discarding the illegal ones, such as the two generals facing down an open file.
- Solve backwards from the end. A position where the side to move has no legal move is a loss for that side, and the result propagates outwards: you win if some move hands the opponent a lost position, you lose if every move hands them a winning one, and anything left over is a draw.
- Apply the 60-move rule afterwards, not during. Whether a game hits it depends on the route taken, not on the position, so the search reports distance to mate and the rule is checked against that. It matters: a win needing 61 moves is a draw in a real game.
Two things make these results differ from what a chess player expects. Having no legal move loses in xiangqi — it is not a draw, which removes the defender’s best resource. And there is no promotion, so a soldier is worth what it is worth for the rest of the game.
The verdicts
Red to move in every count. “Longest win” is the worst case with both sides playing perfectly — not an average, and not a typical game.
| Material | Positions | Verdict | Longest win |
|---|---|---|---|
| Chariot vs bare general 单车对光将 | 3,834 | Win, all of them | 2 moves |
| Chariot vs general + 1 advisor | 33,345 | Win, all of them | 5 moves |
| Chariot vs general + 2 advisors 单车对双士 | 125,118 | Win, all of them | 12 moves |
| Chariot vs general + 2 elephants | 80,328 | Win, all of them | 12 moves |
| Horse vs bare general 单马对光将 | 4,590 | Win, all of them | 7 moves |
| Horse + soldier vs bare general | 276,390 | Win, all of them | 7 moves |
| Soldier vs bare general 单兵对光将 | 4,725 | Win in 4,275; draw in 444 | 13 moves |
| Cannon + soldier vs bare general | 293,823 | Win in 255,144 (86.8%); draw in 38,679 | 10 moves |
| Cannon vs bare general 单炮对光将 | 4,914 | Draw, every one | — |
Every win listed fits inside the 60-move rule, with room to spare. The longest win anywhere in this table is 13 moves.
Two advisors do not save the general
The chariot result is the one worth sitting with. A general with both advisors still loses all 125,118 positions — and the advisors are part of the reason. They occupy the palace points the general would otherwise step to.
Below, Black has both advisors on the board and it is mate in one. There is exactly one move that does it.
Red to play and mate in one.
Xiangqi position. Red: chariot a0, general d0. Black: advisor d9, general d8, advisor d7. Red to move.
The result that surprises people
A horse and a soldier beat a bare general in every one of 276,390 positions. A cannon and a soldier do not — they win 86.8% and draw 38,679.
The reason is the cannon itself. A cannon captures only by jumping exactly one piece, so against a lone general there is often nothing to jump: the soldier has to act as the screen, and a defender who keeps his general away from the one file or rank where that works cannot be mated. A horse needs no screen, which is why the same material with a horse is a clean win.
This is the general shape of cannon endings and the reason a lone cannon draws against a bare general in all 4,914 positions while a lone horse wins all 4,590. More on cannon endings →
What we have not proved
The famous verdict this page does not settle is chariot against a full defence — a general with both advisors and both elephants (士象全), which the manuals call a draw. That position set runs to roughly eleven million positions, past what this solver handles in one pass, so we have not checked it and do not assert it.
The same applies to the larger practical endings — chariot against horse and cannon, two chariots against a full defence. They are on the list. Until they are solved, this page states only what has actually been counted, which is the point of it.
Related questions
Does a single chariot beat a general with two advisors?
Yes, always. We solved all 125,118 legal positions and Red wins every one, in at most 12 moves. The advisors do not save the general — they occupy palace points it would otherwise escape to.
Can a lone cannon mate a bare general in xiangqi?
No. All 4,914 legal positions are drawn. A cannon captures only by jumping exactly one piece, and against a lone general there is nothing to jump. A lone horse, which needs no screen, wins all 4,590 of its positions.
Is a cannon and a soldier enough to win against a bare general?
Usually but not always: 255,144 wins out of 293,823 positions, so 13.2% are drawn. The soldier has to serve as the cannon’s screen, and a general kept away from the squares where that works cannot be mated.
How were these verdicts calculated?
By enumerating every legal position for the material and solving backwards from the positions where the side to move has no legal move — which in xiangqi is a loss, not a draw. The 60-move rule is applied to the resulting distance-to-mate afterwards, since whether a game reaches it depends on the route.
.