Xiangqi endgame

Three rules decide it, and all three surprise a chess player

Xiangqi endgames are sharper than chess endgames, and the reason is structural: there is no promotion to play for, stalemate is a win rather than a rescue, and both generals are locked in their palaces where they can still see each other. Learn the three rules below and you will convert positions that used to slip away. Coming from chess? The diff page covers the rest of the differences.

Rule one: stalemate is a loss

If the side to move has no legal move, that side loses. Not a draw — a loss, recorded with a winner, exactly like checkmate. This one rule removes the entire chess endgame skill of scrambling for stalemate, and replaces it with a second way to win.

So a general with no escape squares is already lost even if nothing is attacking it, and the practical method that follows is take squares away. You do not always need to line up a check; sometimes you just need to leave your opponent with nothing to move. It also means a bare general facing an active piece is in serious trouble, because a general with three legal moves and no help is two accurate moves from having none.

Black to move and nothing is attacking the general — but d0, f0 and e1 are all covered by the two horses, so Black has no legal move at all. In xiangqi that is a loss, recorded with a winner exactly like checkmate. Taking squares away is a second way to win.

Rule two: nothing promotes

A soldier that reaches the far rank stays a soldier, and on that rank it can only step sideways for the rest of the game. There is no queen to play for and no passed-soldier race to calculate.

What this does to endgame technique is easy to state and hard to remember: a soldier is at its most valuable a rank or two before the end, standing next to the enemy palace taking squares from the general — which, by rule one, is how you win. Push it to the back rank and you have converted your best attacker into a spectator. The traditional name for one that has done that is 老兵, the old soldier.

This soldier will never do anything again. On the far rank it can only step sideways — d0 and f0, and that is the whole of its remaining life — and it can never retreat, so the general behind it on e2 is permanently out of reach. A rank earlier it was taking squares from the palace.

Rule three: your general is a weapon

The two generals may never face each other down an open file. That is usually taught as a restriction, and in the endgame it is nothing of the kind — it is a long-range attack that costs you no material. Any piece standing between the two generals is pinned to that file and cannot step off it, and the enemy general can never move onto the file at all while yours is looking up it.

In practice this means a general that has walked to the middle file in an endgame is doing the work of an extra piece, and that a mating net which looks one attacker short often is not.

What the pieces become

Piece	In the endgame
車 Chariot	Unchanged and dominant. It never needed anything from the position, which is why it costs 9.
馬 Horse	Better than it was. Its only weakness is a piece beside it blocking its leg, and there are barely any pieces left.
砲 Cannon	Worse than it was, sharply. It must jump exactly one piece, and an empty board offers none — a lone cannon cannot give a single check.
兵 Soldier	Strong past the river, where it moves sideways as well as forward, and weak once it has nowhere left to go.
士象 Advisors and elephants	The whole defence. They never left home, so in the endgame they are all the general has.

The cannon's borrowed screen

Because a cannon needs something to jump, attacking with one in an endgame usually means borrowing a screen — and the most convenient screen on an otherwise empty board belongs to the defender. A cannon can check a general straight through that general's own advisor.

Which produces the strangest defensive advice in the game: against a cannon, a defender is sometimes better off without a piece in front of the general, and moving an advisor out of the way is a real move rather than a blunder. It is the other half of 缺象怕炮 — the missing elephant fears cannons, and the present one may be working for the enemy.

The cannon on e5 needs exactly one piece to jump, and Black has kindly supplied it: the advisor on e1 is the screen, so the general is in check through its own defender. Move that advisor off the file and the check simply stops — which is why it is a real move here and not a blunder.

Defending: keep the shield, count the attackers

Do not trade advisors for soldiers. The exchange looks fine at 2 against 1, and it is how most lost endgames start. A shield with a hole in it is a different shield.
Two chariots beat a general missing an advisor — 缺士怕雙車, the oldest warning in the book. If you have lost one, trade a chariot off.
Keep your general off the open file unless you are the one using it.

Converting: the short version

Trade pieces, not soldiers, when you are ahead. Every trade shrinks the defender's options and your soldiers are what will actually finish the game.
If you must choose, keep the horse over the cannon. The emptier it gets, the more that choice matters.
Look for squares before checks. Stalemate is a loss, so the last move of a game is often not a check at all.

The puzzle set is mostly forced wins — mate in one, two or three — which is endgame calculation with the fog removed. Solve a few, then play a game and try to reach one on purpose. If you want the attacking shapes that come before this, they are in the middlegame guide.