Horse and Soldier Endgame in Xiangqi: What Wins, What Draws

High, low and bottom

Chinese terminology splits a crossed Soldier three ways. A high Soldier (高兵) has crossed the river but has not gone below the defender's Soldier rank. A low Soldier (低兵) is deeper, on the palace's front or middle rank. A bottom Soldier (底兵) has reached the back rank and can only step sideways.

The reason height matters is simple geometry. A Soldier never moves backward. Once it reaches Black's second rank, it can only ever touch the back two ranks again. Elephants that stand on Black's third rank, at the front of the palace or on the edges, or on the river bank, are out of its reach for the rest of the game. A high Soldier can still choose a side, attack those Elephants, and trade itself for a guard at a moment of its choosing.

The easy cases

Against a bare General, a single Advisor, a single Elephant, two Advisors, or one Advisor and one Elephant, Horse and Soldier win, even when the Soldier is low. The old xqinenglish pages list Horse with a low or bottom Soldier as a sure win against two Advisors and against Advisor and Elephant, and the Chinese reference has a heading for Horse and bottom Soldier beating Advisor and Elephant. Pai's database puts all of these above ninety-nine percent, high or low.

Two Elephants are the first real test. With a high Soldier it is a win, over ninety-nine percent of positions in the database. With a low or bottom Soldier the books only show wins from favourable positions, and the database figure drops to about three in four.

The model win against two Advisors and an Elephant

Horse and Soldier against two Advisors and one Elephant, 马兵必胜单缺象, is a listed book win, and the method is worth learning move by move. The Soldier first blocks the Elephant's road back to the centre. The Horse then locks the Elephant where it stands. The Soldier walks sideways and takes it. Finally the Soldier is traded for an Advisor, which leaves Horse against one Advisor, another book win.

Here is the example from the Chinese reference. Black's General is at home, one Advisor is on the back corner on Red's left and the other on the palace centre, and the Elephant stands in front of the palace on the central file. Red's Soldier is on Black's Soldier rank, on the file of the palace's left edge. Red's Horse is on the central file on Black's side of the river, nearest the river, and Red's General is at home.

The Soldier steps sideways onto the central file and attacks the Elephant, and in the book the Elephant runs back to its starting point just left of the palace. The Horse comes round toward it, Black's General steps to the right back corner, and the Elephant hops out to the left edge of Black's third rank. The Horse then lands right beside it, one point toward the centre on the same rank. Both points the Elephant could hop to are now covered by the Horse.

Nothing in Black's camp can rescue it. The Soldier walks along Black's Soldier rank to the edge file and takes the Elephant from in front, and the engine confirms Red wins it by force within five moves whatever Black does. Notice what made it work: the Soldier was still high, so the Elephant on the third rank was still a target.

With a low Soldier the same material is a hard win. Tu Jingming shows the defence: keep the Elephant and the General on opposite sides of the board, so the Horse cannot cover both. Pai's database puts the high Soldier version at over ninety-nine percent and the lower Soldier at about seventy-two.

One Advisor and two Elephants

Against one Advisor and two Elephants, 单缺士, the pattern repeats. Tu Jingming lists Horse and high Soldier as a book win and Horse with a Soldier on the palace's middle rank as a hard win. The database shows about ninety-five percent against fifty-nine. Two Elephants give the defender a solid central file, and a Soldier that has reached the second rank can never touch an Elephant standing in front of the palace.

Against the full guard

Here the books are clear: Horse and Soldier are not enough against two Advisors and two Elephants. Tu Jingming shows wins only from badly placed guards, and Pai's database finds only about a quarter of positions won with the Horse side to move. That figure includes every clumsy arrangement of the guard. Against a defender who has set up properly, expect a draw.

Even giving the attacker its own Advisor and Elephant changes very little. The database has that ending at almost exactly the same share of wins. What does change it is more force. Put a Chariot in place of the Horse and the same Soldier wins against the full guard in practically every position.

Playing it well

If you have the Horse and Soldier, do not hurry the Soldier. The commentary to Tu Jingming's high Soldier example is blunt: slow and steady, and do not advance the pawn too fast. Every step forward is permanent, and a Soldier that drops to the second rank too early can no longer reach the Elephants that matter.

Bring your General to the central file early, because the last stage, Horse against a bare General, ends in a stalemate in which your General holds that file. Be careful with trades. Horse alone against one Elephant and Soldier alone against one Elephant are both normally draws, so only give up the Soldier when the result is Horse against a lone Advisor or a bare General.

If you are defending, keep the Elephants off the edges where a Horse can lock them, keep at least one Elephant out of a low Soldier's reach, and place the last Elephant on the side away from your General. If the attacker pushes the Soldier down to your second rank early, be glad: from there it can never again touch an Elephant standing on your third rank.

The quickest way to absorb this is to play the positions. Set up Horse and Soldier against two Advisors and an Elephant on a board, try the high and low Soldier versions with a friend, then go looking for these endings in full games against XiangqiPlay's computer, across its eight levels from Easy to God. You will need a free account to play.