Mobei, 119 BCE: The Raid That Broke the Steppe

The problem

For twenty years the Han and the Xiongnu had traded raids across the northern frontier, and the Han had been losing the argument. In 129 BCE the young Wei Qing (衛青) took the Xiongnu sacred site at Longcheng in one long strike, and for the next decade Emperor Wu sent the same two men — Wei Qing, and his nephew Huo Qubing (霍去病) — out again and again. The Xiongnu king, the Chanyu, took Zhao Xin’s advice and pulled his people north of the Gobi, hoping the empty desert would do what no battlefield could.

The campaign of 119 BCE was built to break that. Two columns were to leave the frontier together: Wei Qing from Dingxiang (定襄) in the west, and Huo Qubing from Dai (代郡) in the east. They were to pass through the Xiongnu, not fight a border war, and to destroy the Xiongnu state inside its own pasture. Each column is given in the sources as fifty thousand cavalry and over a hundred thousand infantry, with about a hundred and forty thousand horses mobilised overall — though one version of the same account divides every one of those figures by ten, and we come back to that below.

The plan was changed before it began, for the worst possible reason. A Xiongnu prisoner of war reported that the Chanyu’s main force was at Dai, in the east. The report was false. Emperor Wu, who had been favouring the younger Huo and giving him the best troops, ordered the columns to swap routes — partly to catch the Chanyu, partly to put the better general in front of the larger prize.

Wei Qing’s column was the weaker for it. The elite cavalry had gone east with Huo, and Wei carried five generals expecting prestigious commands, among them Li Guang (李廣), old, famous, and — the emperor privately messaged Wei — unlucky. Li Guang had been promised a vanguard; the promise was quietly withdrawn. He was put with Zhao Shiqi (趙食其) on a barren eastern flanking route instead, and he objected so violently that he walked out of the main camp in anger.

The movement

Huo Qubing’s column left Dai and rode north. The distance is the first thing the sources cannot agree on, and we give all of it rather than pick: the English-language accounts give “over 1,000 miles”, the Chinese-language accounts give “over 1,000 li”, and Huo Qubing’s own biography gives “over 2,000 li”, which is somewhere between 800 and 1,000 kilometres. A thousand miles is nearly twice as far again. Whichever figure is right, this was a march out of the Chinese empire, into the desert, with no supply column and no base to fall back on.

He found the Xiongnu’s Left Worthy Prince — the title the sources give as both “Wise King of the Left” and “Worthy Prince of the East”, without ever explaining the mapping — and destroyed him. The figure the record is most confident about is Huo’s own: 70,443 killed, three Xiongnu lords and eighty-three nobles captured. The column took a fifth of its strength in the doing, and then lived off what it had taken.

Huo did not stop there. He rode on to the mountains the sources call Langjuxu (狼居胥山) and Guyan (姑衍山), where he performed the sacrifices to Heaven as the emperor’s representative — a ritual claim that Han authority had reached the Xiongnu’s own sacred ground. Then he pushed further north to the salt lakes the Chinese call the Han Sea (瀚海), a name modern scholarship identifies with Lake Baikal, though that identification is not certain. A division under Lu Bode (路博德), sent out on a flanking route from Right Beiping (右北平), joined him there with its own tally of 2,800 killed, and the two columns rode home together.

Wei Qing’s column had the harder road. After what the sources put at over 800 kilometres, or in another version over 500 miles or over 2,000 li, it met the Chanyu’s main force of 80,000 cavalry — more than Wei had been told to expect, because the original plan had assumed Huo would be dealing with the Chanyu and Wei with the lesser enemy. The Han were tired, outnumbered, and short of the eastern flanking division, which had not arrived.

The Xiongnu charged. A vanguard of 10,000 cavalry came in, and Wei Qing answered with a formation the Han records call the Wu Gang chariots (武剛車) — heavy armoured war wagons locked into rings, converting the whole army into a set of moving fortresses in which archers and crossbowmen could shoot while protected from the Xiongnu charge. A cavalry force of 5,000 was posted to cut down anything that got inside the ring. The Xiongnu could not break it. The battle settled into a stalemate with neither side able to make ground.

It held until dusk, when a sandstorm came up and put the battlefield in darkness. Wei Qing used it. The Han main force went in under cover of the dust and took the Xiongnu from both flanks; their lines broke, and the Chanyu escaped with only a few hundred men. The Han killed more than 19,000 in the fight, pursued the survivors another hundred miles or so to the Khangai Mountains, besieged and took the fortress of Zhao Xin (趙信) in the Orkhon valley, and burned it to the ground.

The decisive moment

Three decisions made the campaign, and only one of them was good.

The first was the route swap, taken on a prisoner’s report that turned out to be false and on a court favouritism that the sources state plainly. It gave Huo the better force and the eastern theatre, and it gave Wei Qing a command of leftovers and the main enemy. By luck rather than judgement, the arrangement handed the greater prize to the better general.

The second was Wei Qing’s decision to stop hitting and start holding when he was outnumbered. The chariot ring is not a manoeuvre from the text; it is a Han answer to a problem the text does not address, and it cost the army the mobility that had brought it there. Standing still to survive is what Sun Tzu’s text advises when you are weak — but the text assumes you chose the ground, and Wei Qing had not.

The third was Huo Qubing’s decision not to stop at the battlefield. The capture of the Left Worthy Prince was worth 70,443 and three kings. The decision to keep riding — to the sacrificial mountains, to the salt lakes, and out of the country altogether — is what broke the Xiongnu as a power rather than as a battle. It is also what made the campaign defensible at all, because a raid that came home with a number would have been an expense.

The outcome

The Xiongnu did not recover. The Chanyu disappeared for more than ten days, and his own people installed a new leader in his absence, who then had to be removed when he reappeared. The northern tribes retreated further into the steppe, and the Han had bought seven years of quiet — a truce broken only by a Xiongnu raid at Wuyuan in 112 BCE.

The Han paid for it in a way the campaign’s famous numbers do not show. Across the ten years of expeditions from 129 to 119 BCE, the Han army lost close to eighty per cent of its horses — from combat, from the marches themselves, and from sickness contracted where the Xiongnu had fouled the water with dead cattle. On the figure of 140,000 horses mobilised, that is more than 100,000 animals. New taxes were levied to pay for it, the registered population fell through famine and taxation, and the Xiongnu lost millions of livestock in the process. The campaign won the frontier and spent the treasury.

One part of it ended badly. Li Guang’s and Zhao Shiqi’s eastern division lost its way in the desert, missed the battle entirely and rejoined the main force on the way home. Both were summoned before a court martial for failing their orders and putting the strategy at risk. Li Guang, who had spent his career without the battle distinction that would have earned him a marquessate, killed himself rather than face the court.

Huo Qubing died in 117 BCE, aged twenty-three. He had been given a peerage, a house, and a gift of six hundred jin of gold, and he refused the house, saying that while the Xiongnu were not yet destroyed there was no point in having a home. His tomb was built at Maoling in the shape of the Qilian Mountains, with a stone horse standing on a fallen Xiongnu.

What the record cannot tell us

The march distance cannot be settled from the sources in front of us, and it matters more than any other number here, because the whole claim of the campaign rests on how far into the desert the Han cavalry actually rode. The English-language accounts and the Chinese-language accounts of the same event give figures that differ by roughly a factor of two, and the two are not even using the same unit — one tradition counts miles, the other li, and the li figure itself appears as both 1,000 and 2,000.

The size of the armies is equally unstable. One version of the English account gives each column 5,000 cavalry and 10,000 infantry; another gives 50,000 cavalry and 100,000 infantry; the infobox of the same article gives 300,000 cavalry and troops overall and 20,000 for one column. Han losses are given as fewer than 7,000, about 10,000, about 20,000 cavalrymen, or “tens of thousands”. These are not small differences of emphasis. We give the highest consistent figure and flag the rest rather than present a settled number.

Even the best-attested figure needs a caveat. The 70,443 is a Han tally, produced by the people who won, in a campaign the Han state had every reason to inflate. The two theatre totals of 70,443 and 19,000 come to 89,443, which is close to the official figure of around 90,000 — that consistency is mild reassurance and nothing more.

The exact site of the decisive battle is not given, and the endpoint of the pursuit is named two ways in the sources: the Khangai Mountains and the fortress of Zhao Xin, or the Zhaoxin river. The identification of the Han Sea with Lake Baikal rests on modern inference, not on a place named in the record.

And the largest thing the record cannot tell us is what these commanders thought the Art of War was. The biography that praises Huo Qubing most also records, in Sima Qian, that he held innate strategic judgement sufficient for victory without studying ancient texts such as the Sunzi, and that he disdained such writings. The campaign’s methods are explicable as steppe warfare — light cavalry, feigned retreat, composite bows — without any appeal to the text. We cannot show that the text influenced it, and the honest reading is that it did not need to.