The run
Between August 17 and September 16 this year, Yin Sheng played 31 games across three events and did not lose one of them: eleven wins and twenty draws through the qualifier for China's top league, the Shanghai Cup, which he won, and the first stage of the league itself. I ran every game through the same engine analysis Mistboard gives your own games. His accuracy sat between 89 and 100, with a median of 98, and at no point in any of the 31 games did the engine rate his winning chances below 44 percent.
I also ran every other game from the Shanghai Cup and the league stage, 250 games in all, to see what that accuracy means against the field. It means less than it sounds: the median across every player was 96.9, and six players who played six or more games matched or beat his 98. Accuracy is not what separates him. Not losing is.
Some context on the opposition, because the names will mean little outside China. Xiangqi does have ratings: the Chinese Xiangqi Association ran an Elo-style system for over forty years, with the very top players around 2750 to 2800, before retiring it in January 2026 in favour of a tournament-points ranking. The last list it published is from September 2023, so the numbers below are three years old and frozen, but they are the only official ones, and they show the field. Every rated opponent in the run stood above him on that list. Chen Shaobo, who he beat in the final, was too young to be on it at all.
| Player | Rating | Rank in China | Against Yin Sheng this run |
|---|---|---|---|
| Meng Fanrui | 2567 | 17 | 1 win, 6 draws |
| Cao Yanlei | 2551 | 23 | 1 draw |
| Xu Wenzhang | 2546 | 24 | 2 draws |
| Chen Hongsheng | 2517 | 36 | 1 win, 1 draw |
| Huang Guangying | 2501 | 42 | 1 draw |
| He Wenzhe | 2471 | 65 | 2 wins |
| Jin Bo | 2467 | 67 | 1 win, 1 draw |
| Liu Bohong | 2449 | 84 | 2 draws |
| Jin Yuyan | 2448 | 86 | 1 win |
| Yin Sheng | 2454 | 78 | 11 wins, 20 draws, no losses |
How he plays
With red, he opens with a soldier. Ten of his fifteen games as red begin by advancing the third or seventh soldier, the family of openings called "the immortal points the way," and four more by raising an elephant. Not once all month did he play the central cannon, the most direct opening in the game. He is not trying to win the opening; he is trying to reach a middlegame he understands.
With black, the engine can barely mark him. His median accuracy as black was 98.6, against 95.6 as red, and he made one mistake in sixteen games with the black pieces. All four of his blunders came with red, and three of those were in games he went on to win anyway. He errs when he is pressing, not when he is holding.
His draws are short and his wins are long. Half of the twenty draws were over inside thirty moves, while the median win took fifty-two. In nine of the eleven wins, the other side made the first real mistake and the rest of the game was conversion. The other two he ground out against opponents the engine never faulted at all, over 52 and 84 moves of small edges.
In seventeen of the 31 games the engine found nothing on his side worth marking. Across the whole month it found three brilliant moves, three great ones, one missed mate in one, and otherwise a player who simply does not give anything back.
The league qualifier
Hangzhou, August 17 to 19: two wins and three draws. The run starts on day one, with black against Zhang Rui of Tianjin. Zhang's horse move on move 21 is the only error the engine finds in the game, and Yin Sheng needs 38 more moves to turn it into a win. This is the shape of most of his wins: one error from the other side, then no way back.
The Shanghai Cup
September 9 to 13, and he won it: six wins and eleven draws. A knockout with rapid tiebreaks, which is how seventeen games fit into five days. Seven of them were against Meng Fanrui: six draws, one of them 162 moves long, and this one. Move 19 is a pair of edge-soldier pushes, both of them wrong, Meng's more so, and Yin Sheng's chariot move on move 20 is the only move that punishes it.
Against Jin Bo, 88 moves. On move 30 he leaves a chariot where it can be taken, and the engine agrees it should be left there. Then, after black's 55th move, there is a mate in one on the board. He plays a cannon move instead, and wins 32 moves later. The board opens on that position; find the mate before you step forward.
The final against Chen Shaobo went to four games, three draws and this one. Chen's 22nd move gives up 14 points of winning chances, and Yin Sheng plays the rest of the game at 95 percent accuracy. There is no single killing move. He just does not give any of it back.
The league, stage one
Back in Hangzhou, September 14 to 16: three wins and six draws. This is the messy one. Against Chen Hongsheng, with black, the engine counts four blunders for red and one real mistake for Yin Sheng, and the evaluation swings both ways for twenty moves before red's 39th and 42nd hand the game over. It is on the page because the section above would otherwise give you the wrong idea of what these games look like from the inside.
What's next
The league's second stage is in October, and the national individual championship, the title that decides who China's best player is, runs from November 21 to 29 in Changchun. In 2023, on the day he became a master, Yin Sheng said he would give himself five years to a real result there.
All 31 games
Every game of the run is in a Mistboard study, with the engine's judgments and lines on each one, and the three events are in the broadcast archive. Game records from dpxq.com; analysis by Pikafish through Mistboard's review pipeline, September 2026.