GM Gukesh D – GM Ortik Nigmatov

After 17.Nd2 the count stood at roughly a pawn for White, a solid edge out of a London structure that should have been finished off in quiet. Then 18.Ra5, 19.Qa1, 20.Rxa7 picked up the a-pawn with a rook lift; each move was a practical choice and not the best, and after 21.Qxa7 Qd6 the count was back to something under a fifth of a pawn. The advantage was on negotiating ground again.

For 22.c4 Gukesh thought twelve minutes twenty-three. A second candidate was 22.b4, which attacks the c6 pawn directly and stands level in the count. The difference is a fraction of a pawn, not a better move but a matter of taste. Nigmatov answered 22...fxe3, and with the forced reply 23.fxe3 the count jumped back to a good pawn and a half for White.

Probably the long think was not for the better move, but for the certainty to see the plan through to 25.Nf1. For your own games take this: a twelve-minute think in a near-level position with two similar moves does not buy the better move; it buys clarity about the next plan, and that clarity is the reason the knight was ready three moves later.

The position after 24...Qb4 shows where Gukesh wants to go. The knight sits on d2 in its own way, and on f5 it finds a target square that lays claim to the dark squares around the black king. The route is Nf1-Ng3-Nf5, three moves for one piece. 25.Nf1, nine minutes twenty-six on the clock, was the first; it clears d2 and blocks none of its own lines.

Nigmatov played 25...Qxb3 and took the pawn. Called by its name this is a tactical error, not a matter of move order; the one move that held the balance was 25...Bg4 with the attack on the bishop. The refutation was 26.Qe7. The white queen attacks the knight on f6, blocks no piece of its own, and combined with the knight route the defence does not hold; the count jumped to over two pawns for White. Nigmatov played the move in one minute twenty, Gukesh the reply in five minutes thirty-three.

26...Qb7 27.Qd6 28.g4 29.Qe5 built the squeeze, each move the best, and 30.Ng3 put the knight on its stepping square. 31.Nf5 earns the name: the move attacks the black queen on e6, covers the white queen on e5 from behind, and blocks no line of its own. In the human layer at this level most players wanted cxd5; 31.Nf5 was the choice of the few, and Nigmatov had to play 31...Qxe5. For your own games take this: a regrouping over three moves earns its keep when the target square carries three jobs at once.

Between 32.dxe5 and 35...Nxe3 a whole row was cleared: 32...Nxg4 gave up a knight for Nigmatov, 33.Ne7+ and 34.Rxf8 took the rook, 35.Nxc8 the bishop, and 35...Nxe3 the pawn back. After the liquidation the count stood at three pawns for White, the material at bishop and knight and three pawns against knight and five pawns, and Gukesh had been under five minutes since move thirty-one.

With 36.e6 the wrong pawn was pushed. The best move was 36.cxd5 cxd5, which dissolves the black pawn chain and holds the advantage at roughly three pawns for White; 36.e6 gave about one and a half pawns of that back. On the objective count this is a tactical error, and the move is the human reach: instead of trading the pawn White no longer needs, the own passed pawn is pushed a rank further.

Nigmatov did not find the reply. 36...Nf5 stayed within reason, 37.c5 in one minute sixteen held the pawn chain, and 37...Ne7 ended the resistance; the best move was 37...d4 with counterplay against the pawn on e6. From 38.Nxe7 Kxe7 39.Bg4 40.Bf5, all played in thirty-seven, forty-seven and thirty-five seconds, the count was clearly in White's favour again. For your own games take this: in a won position under time pressure, the pawn trade that does not change your own count is usually the quieter move than the pawn push that claims a new square.