Sindarov beats Anand: the rook endgame that turned on one move

Short answer

Javokhir Sindarov beat Viswanathan Anand with White in the opening round of the Global Chess League 2026, in 57 moves. The game stayed within half a pawn of equality for 36 moves and was decided by a single endgame move: 37...Rxb4, when 37...Re5+ held the draw. This article reconstructs the decision, shows the saving line — the rook trading itself for the passed pawn — and explains why the mistake appeared at a moment when Black still had four minutes on the clock.

Event: Tech Mahindra Global Chess League 2026

How this was compiled: Game from round one, played on 5 September 2026. Evaluations measured with Stockfish 19 at depth 24.

The essentials in a few lines

• The game was level until move 36: measured every four moves between move 20 and move 36, the evaluation stays within a tenth of a pawn.
• The decisive move was 37...Rxb4, which takes the evaluation from +0.25 to +4.27 at depth 24; the defence was 37...Re5+, the move the engine calls best.
• The saving idea is to give the rook for the passed pawn: 37...Re5+ 38.Kd2 Rd5+ 39.Kc3 Rxd7 40.cxd7, and only then does the king run to e7.
• After 40.cxd7 only 40...Ke7 holds — 40...Kg7 loses, with the evaluation jumping to +11.28 because the new queen dominates the board before the knight can help.
• Anand played the losing move with four minutes on the clock against Sindarov's two minutes and forty seconds: in this game the mistake did not come from time trouble.

The context: the challenger opening against the champion of 2000

The pairing had the obvious generational appeal: Anand first became world champion in 2000, before Sindarov was born. The technical interest, though, lies elsewhere. Sindarov arrived at the league as the title challenger, in a match that will be the youngest in history — both he and Gukesh will be 20 when the first game is played, on 24 November.

The win over Anand did not come from laboratory preparation. Sindarov chose 5.Bd3 in the Sicilian, a modest move that steers the game away from analysed paths and leaves both sides playing with their heads rather than their memory. Ten moves later the queens were still on and the evaluation had not moved from zero.

The rest of the day is worth recording, because the headline alone overstates things: on that same first day Sindarov drew with Alireza Firouzja as Black, and the following day he lost to Ian Nepomniachtchi, also as Black. The win over Anand is a result, not a demonstration of superiority.

Thirty-six moves without an advantage for either side

After 10.e5 dxe5 11.Qe2 Bg7 12.Bxe5 O-O, White had an extra queenside pawn that was worth nothing and Black had the bishop pair that was worth nothing either. Anand traded the light-squared bishop on f5, Sindarov traded the dark-squared one on g7, and the game became rook and knight against rook and knight with pawns on both wings.

It is the kind of position usually decided by whoever lasts longer playing without a target. Sindarov did not try to invent: 26.Kf1, 27.b3, 28.c4 and 29.Ke2 are slow improving moves, with the king walking towards the centre and the queenside pawns gaining space one step at a time.

The first real movement came on move 33: 33.b4 cxb4 34.c5, creating the passed pawn that would decide everything. Even so the evaluation did not move — the engine stayed in the same band after 34...Rb5 35.c6 Na6. Black had the rook behind the pawn and the knight waiting. Everything was under control.

The move that decided it, and what it decided along with it

After 36.Rd7+ Kf6 37.axb4, the position in the diagram below poses a single question: what does Black do about the c6-pawn, two squares from promotion and supported by the knight on a5? Recapturing on b4 answers the wrong question — it wins a pawn on one wing while the pawn on the other wing runs.

The sequence is relentless and not at all spectacular: 38.Rxa7 Rb2+ 39.Kd1 Nb4 40.c7 Rc2 41.Rb7 Rc3 42.Rxb4 Rxc7. Black does stop the pawn, but pays the knight for it, and rook against rook and knight with pawns on both wings is technically winning. Sindarov needed another fifteen moves to convert, without a scare.

The detail that carries the lesson: Anand played 37...Rxb4 with four minutes and four seconds on his clock, against Sindarov's two minutes and forty. The player short of time was the winner. It is a useful counterweight to the lazy explanation that an endgame mistake in rapid is always the clock — here it was assessment, not haste.

The defence: give the rook for the pawn, then run with the king

The idea runs against the material reflex. Black gives up a rook for a pawn, which on a simple count is losing the exchange and more. But the c6-pawn is the only reason White is playing for a win, and the knight on a6 cannot hold it alone. Once the rook is traded for the pawn, what remains is knight against rook, a balance in which material is far from decisive — because Black keeps pawns and an active king.

The king's timing is the fine point. After 40.cxd7, only 40...Ke7 draws: the king must reach the promotion square before the pawn does. The natural alternative 40...Kg7 loses, with the evaluation running to +11.28, because the new queen dominates the board and the knight on a6 never gets back in time. One move between a draw and a defeat.

It is worth saying what the engine does not say: finding 37...Re5+ with four minutes requires first seeing that the rook is good for nothing except dying for the pawn. That is a judgement call, not a calculation — the calculation after it runs three forced moves deep. Judgement is what was missing, and judgement is what can be trained.

How to train the mistake in this game

The pattern itself is well known and appears in any endgame collection: when the opponent has an advanced passed pawn and your rook is the piece assigned to stop it, trading the rook for the pawn is usually the solution, not a surrender. What makes the pattern fail in a real game is the moment the decision has to be taken — generally the move at which it still looks excessive.

The second habit is easier to train alone. Every time a passed pawn is born in your game, yours or your opponent's, stop and ask the basic question again: who is in a hurry now? In imported games that moment can be located precisely — it is almost always the move where the evaluation starts moving after a stable stretch.

It is also the lesson that separates this endgame from a tactics exercise. There was no combination to find; there was a choice between two plausible moves, one of which loses. Training calculation does not fix that kind of mistake; reviewing your own quiet positions does.

• Find, in your own games, the move where the evaluation leaves a stable band — that is where the decision was taken.
• In endgames with a passed pawn, calculate first the line where you give material to stop it; discard it only after seeing the result.
• Count king tempi before choosing a square: one move separated a draw from a loss in this game.
• Note when a mistake came with time on the clock: that separates a judgement problem from a pace problem.

Frequently asked questions

Update log

• 2026-09-07 — First publication, with both endgame lines checked against the engine.

Sources

• Global Chess League Day 1: Sindarov's American Gambits Begins With Double Win — Chess.com
• FYERS American Gambits - PBG Alaskan Knights: round one games — Lichess
• Geneva to host FIDE World Championship Match 2026 — FIDE

Related guides

• Five essential pawn endgames for breaking the 1800 barrier
• The chess fortress: how to save endgames when you are material down
• The Sicilian Defense for intermediate players: strategic plans and the tactics that decide games