Sindarov beats Firouzja: the refused queen trade and a clock spent on four moves

Short answer

Javokhir Sindarov beat Alireza Firouzja with White on the icon board on day five of the Global Chess League 2026, and the win decided FYERS American Gambits' 8-5 over Triveni Continental Kings. The game was not decided by an attack: Sindarov played a pawn down from move 11 with sound compensation, the position reached 0.00 on move 18, and Firouzja, a pawn up and short of time, declined the queen trade with 18...Qd8, which Stockfish 19 measures at +4.33. This piece reads the clock move by move, measures the decisive moments with the engine and compares our evaluations with the broadcast's automatic annotation.

Event: Tech Mahindra Global Chess League 2026

How this was compiled: Game from round twenty, played on 9 September 2026 in Bengaluru. Clocks read move by move from the official broadcast PGN; evaluations measured by us with Stockfish 19 at depth 24 up to move 19; from move 20, with a limit of 60 seconds per position, the analysis reached depths between 18 and 24, and the figures from that stretch are rounded to one decimal place.

The essentials in a few lines

• Firouzja declined the queen trade with 18...Qd8 in a position evaluated at 0.00; 18...Qxb4 kept the balance, and the move played goes to +4.33 after 19.Ne5, at depth 24.
• It was Sindarov who gave up material: 11...Bxb2 is the engine's preferred move, and White played a pawn down with the evaluation in their favour, between +0.2 and +1.3, from move 11 to move 15.
• Firouzja spent 870 of his 1,200 seconds on four moves — 440 on 10...Qe6 alone — and played his last thirteen moves with less than a minute and a half, in a time control with no increment before move 41.
• Sindarov also went wrong after getting the advantage: 27.Rb4 dropped the evaluation from +7.0 to +3.7, and Firouzja handed it all back with 27...Rc8, played with 7 seconds left.
• The broadcast's automatic annotation and our measurement disagree: it gives +3.00 after 18...Qd8 and +1.62 after 23.a6; Stockfish 19 gives +4.33 and +5.2.

What happened: a level position decided on move 18

The game was the icon board of round twenty, between FYERS American Gambits and Triveni Continental Kings, and it decided the match on its own: the other five boards were drawn, and Sindarov's win with White provided the 3 game points that made it 8-5. It was Sindarov's second win in a row in the league — in round 16 he had beaten Viswanathan Anand with Black — and it took the Gambits to the top on game points at the end of day five. In the first half of the round-robin, in round four, the two had drawn.

The summary of the game that has circulated speaks of a fearless attack and a positional sacrifice. The PGN and the engine tell three different things, each with its own section below: the pawn Sindarov gave up on b2, with sound compensation but no attack; Firouzja's clock, spent on four decisions; and move 18, where a refused queen trade turned 0.00 into +4.33.

The evaluations published here were measured by us, not copied from the broadcast. Lichess's automatic annotation marks 18...Qd8 at +3.00 and 23.a6 at +1.62; Stockfish 19, at depth 24, gives +4.33 and +5.2. The order of magnitude on move 18 is the same, but the reading of move 23 changes: by our measurement, White's advantage was never at risk after move 19.

The b2-pawn: it was Sindarov who gave up material

The game began as a Symmetrical English in which Black exchanged in the centre and fianchettoed the king's bishop: 1.Nf3 Nf6 2.c4 c5 3.g3 Nc6 4.Bg2 d5 5.cxd5 Nxd5 6.Nc3 g6. The first departure from the usual pattern was 7.h4, met by 7...h6, and the first long think was Sindarov's, two minutes on 8.O-O. The position came out of the opening with no advantage for either side: after 10.d3, the engine gives 0.00.

Then came the decision that shaped the rest of the game. Firouzja spent 440 seconds — seven minutes and twenty, more than a third of his time for the whole game — on 10...Qe6, which the engine rates at +0.43, against −0.11 for the simple 10...O-O. The move prepares the capture on b2, and Sindarov allowed it: 11.Be3 Bxb2 12.Rb1. A pawn down, White had the b-file, the g2-bishop on the long diagonal and every piece developed, and the evaluation stayed on their side.

The next mistake was a retreat. Instead of 12...Bd4, which held +0.41 and exchanged White's best bishop, Firouzja played 12...Bg7 and the evaluation rose to +1.05. Sindarov gave some of it back with 15.a4, which drops from +1.19 to +0.18 — the engine wanted 15.Nxd4 cxd4 16.Qb2 — and after 17.Bxd4 cxd4 the position was back to exactly zero. So far there is no fearless attack and no positional sacrifice: there is a pawn given for sound compensation and two sides making small inaccuracies.

The clock: 870 seconds on four decisions

The figures come from the broadcast PGN, which records the clock after every move. The Global Chess League plays 20 minutes with a 2-second increment only from move 41, and this game ended on move 34. In practice both players had exactly 20 minutes for the whole game, and every second spent on one move came out of the budget for all the others.

Sindarov spread his time differently. His single biggest think was 244 seconds on 22.Qf4, with the game already decided on the board, and the second was two minutes on 8.O-O. On 18.Qb4, the move that offered the queen trade, he spent no time at all — the clock shows zero, which points to a prepared move or a premove. Firouzja, on the next move, thought for 200 seconds.

Comparing it with the Anand–Carlsen game we analysed from day three helps separate the cases. There, a player lost on time in a level position, with increment, by spending more than it move after move. Here there was no increment to manage: there were four long decisions taken too early, and the most expensive of them ended in a losing move. Sindarov summed it up after the game: “Alireza spent too much time to think and missed some tactical tricks.”

18...Qd8: the queen trade refused

The diagram shows the moment when the game asked for a decision about its character, not a calculation. Black is a pawn up, on d4, and White offers to exchange queens. Accepting leads to an ending with rooks and bishops in which Black's extra pawn is worth little — the engine gives 0.00 — and in which the player with more time on the clock, Sindarov, would have the more comfortable game. Declining keeps the queens and the tension, but requires finding a square for the black queen.

The three moves the engine considers are 18...Qxb4 (0.00), 18...O-O (+0.32) and 18...b6 (+0.82). Firouzja chose a fourth, 18...Qd8, and it has a concrete flaw: the queen leaves the defence of e5 and the b5–e8 diagonal. 19.Ne5 attacks the g4-bishop and opens the g2-bishop's diagonal; the capture 19...Bxe5 is almost forced, and 20.Qb5+ wins the piece back with check, because the black king has not castled. After 21.Qxe5, the threat on h8 forces 21...f6, and the h8-rook is shut in behind its own pawn.

That is the opposite of the reading that circulated about the game. Sindarov did not need to attack fearlessly or force files open; he offered a simplification at a moment when his opponent had less time and an extra pawn to defend, and waited. Firouzja's mistake is a very common one in club games: declining the trade that draws because the extra material seems to demand more, without the time to calculate what the refusal costs.

After move 19: mistakes on both sides, and an advantage that never left White

The broadcast's automatic annotation marks 23.a6 as a blunder, with the evaluation falling to +1.62, and 23...b5 as a blunder in the other direction. Our measurement tells a less dramatic story. With 23.a6, White fixes a pawn on a6 and prepares the queen's entry on c7, but misses 23.Rb5, which attacks the d4-pawn along the fifth rank; the loss is a little over a pawn of evaluation, in a position that remains winning. In one single-line analysis the position after 23.a6 reached +3.49; in the three-line analysis we use, it stood at +5.17 — the difference is one of measurement, and both leave White winning.

Firouzja played 23...b5 after 32 seconds and was left with 45 on the clock. The move takes the b7-pawn out of the rook's reach, but opens the seventh rank for 24.Qc7, precisely the square 23.a6 had prepared, and the evaluation rises to +7.1. The engine preferred 23...e5, which held +5.2 and closed the g2-bishop's diagonal. With under a minute for eleven moves, the choice between the two was not a matter of calculation: it was reflex.

On move 25 Sindarov was imprecise once more — 25.Rb3, after 26 seconds, gives +6.1 where 25.Qxa7 gave +8.0 — and on move 27 came White's biggest mistake after the turnaround: 27.Rb4, after 59 seconds, dropped the evaluation from +7.0 to +3.7, because it allowed 27...Rb6! 28.Bc6 Bc8 29.Bxb5 Bxa6, with the black rook active and the a6-pawn gone. Firouzja, with 23 seconds, played 27...Rc8 in 7, and after 28.Rxc8 Qxc8 the evaluation was back at +7.4. The game ended after 34.Qc5, with the white pawn on a7 about to queen, Firouzja on 6 seconds and the position at +13.7; the broadcast PGN records 1-0 without saying whether it was a resignation or a flag fall.

Three principles for playing with compensation, and one for the side with the material

The first principle applies to whoever plays Sindarov's side. A pawn down with developed pieces and an open file is a legitimate position when the evaluation confirms it, and the way to check that after the game is to look at the number, not the piece count. From move 11 to move 15, White was a pawn down with the evaluation in their favour on every move.

The second applies to whoever plays Firouzja's side. Declining a trade that draws is a decision that needs a concrete reason in the position — a plan, a better square for the queen — and not the feeling that extra material obliges you to play for a win. When that reason does not appear within two or three minutes, the trade is usually the right choice, especially with less time than your opponent.

The third is arithmetic. With no increment, time spent in the opening is deducted from the middlegame, not from nowhere. Firouzja reached move 18, the most important moment of the game, with 8 minutes 32 seconds, and after deciding had 5:12 for the next fifteen moves. The game shows that clock management is not an endgame topic: it starts on move 10.

• Compensation for a pawn is checked in the evaluation: if it stays in your favour, the pawn was well invested.
• Faced with an offered trade, ask first what the position is worth after it, and only then what you give up by accepting.
• Without increment, treat the clock as the budget for the whole game: seven minutes on move 10 come out of move 20.
• The player with more time can offer simplifications on purpose: the decision weighs more on the one with less.

Frequently asked questions

Update log

• 2026-09-10 — First publication, with clocks read from the round-twenty PGN and evaluations measured at depth 24.

Sources

• FYERS American Gambits - Triveni Continental Kings: round-twenty games, with clocks and evaluations — Lichess
• Global Chess League 2026 Day 5: Sindarov Beats Firouzja To Give American Gambits Lead — Chess.com
• GCL 2026, Day 5: Gambits beat Kings, Pipers defeat Mumba Masters as race to the final heats up — FIDE

Related guides

• How to avoid time trouble in chess: five rules for the last three minutes
• A practical guide to the English Opening: controlling the board with 1.c4 without memorizing endless lines
• How to calculate candidate moves in chess: checks, captures and threats before you commit