Showing posts with label QGD Bf4. Show all posts
Showing posts with label QGD Bf4. Show all posts

Wednesday, 26 February 2025

World Senior Team Championship 65+ Round Nine

ENGLAND 3 faced Czech team ŠK Sokol Vyšehrad in this afternoon's final round.

Jiří Groh (1989) - Spanton (1896)
QGD Bf4
1.Nf3 d5 2.d4 e6 3.c4 Nf6 4.Bf4!?
This position, which goes back to at least the 1899 Janowski-Showalter match, occurs 1,1,28 times in ChessBase's 2025 Mega database, but is very much a sideline, although recently Nakamura has played it
*****
*****
*****
*****
*****
4...Bd6 5.Bxd6 Qxd6 6.e3 0-0 7.Nc3 c5 8.Be2 dxc4!?
This move in not in Mega25, but it is the top choice of Stockfish17 and Dragon1. Known continuations are 8...Nc6, 8...cxd4, 8...Bd7 and 8...b6.
9.Bxc4
The game has transposed to a position occurring 18 times in Mega25, but in 16 of them it is White to play.
9...a6 10.a4
The engines prefer 10.dxc5, 10.0-0 or10.Be2!?
10...Nc6 11.0-0
Now both sides have castled, how would you assess the position?
*****
*****
*****
*****
*****
Black's bishop is hemmed in by its own pawns, but the engines reckon that disadvantage is more than offset by Black's pressure against the white centre.
11...Rd8 12.dxc5 Qxc5 13.Qe2 e5?!
Freeing the bishop, but missing that White has sharp counterplay.
14.Ng5 Rf8
This withdrawal is more-or-less forced.
15.Rac1 Qe7 16.Rfd1?!
Another case of missing an opponent's fairly obvious reply. The black e pawn should be stopped from advancing, perhaps by 16.Nge4.
16...e4 17.Rd2?
JG picked up the rook, intending to move it to d4, but did not let go of the piece, it seems, and played the text instead.
17...h6
Forcing a combination that nearly always favours the side gaining two pieces for rook and pawn.
18.Nxf7
Worse is 18.Nh3? Bxh3 19.gxh3 Ne5.
18...Rxf7 19.Bxf7+ Qxf7 20.Rcd1?
Yet another case of missing a simple reply.
20...Bg4 21.f3 exf3 22.gxf3 Bh3 23.Nd5 Nxd5 24.Rxd5 Rf8 25.Rd6
The engines suggest 25.Rc5 or 25.a5, but agree Black is winning.
25...Qh5!?
This is stronger than grabbing the f pawn.
26.Qc4+ Kh8 27.Qd5 Ne5 28.f4!? Qg4+ 29.Kf2
How should Black proceed?
*****
*****
*****
*****
*****
29...Qh4+
This is fine, but immediately devastating is 29...Rxf4+! 30.exf4 Qxf4+, eg 31.Kg1 Nf3+ 32.Kf2 Nd4+ 33.Kg1 Qg4+ 34.Kf2 Qe2+ 35.Kg3 Qe3+ 36.Kh4 Nf5+, or 31.Ke2 Bg4+ 32.Ke1 Qe3+ 33.Kf1 Bh3+.
30.Kg1 Ng4?!
This is winning, but transposing to the previous note with 30...Qg4+ is better, and if White avoids that by playing 31.Kh1?!, then 31...Nf3 is decisive.
31.Qf2 Nxh2??
This lets White force queens off. Instead 31...Nf6 threatens 32...Bg4, when White apparently has nothing better than 33.Rc1, after which Black has many attractive continuations, eg 33...Bf5 34.Rd4 Re8 35.Qg3!? Qxg3+ 36.hxg3 Rxe3.
32.Qf2 Qxf2+ 33.Kxf2
How would you assess this ending?
*****
*****
*****
*****
*****
Bishop and knight are usually worth more than a rook in middlegames, but that is not always the case in endings, especially when, as here, Black has difficulties getting the minor pieces safely out of the enemy's position.
33...Bf5?
Black can hold the balance by hitting the d1 rook with 33...Bg4.
34.Rd8?
Winning is 34.Kg3 Ng4, and now 35.Rd8, when Black cannot meet White's threat to push the e pawn.
34...Kg8?!
Completely equal, according to the engines, is 34...Rxd8 35.Rxd8+ Kh7.
35.Kg3
Black to play and (probably) hold the draw
*****
*****
*****
*****
*****
35...Ng4!?
Possibly better is 35...Nf1+ 36.Kf2 Nh2, offering to repeat. White should avoid this with 37.Rxf8+ Kxf8 38.Kg3, but 38...Ng4 gives good drawing chances as 39.e4 Be6 40.f5 can be met by 40...Bc4, when 41.Kxg4?? loses to 41...Be2+. Instead the engines reckon 41.Rd8+ leaves White slightly better.
36.e4 Ne3
This is the best try as 36...Be6? can be met by 37.R1d6.
37.R1d3 Nf1+ 38.Kf2 Bxe4 39.R3d4
This is Dragon1's top choice, but Stockfish17 prefers first exchanging on f8.
39...Rxd8 40.Rxd8+ Kf7 41.Kxf1
How would you assess this rook-v-bishop ending?
*****
*****
*****
*****
*****
Dragon1 reckons Black is winning; Stockfish17 reckons the position is equal.
41...Ke6?!
The engines agree activating the king is a mistake. Instead they suggest 41...g5 or 41...Bc6, but disagree as to how the game stands.
42.Kf2 Bc6 43.a5 Kf5 44.Kg3 g6 45.Rf8+ Ke6 46.Kh4 Be4 47.Kg4 Kd5 48.Rf6 Kd4 49.Rb6 Ke3 50.Rb3+ Kd4 51.Rb4+ Ke3 52.Rb3+ Kd4 53.Rb4+ Ke3 54.Rb3+ ½–½
White should not have acquiesced to a draw, according to the engines, but are they right?
Position after 49...Ke3 (and 51...Ke3 and 53...Ke3)
*****
*****
*****
*****
*****
Best play, according to the engines, goes something like 50.Kg3 Kd3 51.Kf2 Bc6 52.b4 Ke4 53.b5!? Bxb5 (53...axb5? 54.Rxc6 bxc6 55.a6 etc) 54.Rxb7 Bd3 (54...Kxf4? 55.Rxb5!) 55.Rb4+ Kd5 56.Ke3 Bb5 57.Rd4+ Ke6 58.Rd1 Kf6 59.Rh1 Kg7 60.Rg1 Kf6 61.Rb1 Bc4 62.Kd4 Be2 63.Rb2 Bf1 64.Rb1 Be2 65.Ke4 Bc4 66.Rb6+ Kg7 67.Ke5 h5 68.Rb7+ Kh6 69.Rb8 Kg7 70.Kd4 Bf1, and so on, and so on. The engines are convinced White is winning (about +5.5, Stockfish17; about +8.5, Dragon1), but do not make progress.
So it seems the final position may be drawn after all, but White could have played on in the hope of inducing an error.
TEAM RESULT
Josef Jurek (IM 2160) 1-0 Brian Hewson (2039)
František Hoštická = (2059) Anthony Ashby (2049)
Jiří Groh (1989) = Tim Spanton (1896)
Karel Šlechta (1874) 0-1 Charles Higgie (1945)
ŠK SOKOL VYŠEHRAD 2-2 ENGLAND 3

Tuesday, 13 August 2024

Olomouc Round Five

Jaroslav Tryhuba (1867) - Spanton (1980)
QGD Bf4
1.d4 d5 2.c4 e6 3.Nc3 Nf6 4.Bf4!?
This was very popular in the early days of the Queen's Gambit Declined. Indeed, in ChessBase's 2024 Mega database, it precedes every continuation except 4.Nf3.
How should Black respond?
*****
*****
*****
*****
*****
4...c5!?
This is fifth-most popular, behind 4...Bb4, 4...Bd6, 4...Be7 and especially 4...c6.
5.e3 a6!? 6.Nf3 Nc6 7.a3 dxc4!?
Capturing on c4, before White has moved the king's bishop, is often frowned on, but here it is the most popular move, and also the top choice of Stockfish16.1 and Dragon1.
8.dxc5!?
After 8.Bxc4 b5, games in Mega24 feature the white king's bishop retreating to a2, b3, d3 and e2, with the engines reckoning Black is at least equal.
8...Qxd1+ 9.Rxd1 Bxc5 10.Bxc4 0-0
Probably not 10...b5?! as White has 11.Nxb5! axb5 12.Bxb5 with strong pressure.
11.0-0
Now both sides have castled, how would you assess the position?
*****
*****
*****
*****
*****
White is ahead on development, and at first glance Black might look in difficulties, but ...
11...b5
... gives the black light-square bishop access to the long light-diagonal with tempo.
12.Ba2!?
This may be a novelty. The engines prefer the known-move 12.Bd3 and also 12.Be2.
12...Bb7 13.b4 Be7 14.Rd2?!
The engines strongly dislike this, suggesting 14.Bd6.
14...Rac8
Even stronger, according to the engines, is 14...Rfc8!?, leaving the queen's rook to support the thrust ...a5.
15.Ne2!?
The engines fluctuate between this and 15.Nb1.
15...Rfd8 16.Rxd8+ Rxd8 17.h3?
White's last move is a mistake - can you see how Black exploits it?
*****
*****
*****
*****
*****
17...h6?
Missing the simple win of a pawn with 17...Rd3. White should have played 17.Rc1 (Stockfish16.1) or 17.Ned4 (Dragon1).
18.Ned4 Nxd4 19.Nxd4 Nd5 20.Bh2?
The engines suggest 20.Be5 or 20.Bxd5!?
20...Rc8?
Much stronger is 20...Nc3, eg 21.Bb1 Bf6 etc, or 21.Bb3 Bf6 etc.
21.Bb1?
The engines reckon 21.Be5 keeps White's disadvantage to a minimum.
21...Bf6?!
Probably stronger are 21...Nb6 and 21...Rc3.
22.Bd6
The engines suggest 22.Be4 or 22.Nb3, but give Black at least the upper hand.
22...Bxd4 23.exd4 Rc3 24.f4?
Thus turns a bad position into a more-or-less hopeless one. After 24.Be4 Rxa3 Black is clearly better, but there is a lot of play left.
24...Ne3
At first glance White's position may not look too bad, but it is almost resignable, even at club level
*****
*****
*****
*****
*****
25.Ne1 Nxg2 26.Be4?!
It is a sign of how bad Black's position is that the engines do not regard this as a mistake.
The game finished:
26....Nxe1 27.Bxb7 Rxa3 28.Kf2 Nc2 29.Bc5 Ra4 30.d5 exd5 31.Bxd5 Nxb4 0-1

Friday, 18 March 2022

Bad Wörishofen Round Seven

U2000 leaderboard pairings at the start of round seven:
tableTNrnameelopts-TNrnameelopts (w) -  (b)
114Bigott,Martin17475.5-1Spanton,Tim R19124.5 - 
212Abed,David17545-6Klein,Reinhard18744 - 
34Eschle,Hubert18764-41Fresow,Boris04 - 
420Steuer,Christopher16564-13Bail,Leonhard17594 - 


Martin Bigott (1747) - Spanton (1912)
Queen's Gambit Declined with Bf4
1.d4 d5 2.Nf3 e6 3.c4 Nf6 4.Nc3 Be7 5.Bf4 0-0 6.e3 c5 7.dxc5 Bxc5
If 7...Na6!?, intending ...Nxc5, my main analysis engines Stockfish14.1 and Komodo12.1.1 give 8.cxd5 Nxd5 9.Nxd5 exd5 10.Bxa6 Qa5+ 11.b4!? Qxa6 12.Nd4, when Black's bishop-pair gives some compensation for being a pawn down.
8.Be2!?
The main moves in ChessBase's 2022 Mega database are 8.Qc2 and 8.a3.
8...Nc6 9.0-0
MB offered a draw.
9...Re8!?
9...dxc4 and 9...a6 are the main moves in Mega22.
10.Bg5 Be7?!
Probably better is 10...dxc4, when the engines reckon White has a slight edge..
11.Bxf6 Bxf6 12.cxd5 exd5 13.Qxd5 Be6!?
13...Qxd5 14.Nxd5 Bxb2 15.Rab1 Be5 16.Nxe5 Rxe5 gives White the upper hand, according to the engines.
14.Qxd8 Rxd8 15.a3 Na5 16.Nd4?!
Giving back the pawn to simplify, and hopefully hold on for a draw. Best, according to the engines, is 16.Rfd1.
16...Bxd4 17.exd4 Nb3 18.Rad1 Nxd4 19.Bf3?!
Seeking more simplification, but it comes at some cost.
19...Nxf3+ 20.gxf3 Bb3 21.Rxd8+ Rxd8 22.Rf1 Ke8 23.Kg2
How should Black proceed?
*****
*****
*****
*****
*****
23...Rd6
I rejected 23...Rd2 because of 24.Re2, but the engines reckon 24...Rxe2 25.Nxe2 Ke7 is promising for Black (winning, according to Stockfish14.1; slightly better, according to Komodo12.1.1). However they also reckon 24.Re4!? Rxb2 25.Rb4 is dead-equal, albeit preferring Black after 24...b6 or 24...Be6.
24.Re4 Be6 25.f4?!
This seems weakening. Perhaps best is 25.Rb4 b6 26.Nb5 Rd7, which the engines reckon favours Black.
25...Rc6
The engines like 25...Bd7!?, eg 26.Re3 Bc6+ 27.f3 Rd2+ 28.Re2 Rd3 29.Rf2 Ke7, when Black is more active.
26.Kg3 Rb6 27.Re2 Bc4 28.Re4!?
The more modest 28.Rd2 may be better.
28...Rg6+
The engines' 28...Bf1!? could be best.
29.Kf3 Bf1
How should White proceed?
*****
*****
*****
*****
*****
30.f5!
Giving up a pawn, but only temporarily.
30...Bg2+ 31.Kf4 Bxe4 32.fxg6 Bxg6
Black has won a pawn, but after ...
33.Nb5 a6 34.Nd6 b5 35.b4 Ke7 36.Nb7 Ke6 37.Nc5+ Kd5 38.Nxa6
... White has won it back.
How would you assess this ending?
*****
*****
*****
*****
*****
Pawns are equal, but Black has bishop versus knight in an ending with rival pawn-majorities. Black also has the slightly more active king. Nevertheless the engines reckon the ending is dead-equal.
38...Bd3 39.Nc5 Bf1
39...Bc2 prevents 40.a4, but White has 40.Na6, and if 40...Kc4 then 41.Nc7.
40.a4 bxa4 41.Nxa4 Kc4 42.Nc5 Kxb4 43.Ne4 f6
There is no way for Black to get an advantage, according to the engines.
44.Nd6 Bd3 45.Ne8 Kc5 46.Nxg7 Kd5 47.Nf5
MB offered a draw.
47...Ke6 48.Ne3 h6 49.Ng4 h5 50.Ne3 Bg6 51.h4 Bd3 52.Nd1 Kd5 53.Ne3+ Kd4 54.Nf5+ Kd5 55.Ng7 Bg6 56.Nxh5!? ½–½

U2000 leaderboard pairings for round eight:
tableTNrnameelopts-TNrnameelopts (w) -  (b)
11Spanton,Tim R19125-12Abed,David17546 - 
213Bail,Leonhard17595-14Bigott,Martin17476 - 
322Klein,Robert04.5-4Eschle,Hubert18764.5 - 
441Fresow,Boris04.5-9Litfin,Boris18184.5 - 
516Kampfmeier,Dirk17334-11Kopischke,Maik18004.5 - 

Thursday, 24 June 2021

Opening Lessons From Crete III

IN round four I faced Tomasz Sielicki, a Pole listed in the tournament as having a Fide of 1938, although his rating had slipped in the June list to 1905.
The full game can be seen at https://beauchess.blogspot.com/2021/06/crete-round-four.html, but here I will look in some depth at the opening.

Sielicki - Spanton
QGD Bf4
1.d4 d5 2.c4 e6 3.Nc3 Nf6 4.Nf3 Be7
This is Black's most-popular reply, with more than 46,000 examples in ChessBase's 2021 Mega database. Other popular moves include 4...c6 (27,000+), 4...Bb4 (16,000+) and 4...Nbd7 (14,000+).
5.Bf4
This scores a very good 63% in Mega21, whereas the more-popular 5.Bg5 scores 58%. The analysis engines Stockfish13 and Komodo12.1.1 prefer the text, although it is often unwise to place much reliance on engine choices early in an opening unless there is a major tactical point.
5...0-0 6.e3
6.Nb5?! has been played in a handful of games in Mega21. The engines give the interesting, and apparently unplayed, 6...dxc4!?, the point being that 7.Nxc7? runs into 7...Bb4+, when 8.Nd2 e5! 9.Bxe5 Ne4 10.Bf4 Qxd4 wins for Black. In this line, 9.dxe5 is simply met by 9...Qxc7, when the e5 pawn is pinned. Also bad for White is 7.Bxc7? Bb4+ 8.Nd2 Qd7.
Also strong after 6.Nb5?! is 6...c6!?, which has similar ideas to 6...dxc4 and also does not appear in Mega21.
6...Nbd7
This is second in popularity to the more-aggressive 6...c5. The engines' top choice, but by a small margin over other moves, is 6...b6, which has been played by Magnus Carlsen.
7.a3
The main line in Mega21 runs 7.c5!? c6 8.Bd3 b6 9.b4 a5 10.a3 Ba6 11.0-0 Qc8, when the engines give White an edge.
7...a6!?
Easily most popular is 7...c5, when the commonest continuation is 8.cxd5 Nxd5 9.Nxd5 exd5 10.dxc5 Nxc5 with an IQP position that looks very reasonable for Black.
Anatoly Karpov has played 7...Nh5, which I rejected because after 8.Bg3 (Karpov's opponent preferred 8.Be5) I felt it would be dangerous for Black to follow up with 8...Nxg3 as 9.hxg3 half-opens the h file.
The text has been played by grandmasters, but may be too slow.
How should White proceed?
*****
*****
*****
*****
*****
8.Qc2?!
White is set up to play 8.c5, and probably should.
8...c5
Thematic in this type of position is 8...dxc4 9.Bxc4 b5, although Black would ideally wait for White to spend a tempo on Bd3 or Be2 before capturing on c4.
9.dxc5 Bxc5?!
Probably better is 9...Nxc5, when 10.Rd1 Qe8!? 11.cxd5 Nxd5 12.Nxd5 exd5 is only slightly better for White, according to the engines, the point being that 13.Rxd5 is met by 13...Be6!?, eg 14.Rxc5 Bxc5 15.Ng5 g6 16.Qxc5 Rc8 with an initiative for Black.
10.0-0-0!?
Stronger, according to the engines, is 10.cxd5 with probably short-castling to follow. However the position after the text is far from clear.
10...Nb6!?
This may be too defensive. The engines prefer 10...b5 or 10...dxc4.
Here the engines come up with a very aggressive continuation for White
*****
*****
*****
*****
*****
11.cxd5
The engines like 11.g4!?, when 11...Nxg4 12.Rg1 Nf6 13.Bh6 seems strong for White. Instead the engines reckon Black should probably play 11...Bd7, when they agree 12.g5 Nh5 13.Be5 is better for White.
11...exd5 12.Kb1
If 12.e4!? the engines like Black after 12...Bd7, the point being that 13.exd5?! Rc8 seems very good for Black.
12...Qe7
Getting out of a pin, but the engines prefer 12...Be6 or 12...Bd7.
13.Ng5?!
It was probably better to get on with development and connect the rooks with 13.Bd3.
The text threatens 14.Nxd5! Nbxd5 15.Rxd5, but that is easily parried.
13...h6 14.Nxd5?
TS suggested 14.h4!?, which is much better than the text, although the engines prefer Black after 14...Rd8.
14...Nbxd5 15.Rxd5 hxg5 16.Bxg5 Bd6!?
The bishop is not as safe here as it would be on b6 or a7.
17.Bd3
White has two pawns and an initiative, but that is not enough for a knight.

Monday, 14 June 2021

Crete Round Four

Weight
Friday: 80.4kg (177lb)
Saturday: 80.7kg (178lb)
Yesterday: 80.4kg (177lb)
Today: 80.2kg (177lb)

FACED a Pole today.
Tomasz Sielicki (1938 Fide) - Spanton (1998 ECF)
ACO World Senior Championship Group C (Fodele Beach, Crete) Round 4
QGD Bf4
1.d4 d5 2.c4 e6 3.Nc3 Nf6 4.Nf3 Be7 5.Bf4
This scores 63% in ChessBase's 2021 Mega database, while the more popular 5.Bg5 scores 58%. The analysis engines Stockfish13 and Komodo12.1.1 prefer the text.
5...0-0 6.e3 Nbd7 7.a3!?
This is second in popularity to 7.c5!?, which is preferred by Komodo12.1.1, whereas Stockfish13 vacillates between the two moves.
7...a6!?
This has been played by grandmasters, but may be too slow. Overwhelmingly more popular is 7...c5.
8.Qc2
White is set up to play 8.c5!?, and should do, according to the engines.
8...c5
Black in turn is set up to play 8...dxc4 9.Bxc4 b5 followed by a quick ...c5, although the engines are OK with the text.
9.dxc5 Bxc5!?
Possibly a novelty. Known is 9...Nxc5, which may well be better.
10.0-0-0!?
The engines reckon White should prefer the plan of exchanging on d5 followed by castling short.
10...Nb6
Somewhat defensive - the engines much prefer 10...b5!?
11.cxd5
The sharp 11.g4!? is strong, according to the engines.
11...exd5 12.Kb1
White is in massive trouble after 12.Nxd5? Nbxd5 13.Qxc5 Bf5.
12...Qe7
Getting the queen out of a pin while protecting the c5 bishop seems natural but the engines prefer developing the light-square bishop to e6 or d7.
13.Ng5!?
This threatens 14.Nxd5!, but the engines prefer development with 13.Bd3.
13...h6
How should White proceed?
*****
*****
*****
*****
*****
14.Nxd5?
Capturing on d5 does not work after 13...h6. As TS pointed out after the game, better is 14.h4!, when the engines slightly prefer Black, but the position is very sharp.
14...Nbxd5 15.Rxd5 hxg5 16.Bxg5 Bd6?!
The bishop is a little vulnerable here, despite being protected by the queen, so probably better is 16...Bb6 or 16...Ba7.
17.Bd3
White has two pawns and an attack for a knight, so Black has to be careful
*****
*****
*****
*****
*****
17...Rd8!
The engines agree this developing move, giving the black king luft, is best.
After 17...Be6 18.Bxf6 gxf6 (18...Qxf6?! 19.Rxd6 Ba2+ 20.Kxa2 Qxd6 21.Rd1!? favours White, according to Stockfish13, although Komodo12.1.1 reckons the position is equal) White gets play with both 19.Rh5 and 19.Qa4!?, although objectively Black is doing well.
I also rejected 17...Qe6, when White has chances after 18.Rd4, although again Black should come out on top.
18.Rd1 Be6 19.Bxf6
How should Black recapture?
*****
*****
*****
*****
*****
19...gxf6
I thought White would be fine after19...Qxf6?! 20.Bh7+ Kf8 21.Rxd6 Rxd6 22.Rxd6 Ba2+ 23.Kxa2 Qxd6 24.Be4, and the engines agree, but Black can seemingly improve with 22...g6!?, when the engines give 23.Bxg6 fxg6 24.Qc5, after which White has three pawns and the safer king for a bishop. Black is better but only has the upper hand, according to Komodo12.1.1 (Stockfish13 reckons Black is winning).
20.Rh5!? Bg4
This is winning, but the engines slightly prefer 20...Ba3!?, the point being that 21.bxa3 Qxa3 leaves White facing murderous threats, eg 22.Bh7+ Kg7!, when the best the engines come up with for White is 23.Rd4 Rxd4 24.exd4 Qb4+ 25.Qb2 Qxb2+ 26.Kxb2, but after 26...Rh8 White will be a bishop down with effectively no compensation.
21.Rh4 Bxd1 22.Qxd1
Black is winning but still has to survive something of a storm
*****
*****
*****
*****
*****
22....Qe5?!
This is probably good enough but the engines much prefer 22...f5 or 22...Bd7.
23.Rh5
If 23.Qg4+, then not 23...Qg5?, after which 24.Qh3 Kf8 25.Rh5! Qg7 26.Rh8+ Ke7 27.Rh7 Qg8 28.Bc4 gives White an edge, according to the engines, but instead the simple 23...Kf8 seems to get Black through the complications with a winning advantage, eg the engines' main line runs 24.Rh5 Qe6 25.Bf5 Qd5 26.Bc2 f5! 27.Rxf5 Qe6, when Black is more-or-less out of the woods and is up a rook for three pawns.
23...Qe6 24.Bf5 Qc4 25.Bd3 Be5! 26.Rh8+!?
A last desperate throw of the dice.
Position after 26.Rh8+!?
*****
*****
*****
*****
*****
26...Kxh8??
Winning is 26...Kg7 27.Rh7+ Kf8.
27.Qh5+ Kg8 28.Bxc4 Rd7
Or 28...Rf8 29.Qg6+ Kh8 30.Qh6+ Kg8 31.Bd3 and mates.
Objectively best, according to the engines, but hopeless, is to allow White to capture on f7 with check, eg 28...Rd2 29.Qxf7+ Kh8 30.Qh5+ Kg7 31.Qg4+ Kh8 32.f4 Bxb2 33.Bf7 Bxa3 34.Qg6, but mate follows.
29.Qg4+ 1-0