Miniatures 45

IN this occasional series I am going through my decisive games of 20 moves or fewer.

Spanton (164 BCF) - D Dane (130 BCF)

Middlesex League 1992

Scotch Gambit

1.e4 e5 2.Nf3 Nc6 3.d4 exd4 4.Bc4!?

This scores two percentage points more than 4.Nxd4 in ChessBase's 2026 Mega database.

How should Black respond?

*****

*****

*****

*****

*****

4...Bb4+!?

This is only fourth-most popular in Mega26, but is liked by Stockfish18 and Dragon1 along with the much more common 4...Nf6.

5.c3 dxc3 6.0-0!? Qe7?

Probably best is 6...Nf6, while 6...cxb2!? 7.Bxb2 seems to give White plenty of compensation for two pawns.

7.a3 Ba5

The engines suggest 7...c2!? 8.Qxc2 Bd6, but much prefer White.

8.b4 Bb6 9.Nxc3 Nf6?!

Apparently a novelty. The engines prefer 9...d6, but agree White is winning.

10.Bg5 0-0 11.e5!? Nxe5 12.Nd5 Nxf3+

This seems best, but is losing.

13.Qxf3 Qe5

The engines give 13...Nxd5!? 14.Bxe7 Nxe7, but Black does not have enough for a queen.

14.Nxf6+ gxf6 15.Bxf6 1-0

LESSON: old gambits still pack a punch if the defender errs quickly.

Winning Chess

GRANDMASTER Keith Arkell won the British 65+ championship, scoring +6=1-0, finishing a point clear of the field.

Arguably the key game came in round six, where he was Black against Welsh international master John Cooper.

Cooper (2252 ECF/2194 Fide) - Arkell (2512 ECF/2390 Fide)

English Symmetrical

1.c4 c5 2.Nc3 g6 3.e3 Nf6!?

The point of playing this, rather than the 'natural' 3...Bg7, is so ...

4.d4 cxd4 5.exd4

... can be met by:

5...d5

Nevertheless, after 3...Bg7 4.d4 cxd4 5.exd4, Black can still play 5...d5!?, meeting 6.Nxd5 with 6...Be6 or 6...e6.

6.cxd5 Nxd5 7.Bc4 Nb6 8.Bb3 Bg7

The position is reminiscent of a Tarrasch Defence to the Queen's Gambit, but with colours reversed

*****

*****

*****

*****

*****

9.d5!?

The main move in ChessBase's 2026 Mega database is 9.Nf3, but it scores only 49%, whereas the rare text has racked up an impressive 61%.

9...0-0 10.Nge2 N8d7 11.Be3 Nf6

Stockfish18 and Dragon1 suggest 11...Ne5!?

12.0-0

Now both sides have castled, how would you assess the position?

*****

*****

*****

*****

*****

White's isolani has gained space in the centre, and does not look vulberable, which may be why Dragon1 gives White a slight edge. Nevertheless, it is an isolated pawn, and it is not obvious how White can use the pawn for attacking chances, eg the pawn has not created outposts for White's knights, except, perhaps, at d4. Stockfish18 calls the game equal.

12...Bd7 13.h3!?

This may be a novelty. The known move is 13.Re1.

13...Nc8!?

Heading for the blockading square d6.

14.Nd4 Nd6 15.Qe2 a6 16.a4

The engines reckon White has a slight edge after 16.Rad1 or 16.Rfe1.

16...Qa5 17.Rfd1 Rfe8 18.Bf4!? Rac8 19.Be5 Nfe4!? 20.Bxg7 Nxc3 21.bxc3 Kxg7 22.c4 Qc5!? 23.a5 Kg8!?

This retreat - perhaps withdrawal is a better word - is the engines' top choice.

24.Rac1?!

The engines do not like this pawn sacrifice

24...Qxa5 25.Ba2?!

This looks rather slow. The engines suggest 25.Qe3 or 25.Nf3.

25...Ba4 26.Bb3 Bxb3 27.Nxb3 Qb4 28.Nd2

It is hard to see what White has for a pawn

*****

*****

*****

*****

*****

28...b5!? 29.Rb1 Qc5 30.cxb5 axb5 31.Nb3 Qc4 32.Qe5 Qe4 33.Qg5 Qf5 34.Qe3 Qe4 35.Qg5 Rc4 36.Qd2?!

Probably better is the engines' 36.Na5, although Black remains on top.

36...Rc2 37.Qg5 Qe2 38.Qf4 Rb2 39.Nd4 Rxb1 40.Nxe2?

Better is 40.Rxb1, although 40...Qa2 gives Black the upper hand, according to the engines. The text wins the white queen, but after ...

40...Rxd1 41.Kh2 Rxd5

... Black has two rooks and two pawns for it.

0-1

Multi-Millionaire!

AT some point today my blog passed two million hits, and, as I type this, has reached 2,051,031 hits.

However, there is apparently no way of discovering how many of these views are by humans, and how many by bots.

Nevertheless, the figure looks impressive, especially when compared with the 183,062 hits for my other main blog, 10mm and all that!

And my third blog, Chess Tip Of The Day, which is really just a feeder blog for the PC/laptop version of Beau Chess, has just 18,561 hits, albeit it has been going for a much shorter period.

Summing Up The British

MY score in the 65+ of +3=2-2 gained 7.8 ECF elo but lost 13.2 Fide elo.

Euros In Crete

THE chess-results.com pages for the European senior team championships in Crete later this month have been updated, showing there are 18 teams in the 65+, including Ageless of Crete, and 14 teams in the 50+, including Austria's catchily named Data Technology Pressbaum/Eichgraben.

British 65+ Round Seven

Spanton (1940 ECF/2003 Fide) - Mark Page (2123 ECF/2050 Fide)

Sicilian Nimzowitsch

1.e4 c5 2.Nf3 Nf6 3.e5

This and 3.Nc3 are almost equally popular, but the text scores five percentage points more in ChessBase's 2026 Mega database.

3...Nd5 4.Nc3 e6 5.Ne4

This is a reasonably common alternative to the mainline 5.Nxd5 exd5 6.d4.

5...f5 6.Nc3 Ne7

Nimzowitsch played 6...Nxc3 in the stem game, a loss to Yates at Carlsbad 1929, but the text, which can perhaps be regarded as the chief modern continuation, scores a fabulous 61% in Mega26.

How should White proceed?

*****

*****

*****

*****

*****

7.d4

Stockfish18 and Dragon1 like the text and 7.Nb5!?

7...cxd4 8.Nb5!? Nec6!? 9.Nbxd4

The engines are not keen on this, more-or-less settling on 9.a4!? or 9.Be2.

What should Black play?

*****

*****

*****

*****

*****

9...Nxd4

The engines reckon Black is slightly better after 9...Nxe5!? 10.Nxe5 Qa5+ 11.c3 Qxe5+, a position reached in the game.

10.Nxd4!?

10.Qxd4 avoids transposing into the previous note, and is slightly preferred by the engines.

10...Qa5+ 11.c3

This is almost certainly best.

11...Qxe5+ 12.Be2

How would you assess the position?

*****

*****

*****

*****

*****

Black is a pawn up and has the only pawns on the two centre files. But White has an edge in development, with the promise of harassment of the black queen to come. The engines reckon Black is slightly better (Stockfish18) or at least has the better part of equality (Dragon1).

12...a6

But the engines reckon this is unnecessary, preferring development with 12...Be7 or 12...Nc6.

13.0-0 Nc6 14.Nf3?!

Avoiding an exchange of knights, while hitting the black queen. But the engines strongly dislike it, claiming 14.Re1!? gives White the upper hand, one point being 14...Nxd4?? loses to 15.Bh5+.

14...Qf6!? 15.Bg5 Qf7

If 15...Qg6?!, then 16.Re1 is difficult to meet, eg 16...Be7 17.Bxe7 Nxe7 (17...Kxe7!? might be better) 18.Ne5 Qf6 19.Bh5+ g6 20.Bf3, with good play for a pawn.

16.Re1 Be7 17.Bxe7 Qxe7 18.Nd4!? 0-0

Now both sides have castled, how would assess the position?

*****

*****

*****

*****

*****

Black remains a pawn up, with the only central pawns, and has almost caught up in development, but has a bad bishop. The engines more-or-less agree Black has a slight edge.

19.Bf3 Nd8!?

The engines like this, along with 19...Qd6.

20.Qc2 Qc5 21.Rad1 Nf7 22.b4 Qc7 23.c4 Ng5 24.Qd3 Rb8 25.h4!?

This is the engines' top choice ...

25...Nf7

... and they agree with not capturing on f3.

26.c5 g6?!

Reinforcing f5, to make ...e5 possible, but the engines prefer 26...b6 or 26...Ne5.

How can White best exploit Black's last move?

*****

*****

*****

*****

*****

27.Nb3

The engines reckon White has much more than enough compensation for a pawn after 27.Qe3!?, eg 27...b6 28.c6!? f4 29.Qa3 e5 30.b5! exd4 31.Rxd4 d6 32.Rxf4, with what they agree is a winning attack.

27...b6 28.cxb6?!

This is probably the wrong idea. Instead White is slightly better after something like 28.Qc3, according to the engines.

28...Rxb6 29.Rc1 Qb8 30.Nc5?!

Probably better is 30.a3.

30...Ne5 31.Qd2 d6 32.Rxe5?

This is not good. The engines give 32.Nd3!? Nxf3+! 33.gxf3, but award Black the upper hand after both 33...e5 and 33...Qb7, but note that 32...Nxd3 33.Qxd3 Rxb4, winning a second pawn, is problematic after 34.h5, eg 34...e5 35.Qd2 Be6 36.Qh6 Rf6 37.Rc6, with equal chances, according to the engines.

32...dxe5 33.a3 Rd6

Black is untying his position, emerging up the exchange and a pawn

*****

*****

*****

*****

*****

34.Qg5 Qa7 35.h5 Qg7 36.hxg6 hxg6 37.Re1 e4 38.Be2 Qf6 39.Qe3 Kg7 40.f3!? exf3 41.Bxf3 Rfd8 42.Bb7!? Kf7?!

There is nothing wrong with 42...Bxb7, as 43.Nxb7 is simply met by 43...Rd3.

White to play and seriously reduce Black's advantage 

*****

*****

*****

*****

*****

43.Qh6

Missing 43.Bxc8 Rxc8 44.Ne4!, after which 44...Qd4 45.Nxd6+ Qxd6 leaves Black a pawn up and, according to the engines, winning, but the position is far from trivial.

43...Qd4+ 44.Qe3 Qxe3+ 45.Rxe3 Bxb7 46.Nxb7 Rd1+ 47.Kh2 R8d3

But this position is trivially winning.

0-1

British 65+ Round Six

Paul Kenning (1831 ECF/1778 Fide) - Spanton (1940 ECF/2003 Fide)

Vienna Game

1.e4 e5 2.Nc3 Nf6 3.d4!?

This early central advance, reminiscent of the Centre Game (1.e4 e5 2.d4) scores 59% in ChessBase's 2026 Mega database, the same as the main move 3.f4, and up to 11 percentage points better than the other popular continuations, 3.Bc4, 3.Nf3 and 3.g3.

3...exd4 4.Qxd4 Nc6 5.Qd3!?

This is the main continuation, upping White's score in Mega26 to 62%.

How should Black respond?

*****

*****

*****

*****

*****

5...Bb4

This is the commonest continuation in the Centre Game, after the moves 1.e4 e5 2.d4 exd4 3.Qxd4 Nc6 4.Qe3 Nf6 5.Nc3, and it is also the commonest continuation in this line of the Vienna, but after it White's score in Mega26 climbs to 67%. Stockfish18 and Dragon1 suggest 5...Bc5 or 5...d5!?, one point being 5...d5!? 6.exd5 can be met by 6...Nb4.

6.Bd2 0-0 7.Nge2!?

Stockfish18 marginally prefers the much more common 7.0-0-0, but Dragon1 fluctuates between the two moves.

7...Re8 8.0-0-0

Now both sides have castled, how would you assess the position?

*****

*****

*****

*****

*****

White has the only pawn on the fourth rank, but, as is often the case with opposite-side castling, the position is both tricky to assess and to play. The engines reckon chances are equal.

8...Qe7?!

But not anymore! The engines suggest 8...Ne5 or 8...d6.

9.Bg5!

Moving the bishop for the second time, but it is worth it as the pin is hard to meet.

9...Ne5?!

Apparently a novelty, and probably not a good one, although it does come to be the engines' second choice, at least for a while, behind getting out of the pin immediately with 9...Qe6, after which the engines suggest another apparent-novelty in 9...f3!?, claiming White has at least the upper hand, a sample line running 10.Bh4 Ne5 11.Qd4!? Ba5 12.Qa4!? Qb6, reaching a position where the engines disagree as to how White should proceed, but agree White is winning.

10.Qg3

White's main threats are 10.f4 and 10.Nd5

*****

*****

*****

*****

*****

10...Bxc3!?

The engines suggest 10...c6, but reckon 11.a3!? (as well as 11.f4) is very strong, one line running 11...Bxa3!? 12.Bxf6!? Bxb2+ 13.Kxb2 Qxf6 14.Nd4, when Black does not seem to have enough for a knight. Alternatively, 11...Bc5 12.f4 Neg4!? 13.e5 d5!? 14.Bxf6!? Nxf6 15.exf6 Qxf6 also leaves Black significantly down on material.

11.Nxc3 Qe6 12.f4!?

This seems even stronger than 12.Bxf6 Qxf6 13.Nd5 Qd8 14.f4.

12...Nc6 13.e5 Nh5 14.Qf3 h6!?

Giving up a pawn to keep the game going, as 14...g6?! 15.g4 Ng7 16.Nd5 is horrendous.

15.Qxh5 gxh5 16.Qxg5

How big is White's advantage?

*****

*****

*****

*****

*****

White is roughly the equivalent of being up a piece and a pawn, according to the engines.

16...Nb4!?

If 16...d6, then 17.Nd5.

17.Kb1 d6 18.a3 a5!?

Desperation, but the game is lost anyway.

19.axb4 axb4 20.Nd5 dxe5 21.Qxe5 Qd7

White to play and win

*****

*****

*****

*****

*****

22.Nf6+?

White has several strong moves, the most convincing being 22.Qg5 and 22.Bb5!?, eg 22.Qg5 Qa4!? 23.Nf6+ Kf8 24.Qc5+ Re7 25.Rd8+ Qe8 26.Rxe8#, or 22.Bb5!? c6 23.Nb6 Rxe5 24.Rxd7 Bxd7 25.fxe5.

22...gxf6 23.Rxd7 fxe5 24.Rxc7

PK offered a draw.

Objectively, should Black accept?

*****

*****

*****

*****

*****

Stockfish18 reckons the game is completely equal after 24.Ra5 or 24.exf4. Dragon1 is less convinced, but reckons White has no more than the better part of equality.

½–½

British 65+ Round Five

Spanton (1940 ECF/2003 Fide) - Francis Best (1893 ECF/1813 Fide)

Scandinavian

1.e4 d5 2.exd5 Qxd5 3.Nc3 Qd6 4.d4 Nf6 5.Nf3 c6 6.g3!?

There are 1,052 examples of this move in ChessBase's 2026 Mega database, but the main continuation remains 6.Ne5.

How should Black respond to the text?

*****

*****

*****

*****

*****

6...Bf5

This has been played by many grandmasters, but the main line in Mega26 runs 6...Bg4 7.Bg2 e6 8.0-0 Be7 9.Bf4, when Black usually chooses between the aggressive 9...Qb4 and the passive 9...Qd8 (3...Qd6 guru Tiviakov has played both moves), with Stockfish18 and Dragon1 reckoning White is clearly better in either case.

7.Bg2 e6 8.0-0 Qc7!?

Prophylaxis, the idea being to meet 9.Bf4 with 9...Bd6, but Tiviakov has preferred the more-common moves 8...h6 and 8...Be7.

9.Re1!? Be7!?

The problem with 9...Bd6, after 9.Re1, is White has 10.d5!?, when 10...Nxd5 11.Nxd5 cxd5 12.Nd4 obliges Black to allow 12...0-0 13.Nxf5, as 12...Bg6? can be met by 13.Nb5 Qe7 14.Bxd5!, or 13.Nxe6!? fxe6 14.Rxe6+, with the engines reckoning White is winning comfortably in both lines. Probably even worse is 10...cxd5?! 11.Nd4.

10.Bf4 Qd8

This position occurs 22 times in Mega26, but in all but one game it is Black to move

*****

*****

*****

*****

*****

11.Ne5 0-0?!

The engines suggest 11...Nbd7.

How should White proceed?

*****

*****

*****

*****

*****

12.g4!

This type of move is often double-edged, but here it definitely works.

12...Bg6 13.h4 h5?!

The engines prefer 13...Nd5!?, but after 14.Nxd5 cxd5 15.h5 Be4 16.Bxe4 dxe4 17.Rxe4 White has won a sound pawn.

14.g5

There are other good continuations, including 14.Nxg6 fxg6 15.Qe2!?

14...Nd5 15.Nxd5 exd5 16.Bf3

Even stronger, according to the engines, is 16.Nxg6 fxg6 17.Bg3.

16...Nd7 17.Bxh5

Stockfish18 again likes Nxg6, but Dragon1 prefers the text, at first, albeit eventually evaluating the two moves as roughly equally good.

17...Nxe5 18.dxe5

And here the engines are unsure whether the text or 18.Bxe5 is better. Note that 18.Bxe5 cannot be met by 18...Bxg5? as White has 19.Bxg6!, before capturing on g5.

18...Bf5 19.Bg4 Qd7 20.Bxf5 Qxf5 21.Bg3

With the tactics having settled down, how big is White's advantage?

*****

*****

*****

*****

*****

Black has the better bishop and the safer king, but White has an extra pawn, and chances of using the white majority for an attack on Black's king. The engines agree White is winning.

21...Bc5

21...Qh3 is not dangerous for White after 22.Qf3.

22.Kg2 Rae8

The engines suggest plunging in with 22...f6!? 23.exf6 gxf6, but it is far from clear whose king is more vulnerable.

23.Qf3!?

The engines like 23.Qd3, not being worried about the nuisance-value of 23...Qxd3 24.cxd3 Bd4.

After 23.Qf3!?, can Black get his pawn back?

*****

*****

*****

*****

*****

Yes! The point being 23...Qxc2!? 24.Rac1 Qd2!? saves the black bishop. However, after 25.Red1 Qa5 26.a3 Bb6, White has gained valuable time, and the engines reckon 27.h5 leads to a winning kingside attack.

23...Qxf3+?!

Nevertheless, the engines agree Black should keep queens on, one way or another, eg by 23...Qe6 or 23...Qg6.

24.Kxf3 g6!? 25.Kg4 Kg6 26.Rad1 Re6 27.f4 Rh8 28.Rh1 Ree8 29.Rde1 Bb6 30.Re2 d4!?

Going for counterplay. It seems unlikely that passively defending on the kingside would work in the long run.

31.Rd1 c5 32.b3 Ba5 33.a4 a6 34.f5!?

This lets Black split White's majority, but in return the white king advances, and the h pawn becomes passed.

34...gxf5+

This is best, according to the engines.

35.Kxf5 Re6 36.Rf2 Bc7 37.Re1 Rhe8

Not 37...Rh5? 38.Kg4 Rh8 39.Rhf1.

38.Re2 b5 39.axb5 axb5 40.h5 c4 41.bxc4 bxc4 42.Rd1 Bb6 43.Re4

Not 43.Bf2? d3!

43...Rh8 44.h6+ Kh7 45.Bf2 d3 46.Bxb6 dxc2 47.Rc1 Rxb6 48.Rxc2

Black is losing a second pawn

*****

*****

*****

*****

*****

48...Re8 49.Rcxc4 Rbe6 50.Rc7 Kg8 51.Rd7 Rg6 52.Rd6 Rge6 53.Rxe6 Rxe6 54.Rd4 1-0

Outdoor Chess

 

The British Chess Championships attract players at all levels

British 65+ Round Four

Chris Bellers (1916 ECF/1919 Fide) - Spanton (1940 ECF/2003 Fide)

English

1.c4 e5 2.g3 Nc6 3.Nc3 g6 4.Bg2 Bg7 5.e4!?

Aiming for a Botvinnik setup, as recommended in Tony Kosten's The Dynamic English.

5...d6 6.Nge2

Nimzowitsch pioneered the V formation (e4-d3-c4) in the English, but developed the king's knight to f3. Botvinnik played the same V formation as Black against the Closed Variation of the Sicilian, but refined it by developing the king's knight to e7 rather than f6.

6...Nge7 7.0-0 0-0

Now both sides have castled, how would you assess the position, which occurs 779 times in ChessBase's 2026 Mega database?

*****

*****

*****

*****

*****

White's grip on d5 is generally regarded as at least making up for the hole at d4, and Stockfish18 and Dragon1 give White a slight edge. Note that if the white c pawn were on c2, and White had instead played d3, the position would be symmetrical. 

8.d3

Now we have reached a position occurring 2,035 times in Mega26, with White scoring 53%.

8...f5!?

The main line in Mega26 runs 8...Be6 9.Nd5 (Kosten recommends holding back on this move until Black has played ...Be6, after which Black can no longer reply ...Nxd5 - at least not while there is a knight on c6 as well as a bishop on e6) Qd7 10.Be3 f5 11.Qd2, with a slight edge for White, according to the engines.

9.Rb1 Be6

This is a little-more common in Mega26 than the restraining 9...a5.

10.Nd5 Qd7 11.b4 Nd8!?

Preparing to evict the d5 knight.

12.Bg5 Nc8!?

Spanton (2018) - Russell Finch (1650), Guernsey 2005, saw 12...Bxd5!? 13.cxd5 c6 14.d4!? cxd5 15.exd5 h6 16.Be3, after which the engines reckon 16...e4 would have kept Black's disadvantage to a minimum (1-0, 31 moves). The somewhat higher-level encounter, Normunds Miezis (2515) - Alberto Escobedo Tinajero (2370), Olympiad (Elista, Russia) 1998, went 12...Nxd5 13.cxd5 Bf7 14.d4!? fxe4 15.dxe5 Bxe5 16.Bxe4, with the upper hand for White, according to the engines (1-0, 40 moves).

13.Qd2?!

Probably better is 13.exf5!?, when the engines reckon White has at least a slight edge.

13...c6

Even stronger is first capturing on e4, and then evicting the knight, according to the engines.

14.Ndc3 Nf7 15.Be3 Ne7

How should White proceed?

*****

*****

*****

*****

*****

16.a4!?

This apparent-novelty seems an improvement over Mária Machalová (2184) - Radomír Háčik (2158), Slovak Team Championship 2011, when 16.f4 could have been answered by 16...fxe4, with an edge for Black, according to the engines, although Black came out on top anyway in the complications following 16...d5 (0-1, 60 moves). However, possibly best is another apparent-novelty, suggested by engines, 16.b5!?

16...d5 17.cxd5

The engines give 17.exd5 cxd5 18.d4!?, claiming a slight edge for White.

17...exd5 18.exd5 Nxd5 19.Nxd5 Bxd5 20.Bxd5 Qxd5

Following a flurry of exchanges, how would you assess the middlegame?

*****

*****

*****

*****

*****

Black's position looks better, and indeed Stockfish18 gives Black a slight edge, but Dragon1 for a long time calls the game equal, before also settling on a slight edge for Black.

21.Nc3

The engines reckon White should take the open file with 21.Rbc1!?, or push on with 21.a5!?

21...Qd7

21...Qf3 22.Qd1 Qxd1 23.Rfxd1 helps White, as does 22...Qc6 23.Rc1.

After the text, Black threatens to win with 22...f4! 23.gxf4 Qg4+ 24.Kh1 exf4 25.Bxf4 (25.Bc5 Qf3+ 26.Kg1 Ng5 27.Rfe1 Qh3) Bxc3, so what should White play?

*****

*****

*****

*****

*****

22.f3!?

This is Stockfish18's second choice, behind 22.Qa2!?, when 22...Qxd3? loses to 23.Nd5!, after which the threat of 24.Rbd1 cannot be successfully met. Instead, Black should play 22...b6 or 22...Rfd8, with at least a slight edge, according to the engines. Dragon1 suggests 22.Rbc1, but 22...f4!? is still interesting, eg 23.gxf4 exf4!? 24.Bxf4 g5! 25.Bg3 h5.

22...Rfd8 23.Rfd1?

The wrong rook.

23...Qc6

The f3 pawn cannot be reasonably saved, eg 24.Kf2?! runs into 24...e4

*****

*****

*****

*****

*****

24.Rbc1

Perhaps 24.b5 is a slightly better way to give up the f3 pawn, although after 24...Qxf3 the engines reckon that, as in the game, White's best is to play for a swop of queens (with 25.Qe2).

24...Qxf3 25.Qf2

The engines marginally prefer 25.Qe2, or 25.Rf1 Qc6 26.Nb5 Qd7 27.Rc7, but in that case agree 27...Qd5 is winning.

25...Qxf2+ 26.Kxf2 Rd7 27.Ne2

Perhaps hoping to eliminate the d3 weakness, but the engines prefer active play with 27.Nb5 or 27.a5.

27...b6 28.a5

Not 28.d4? exd4 as both 28.Nxd4 and 28.Bxd4 are met by 28...Rad8.

28...bxa5 29.bxa5 Rad8 30.a6 e4 31.d4 h6?

Preparing ...Ne5, but 31...Ne5!? can be played immediately as 32.Bg5? Nd3+ 33.Rxd3 exd3 34.Bxd8 can be simply met by 34...Rxd8!, eg 35.Nf4 Bxd4+ 36.Kf3 d2 37.Rd1 Bc3, after which White is helpless. The more-prosaic 31...Rd6 is also strong.

32.Rc6?

The engines give 32.Rd2, reckoning White is no more than slightly worse.

32...Ne5!

The only winning move, but, bearing the preamble, not difficult to find.

33.Rc2??

Better is 33.Rc3, but 33...Ng4+ 34.Kg2 Nxe3+ 35.Rxe3 Bxd4 gives Black an overwhelming position.

33...Ng4+ 0-1