Opening Lessons From Wrocław IX

IN round nine I had white against Kinga Marczyk (1271)

English Botvinnik

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

More popular are 5.d3 and 5.e3, but there are 2,207 examples of the text in ChessBase's 2021 Mega database. It received a major boost with publication of The Dynamic English by Tony Kosten (Gambit 1999).

5...d6 6.Nge2

The game has reached the basic set-up recommended by Kosten. As he points out, the pawn-formation with a hole at d4 was pioneered by Aron Nimzowitsch and refined by Mikhail Botvinnik, who developed the king's knight to e2 rather than Nimzowitsch's preferred f3.

6...Nge7

Kosten says this is "considered Black's best," and it is the main move in Mega21.

7.0-0 0-0 8.f4!?

There are lines in the Botvinnik system where a quick f4 can inconvenience Black, but that is usually when White threatens to continue with an immediate f5. That is not the case here, so 8.d3 is more flexible, and is very much the main move in Mega21. Basically in the Botvinnik system of the English, White has three possible plans: break on the kingside with f4; break on the queenside by pushing the b pawn; advance in the centre with d4. The last is not often on as Black can drop a knight into d4 at will, and it is probably too early to decide whether to go for a kingside or queenside break, hence 8.d3 is probably better than my choice.

8...exf4!?

Probably a novelty. A 2527 in Mega21 played the standard-type move 8...Be6, while Stockfish14 and Komodo12.1.1 like 8...Bg4. The text makes the f4 square a good post for a white knight, while it is not clear a black piece could remain for long on e5.

9.Nxf4

Stockfish14 narrowly prefers 9.gxf4. My other main analysis engine, Komodo12..1.1 prefers the text, which seems to fit in more with Kosten's general advice.

9...Nd4

Thanks to 8...exf4!?, Black has occupied d4 with  aknight at a moment when White cannot simply swop it off, and the knight is in White's half of the board. On the other hand the white knight is also well-posted, and the half-open f file looks more useful for White than the half-open e file is for Black.

10.d3 Bd7

This is an uninspiring post for the bishop, but the bishop will have to be developed somewhere at some point and it is far from obvious Black will find it a better square.

11.Be3 Nec6 12.Qd2

White has connected rooks in just 12 moves

Komodo12.1.1 at first assesses this position as slightly better for White, while Stockfish14 reckons the position is more-or-less equal. Given more time, the engines swop assessments, and generally fluctuate between having White slightly better or effectively equal. The full game can be seen at https://beauchess.blogspot.com/2021/08/wrocaw-round-nine.html

Summer Prague Round Five

FACED a German junior this afternoon in round five of Summer Prague U2000.

Spanton (1831) - Arne Herter (1613)

English Botvinnik

1.c4 Nf6 2.Nc3 g6 3.e4 d6 4.Nge2 Bg7 5.g3 Nc6 6.Bg2 0-0 7.0-0 e5 8.f4!?

Much more popular is 8.d3, but the text was recommended by Tony Kosten in The Dynamic English (Gambit 1999), although it is not liked by the analysis engines Stockfish11 and Komodo11.01.

8...Nh5!?

The main line in ChessBase's 2020 Mega database goes 8...exf4 9.gxf4 Bg4 10.h3 with a position in which the engines' evaluations fluctuate quite a bit.

9.f5

Getting in this move is the point of omitting d3.

9...gxf5!?

The engines prefer 9...Nd4 but reckon White is better.

10.exf5 Bd7 11.g4 Nf4!

The best move, after which the game becomes very sharp.

12.Nxf4 exf4 13.Ne2?!

I rejected 13.Rxf4 because of 13...Be5 14.Rf2 Qh4, but the engines reckon White is much better after sacing the exchange with 14.d4! Bxd4 15.Bxf4, when Black's kingside dark squares are very weak. Also probably better than the text is the simple 13.d3 - I think I rejected it because I over-feared 13...Bd4+.

13...Qh4 14.Rxf4!?

Sacing the exchange. Stockfish11 comes to quite like the move, but both engines prefer 14.d3.

14...Be5 15.d4 Bxf4 16.Nxf4 Rae8

Black threatens 17...Re1+.

17.Bd2 Nxd4 18.Be1 Qf6

The engines reckon Black should return the exchange with 18...Rxe1+! 19.Qxe1 Qxg4, which leaves Black a pawn up and with a second pawn certain to fall.

19.Nd5 Qe5 20.Nxc7?

Black is up the exchange but White has some counterplay after the engines' 20.Bh4 or 20.Kh1.

20...Rc8?

20...Ne2+ 21.Kh1 Qc5+ wins the white knight thanks to the threat of 22...Qg1#.

21.Nd5 Rxc4

The engines prefer Black after 21...Ne2+ 22.Kf1 Nf4.

How should White proceed?

*****

*****

*****

*****

*****

22.Bh4!

Now Black's king is the more vulnerable - so much so that Black has to jettison material.

22...f6

This is best, according to the engines, as 22...Ne2+?! 23.Kh1 Rd4 runs into 24.Bf6 Rxd5 25.Bxe5 Rxd1+ 26.Rxd1 dxe5 27.Rxd7, when material is level but White's pieces are much more active than Black's.

23.Bxf6 Rxf6 24.Nxf6+ Qxf6 25.Bd5+ Be6!?

This should probably hold but the engines give 25...Kg8! 26.Bxc4 Bc6, when White is the exchange and a pawn up but the exposed white king means the position is apparently drawn, eg 27.Kf1 (if 27.Rc1 then 27...Nf3+ 28.Kf1 Nxh2+, when White has to play carefully to avoid mate or at least decisive material loss) Qe5 28.Qd2 Qe4 29.Qh6 Qg2+ 30.Ke1 Qe4+ 31.Kf2, and Black has to carry on checking as otherwise he gets mated.

26.fxe6 Rc8

This looks, certainly at first sight, as the best way to stop the passer, but Black can draw by giving up the rook with 26...Qe5!, the point being White cannot escape perpetual check after 27.e7+ Kg7 28.Bxc4?! Qe3+. However the engines point out the superior 28.Qd3! Ne2+ 29.Kh1 Rf4 30.Bf3 Bd4 31.Bxb7 Qxe7 32.Bg2 Rxg4, an intricate line - to say the least! - at the end of which material is level but the engines reckon White has slightly better chances.

27.Qf1 Qg6 28.Re1!?

Possible because 28...Qxg4+ 29.Kh1 leaves the black queen badly out of position. Indeed, best seems to be to return to g6 with 29...Qg6, after which 30.Qf4 threatens 31.Rg1 and 31.Qxd4.

28...Re8 29.Qf7+ Kh8 30.Qxg6

Even stronger is 30.e7 as ...Qxg4+ still does not work for Black.

The game finished: 

30...hxg6 31.Re4 Nf3+ 32.Kg2 Ng5 33.Re2 b6 34.h4 Ng7 35.g5 Re7 36.Kf3 Kg7 37.Kg4 Kf8 38.Re4 Rg7 39.Rc4 Ke7 40.Kf4 b5 41.Rc7+ Kf8 42.Rxg7 Kxg7 43.e7 1-0

Learn From The Greats (part five)

WHEN Garry Kasparov visited London in April 2000 to publicise his upcoming world championship match against Vladimir Kramnik, he agreed to play me - I was a writer for The Sun - in his rooms at The Savoy.
I turned up only to find he did not have a chess set - something I could have anticipated as, after all, how many players travel with anything other than a laptop and a pocket set?
A lackey was sent out to quickly get a full-sized set, and we sat down to play.
I was not allowed to keep score. This was not Kasparov's doing - rather it was The Sun photographer who reckoned a scoresheet and pen would clutter the playing table.
Spanton - Kasparov
Savoy Hotel (London) 2000, Game 1
English Botvinnik
1.c4
Kasparov reacted to this by saying "Oh," apparently surprised at the 'sophistication' of my opening choice. This was the only time he spoke during the game, which we played at blitz speed (but without clocks).
1...e5 2.g3 Nc6 3.Bg2 g6 4.Nc3 Bg7 5.e4
Aiming for a Botvinnik set-up, as recommended in Tony Kosten's The Dynamic English, which had been published by Gambit the previous year.
5...d6 6.Nge2 h5!?
Kosten: "This move is favoured by some aggressive players." It turns out to be also Stockfish10's choice.
7.h3!?
Dubious, according to Kosten, who recommended the much-more popular 7.h4. But the text is preferred by Stockfish10.
7...Be6
Kosten warned that "White is well on the way to ending up with a bad bishop" after 7...h4 8.g4?! (Kosten's punctuation) Bh6. He added: "White does not want to place too many of his pawns on light squares, at least not unless he can exchange the king's bishop."
However, 8.g4 is the most-popular move in ChessBase's 2020 Mega database after 7...h4, and is the choice of Stockfish10 and Komodo10. Black did not reply 8...Bh6 in any of the 16 games in Mega20.
The engines strongly dislike 8...Bh6, giving 9.d4, reckoning White has the upper hand.
8.Nd5 Qd7 9.d3 Nd8!?
The idea is to expel the white knight by ...c6, without allowing the knight to exchange itself.

How should White proceed?

*****
*****
*****
*****
*****
10.Be3
After the game was over, Kasparov dictated the moves to me, making occasional comments. At this point in the dictation he said: "I played d4 against Nigel Short in Linares in '93, no '90. It is a better move."
In fact that game, which was indeed in 1990, went 1.c4 Nc6 2.Nc3 e5 3.g3 g6 4.Bg2 Bg7 5.d3 d6 6.e4 Be6 7.Nge2 Qd7 8.Nd5 Nce7, and now 9.d4.
Here is the position after 8...Nce7:

Kasparov - Short, Linares 1990

The position in Kasparov - Short is clearly not the same as in Spanton - Kasparov, but the pawn-structure and piece-placements are very similar. The point, I believe, is that, if Black allows it, White should seize the chance to play d4 in such a structure.
10...c6 11.Ndc3 Ne7 12.Qd2 d5 13.cxd5 cxd5 14.exd5?
Giving Black a central space advantage and leaving White with a weak d pawn. The engines reckon White is slightly better after 14.d4.
14...Nxd5 15.Nxd5?!
Kasparov: "You had to play Ne4." He added, and I cannot recall his exact words, that capturing on d5 is too weakening. I presume I wanted to avoid giving up the bishop-pair to ...Nxe3.
For what it is worth, Stockfish10 for quite some time prefers the text, but eventually agrees with Komodo10 (and Kasparov) that 15.Ne4 is better.
15...Bxd5 16.Bxd5 Qxd5 17.0-0 Nc6 18.Nc3 Qd7 19.Kg2 f5 20.Rad1?
Kasparov: "Bad. You should play Bg5, when White has chances to resist."
20...h4 21.f4?
The engines suggest giving up a pawn by 21.g4!? fxg4 22.Rg1, but clearly Black is much better.
21...hxg3 22.Kxg3 g5 0-1
Kasparov: "That was a good game."
It was also a quickly-played game - so much so that he offered me a chance (however hypothetical) for revenge, insisting I again take the white pieces.
**********
Spanton - Kasparov
Savoy Hotel (London) 2000, Game 2
Morra Gambit
1.e4 c5 2.d4 cxd4 3.c3 dxc3 4.Nxc3 Nc6 5.Nf3 e6 6.Bc4 d6 7.0-0 Nf6 8.Qe2 Be7 9.Rd1 e5
This is one of the main lines of the Morra, with more than 1,000 examples in ChessBase's 2020 Mega database.

How should White proceed?

*****
*****
*****
*****
*****
10.Bg5?!
Most popular is 10.h3, which Graham Burgess treated as the main line in Winning With The Smith-Morra Gambit (Batsford 1994). However, he expressed sympathy for 10.Be3, which is recommended by Marc Esserman in Mayhem In The Morra! (Quality Chess, 2012).
Kasparov criticised my choice, saying I should have played 10.b4!?, one point being that 10...Nxb4? runs into 11.Nxe5. A game between two future grandmasters, Pavel Šimáček (2481) - Milan Pacher (2255), Česká Třebová (Czech Rep) 2007, saw 10.b4!? Bg4 11.Be3 0-0 (11...Nxb4?! 12.Rab1 and Rxb7) 12.Rab1 Rc8 13.h3 Bxf3!? 14.Qxf3, when the engines reckon White has full compensation for the pawn (1-0, 25 moves).
10...0-0!?
Burgess reckoned the problem with White's 10th move is 10...Bg4, but ...Bg4 and castling are almost equally popular in Mega20.
11.Bxf6!?
"Too early," Kasparov said after the game, but it is Stockfish10's choice.
11...Bxf6 12.Nd5?
The engines give 12.h3 (Stockfish10) or 12.Qd3 (Komodo10), in each case reckoning Black has, at best, a slight edge.
12...Bg4 13.h3 Nd4 0-1
Kasparov reckoned White has "reasonable drawing chances" after 14.Rxd4! Bxf3 15.Qxf3 exd4 16.Nxf6+ Qxf6 17.Rd1.
LESSONS FROM THESE GAMES
In both games I had decent positions out of the opening (assuming you do not regard the Morra as simply bad), but my understanding of what to do in the positions was well below my ability to memorise the moves needed to get to those positions. Both the English Botvinnik and the Sicilian Morra are quite sophisticated opening choices in that they lead to unbalanced positions in which positional understanding is very important. It therefore makes sense to either learn the openings more thoroughly, or, if time or inclinations do not permit, to play something simpler.

A City Called City

MANY visitors to Malta admire the modern capital Valletta, others like the beauty of the Blue Lagoon, but for me the most interesting site is the old capital Mdina.
It got its name, which comes from the Arabic for city (as does Medina in Saudi Arabia), after the Christian population was massacred and replaced by Muslims in 870.
They were in turn conquered by Normans from Sicily, and in 1530 the whole of the island was given by the Holy Roman Emperor Charles V to the Knights of St John.
It was the Knights who effectively transferred the capital to the coast, alongside the Grand Harbour, which was the main setting for the Great Siege of Malta in 1565.
But Mdina - pronounced Im-deen-er by the Maltese - still had symbolic and military value.
So the besieging Turks, looking for a way to boost morale and secure a base for the winter, advanced to attack the city when their efforts to force the Grand Harbour and its fortifications were held up.
The Mdina garrison was low in numbers and ammunition, but opened fire with its guns at a ridiculously long range, leading the Turks to cancel their planned assault as they assumed the city was abundantly supplied.

Main entrance to Mdina

Close-up of the gate

A typically walled and paved street

View from a bastion with the Mediterranean in the distance

Almost certainly St Paul - his shipwrecking on the way to Rome in about AD59 is traditionally located near here

Iluminada Minguez Greciano (1765) - Spanton (1934)
Malta Round 6
English Opening
1.c4 e5 2.Nc3 Nc6 3.g3 g6 4.Bg2 Bg7 5.e3
5.e4 is common at club level in the UK, not least thanks to Tony Kosten's The Dynamic English, but the text and 5.d3 remain the most popular moves at higher levels and abroad.
5...d6 6.Nge2 Nge7 7.Nd5!?
Unusual this early, but it seems reasonable.
7...0-0 8.Nec3!?
This may be a novelty. White's knight-dancing looks strange, but I could see no way to exploit it.
8...Be6 9.0-0
The game has transposed to a known position.
9...Qd7 10.Re1 f5
Now we leave Mega20 for good. 10...Nd8 was played in Vladimir Botto - Milan Cenker, Slovakian championship 1997.
11.d3 Rf7 12.Bd2 Raf8 13.Rb1 Nd8 14.b4 c6 15.Nxe7+ Qxe7 16.b5 Qd7 17.Qa4

How should Black meet the threat to the a7 pawn?

*****
*****
*****
*****
*****
17...f4!
Offering the a7 pawn, but a line such as 17...cxb5 18.cxb5 b6 was distinctly unpalatable.
18.bxc6
Black gets a strong attack after 18.Qxa7, but the position is unclear. I intended 18...fxg3 19.fxg3 Rf2, when Stockfish10 and Komodo10 reckon White is OK after 20.Re2, but that Black has full compensation for the pawn.
18...bxc6 19.exf4 exf4 20.Ne4
A great square for the knight. Black will have difficulty evicting it while ...d5 can be met by Nc5.
20...fxg3 21.hxg3 h6
Just about playable is 21...Bh3? as 22.Bxh3 Qxh3 23.Ng5 can be met by 23...Qh6, but the text is better.
22.Bb4 Bd4 23.Rf1
The engines very slightly prefer to sac the exchange with 23.Bxd6!? Bxf2+ 24.Kh2 Bxe1 25.Rxe1 Re8, when they reckon White has full compensation.
23...Nb7 24.Bd2?
White needed to play something like 24...Bc3.
24...Kh7?
I looked at 24...Nc5 25.dxc5 dxc5 and rejected it because of 26.Bxc6?, but I did not look far enough as the engines point out 26...Qc7 (26...Rxf2!? also wins). So White has to allow 26.Be3 Bxe3 27.fxe3 Qxd3, or 26.Bf4 Bh3, when Bxc6 this time loses to 27...Qf5 with a huge attack for a pawn.
25.Be3 Bxe3 26.fxe3 Nd8 27.Nd2?
A very strange retreat as the knight was performing great duty on e4.
27...Bg4?!
27...Bf5 seems to give more pressure.
28.Rbe1?!
Preventing 28...Be2, but a less passive way to save the d3 pawn was 28.Rxf7+ Rxf7 29.Ne4, when 29...Be2?! can be simply met by 30.Qc2 (but not 30.d4?? Bd3 31.Rd1 Bxe4 32.Bxe4 Qg4 etc).
28...Bh3!?
Again ...Bf5 may be objectively slightly better, but the text exchanges off a major defender of the white king's position, making it easy for White to go wrong.
29.Rf4 Bxg2 30.Kxg2 Ne6 31.Rxf7+ Rxf7 32.Ne4 d5 33.Nf2?
White is fine after 33.cxd5 Qxd5 34.Qc4.
33...Nc5
Even stronger, according to the engines, is 33...Ng5.
34.Qc2?
Better is 34.Qd1 when 34...Qf5, as in the game, is not so catastrophic, although after 35.Rf1 dxc4 36.dxc4 Nd7 37.Qe2 Ne5, White's four isolated pawns will start dropping off.
34...Qf5
Threatens 35...Qxf2+! 36.Qxf2 Rxf2+ 37.Kxf2 Nxd3+ etc.
35.Re2 Qf3+ 36.Kh2 Rf5 37.Nh3??
Much better is 37.g4, but Black is more-or-less winning after 37...Rf7, eg 38.cxd5 cxd5 39.Nh3 (39.Qxc5?? Qxe2) Rc7, when White's g or e pawn will fall and her king remains exposed.
37...Rh5 38.Re1 Qg4 39.Qg2 Nxd3 40.Rf1 dxc4 41.a4 c3 42.Kg1 Rxh3 0-1

Guernsey Round Seven

Kees de Kruif (1985) - Spanton (1927)
Botvinnik English
1.c4 e5 2.g3 Nc6 3.Bg2 g6 4.Nc3 Bg7 5.d3 Nge7 6.e4
The most popular move in ChessBase's 2019 Mega database, setting up a Botvinnik formation, assuming the king's knight goes to e2 (if it goes to f3 we have an earlier, similar formation pioneered by Nimzowitsch). An excellent source for understanding the Botvinnik set-up is Tony Kosten's The Dynamic English.
6...0-0 7.Nge2 d6 8.0-0 f5 9.Rb1 Be6 10.b3?
A novelty, and not a good one. This move is bad positionally and tactically (I only managed to spot the former drawback).

Black to make his 10th move

*****
*****
*****
*****
*****
10...Qd7
The positional problem with White's move is that Rb1 and b3 do not go well together. The former, in this type of position, is normally a prelude to queenside expansion with b4.
The second problem is that Black can reply with the positionally desirable 10...f4! thanks to the tactical point 11.gxf4 exf4 12.Nxf4?! Rxf4 13.Bxf4, dropping the queen's knight to 13...Bxc3. Black ends up with two knights and the less-exposed king for rook and pawn.
Note that 10.b4? can be met in a similar way, although it was not the three times 10.b4? appears in Mega19. Most strong players who have reached this position as White have preferred 10.Nd5.
11.Nd5 Rf7 12.f4 Raf8 13.Be3 a6 14.Qd2 Kh8 15.Kh1
KdK offered a draw.
15...Ng8 16.Rbd1 fxe4 17.Bxe4?!
Stockfish10 and Komodo10 prefer 17.dxe4.
17...exf4?!
I somehow missed White's reply. Stockfish10 reckons Black has the upper hand after 17...Nd4!? Komodo10 also likes the move, but is not so enthusiastic about it.
18.Rxf4 Rxf4 19.Ndxf4 Bf5 20.Bg2 Nf6 21.Nd4 Nxd4 22.Bxd4 c6 23.Re1 Re8 24.Rxe8+ Qxe8 25.Kg1 Ng4!
At first glance it might seem strange to allow an exchange of bishops with check, but White has to find a fairly precise defence.
26.Bxg7+ Kxg7 27.Bf3!
Best, according to the engines, as the 'obvious' 27.Qc3+ can be simply sidestepped with, for example 27...Kf7, when Black gets a bit of an initiative.
27...Qe5 28.h3?!
Almost certainly better was 28.Bxg4, hoping to take advantage of the fact that a knight and queen often work well together, although here the knight does not have an unchallengeable outpost (=, according to the engines), or 28.d4, when I planned 28...Qe3+ 29.Qxe3 Nxe3, but the engines give 30.Kf2 as also =.
28...Qd4+ 29.Kg2??
Correct was 29.Kf1, when 29...Ne5 30.Be4 Qa1+ is a little better for Black as White will get an isolated e pawn.
29...Ne3+ 30.Kg1
Forced as 30.Kh1?? loses a piece to 30...Qa1+. Note that the white king cannot go to h2, either immediately or on move 31, because of the fork ...Nf1+.
30...Nxc4+ 31.Qf2 Qxf2+ 32.Kxf2 Ne5 33.d4 Nxf3 34.Kxf3 a5 35.a3 Kf6 36.b4 axb4 37.axb4 g5 38.Nh5+ Ke7 39.h4
The only realistic way to keep the game going was 39.g4 Bg6 40.Ng3, but the position is hopeless for White.
39...Bg6 40.Ng7 Kf7 41.h5 Kxg7 42.hxg6 Kxg6 0-1
My final score of +1=4-2 saw me gain 1.6 Fide elo.

A Tale Of Two Islands

Spanton (1914) - Ian Heppell (2001)
Jersey Round 8
Botvinnik English
1.c4 e5 2.Nc3 Nc6 3.g3 f5
Black is playing the Grand Prix Attack with colours reversed and a tempo less.
4.Bg2 Nf6 5.a3!?
This move goes all the way back to a game of Bird's in 1889 (he lost) and has been played in more modern times by Nakamura, Naiditsch and other strong players. The idea - at least, my idea in playing it - is to stop Black playing ...Bb4 (Bb5 being a popular move for White in the Grand Prix Attack proper).
The most popular line in ChessBase's 2019 Mega database runs 5.d3 Bb4 6.Bd2 0-0 7.a3 Bxc3 8.Bxc3 d6 9.Nf3 Qe8, reaching a position from which Black scores a stupendous 63%, but which my main analysis engines Stockfish10 and Komodo9 regard as quite a bit better for White.
The last time I had White against IH, also in the Channel Islands, the game continued 5.e3!? (Tony Kosten warns in The Dynamic English against playing this) d5!? 6.Nxd5 Nxd5 7.cxd5 Nb4 8.d3 Nxd5, when, according to Kosten, "Black has a reasonable Sicilian structure." Spanton (2030) - Heppell (2106), Guernsey 2013 (½–½, 17 moves).
5...d6
Most popular is 5...a5, but it is by no means clear that Black has to stop White playing b4.
6.d3 g6!?
Now Black is playing a Closed Sicilian with colours reversed and a tempo down.
7.e4
Arguably more consistent was 7.b4 but the text, which establishes a Botvinnik formation, has been the choice of  at least two 2500+ players.
7...Bg7 8.Nge2 0-0 9.0-0 Ne7
This may be a novelty. Most popular is 9...Be6.
10.b4 c6 11.Bb2 Be6 12.Rc1
Preparation for the c file being opened after a later ...d5.
12...Qd7 13.a4?!
Black's next move shows White's queenside play is too slow.

Can you find Black's thematic move in this type of position?

*****
*****
*****
*****
*****
13...f4! 14.gxf4 Nh5 15.fxe5
The engines agree this is White's best try.
15...dxe5!?
I thought 15...Bxe5 was more dangerous, but the engines prefer the text. The position is certainly unclear, eg after 15...Bxe5, Komodo9 likes 16.f3, which it reckons is slightly better for White, but Stockfish10 reckons 16.f3 is slightly better for Black.
16.Rc2
Again the position is unclear. Komodo9 suggests 16.b5 with equal chances, but Stockfish10 reckons Black is close to winning after meeting 16.b5 with 16...g5 or 16...Rf7.
16...Bh3
The engines prefer a slow build-up with 16...g5 or 16...Rad8.
16...Bh6 looked strong in the postmortem, but as with many moves around here the best continuation is far from clear.
17.f3 Rad8 18.Bc1 Bxg2
The engines want to prepare ...g5 with 18...h6 or 18...Bf6.
19.Kxg2 Nf4+?!
This seems to dissipate what remains of Black's advantage. Again ...h6 and ...Bf6 are liked by the engines.
20.Nxf4?!
The engines agree White is quite a bit better after 20.Bxf4 exf4 21.d4.
20...exf4 21.Ne2 g5 22.d4 Ng6 23.Kh1 Nh4?!
The engines give 23...g4 24.d5 cxd5 25.Qxd5+ Qf7 26.Qxf7+ Rxf7 27.fxg4 Re8, when Black is temporarily two pawns down, but is about to get one back and apparently has almost full compensation for the other one.
24.Qe1?
Better was 24.d5 as the text allows Black a neat combination.

Black to make his 24th move

*****
*****
*****
*****
*****
24...Qh3
This is fine if followed up correctly, but even stronger seems to be 24...Bxd4! We both thought it failed to 25.Rd2, but then comes 25...Qh3 26.Nxd4 Rxd4 27.Rdf2 (27.Rxd4?? Qg2#) Rd3 28.Qe2 Rfd8 with what the engines reckon is a winning attack.
25.Ng1 Qd7
Better is 25...Qc8! with 26...Bxd4 to follow, when Black has restored material equality and still has an attack, although the engines disagree just how much better this is for Black.
26.Ne2
As IH pointed out in the postmortem, White has an edge after 26.Bb2.
The game finished:
26...Qh3 27.Ng1 Qd7 ½–½