The Kibitzer
by Tim Harding
Have a Hack with the Hennig-Schara
THIS MONTH'S COLUMN deals with an aggressive way to meet
the Queen's Gambit1 d4 d5 2 c4 e6 3 Nc3 c5 4 cxd5 cxd4!?, the
Hennig-Schara Gambit (ECO code D32).
This gambit, or counter-gambit, can be quite effective. In its early
days it was employed by Alekhine and Tartakower and quite a few
Soviet masters also tried it.
It might appear to be primarily suited to rapid-play events but in
fact has been equally successful in both over-the-board and
correspondence events, and there is good reason to believe that the
gambit has a sound positional and dynamic basis.
When preparing this article, I reflected that there was one very
solid opponent whom I met with Black in three successive years in
the Irish CC Championship. He drew twice and the game I won
was the one in which I played the gambit. He got confused in the
complications, gave up the bishop pair, returned the pawn to
simplify and I won the endgame. So you don't always have to play
for mate though of course it's pleasant when something like the
following happens: 1 d4 d5 2 c4 e6 3 Nc3 c5 4 cxd5 cxd4 5 Qxd4
Nc6 6 Qd1 exd5 7 Qxd5 Bd7 8 Qd1 Nf6 9 Nf3 Bc5 10 e3 Qe7 11
Bb5 (Not White's best.) 11...0-0-0 (Black can castle either side in
this gambit but queenside is more aggressive and usually better.)
12 Qe2 a6 13 Bxc6 (If 13 Ba4 g5! with interesting complications.)
13...Bxc6 14 0-0 (My opponent tried 14 Bd2 here, which is better
and unclear.) 14...Ne4! 15 Bd2 Nxc3 16 bxc3 (16 Bxc3 Bb5)
16...Qe4! 17 c4 Rd6 18 Bc3 Rg6 19 Kh1 Rxg2! 0-1
Kuznetsov-Lerner, USSR 1977.
So I thought it might be interesting to do a ChessBase "opening
report" on the gambit. The statistics, with over 1000 games with
the gambit in my reference database, should be quite significant
Black scores above average (47%). Games (whatever result) are
shorter than average.
The two main moves 5 Qxd4 (as above) and 5 Qa4+ are played
with almost equal frequency. As we shall see, they can and often
do lead to the same position a little further on.
I don't want to waste time arguing about names. On the continent
of Europe the name of Schara is often given first and Eric
Schiller's 1992 monograph is entitled "Von Hennig-Schara
Gambit" which is perhaps strictly correct. He says that "Schara was
the first to do any serious analysis of the line and Von Hennig
introduced it into serious tournament play." Schiller has a game
from him played in 1929.
Anton Schara used the gambit to defeat Ernst Gruenfeld in a short
game played in Vienna in 1918 which isn't in Schiller's booklet.
Perhaps it wasn't a serious game but here it is: 1 d4 d5 2 c4 e6 3
Nc3 c5 4 cxd5 cxd4 5 Qxd4 Nc6 6 Qd1 exd5 7 Qxd5 Bd6 8 Bg5
Nge7 9 Qd2 f6 10 Bh4 Qb6! 11 Nf3 (11 Qxd6? Qxb2) 11...Bb4 12
e3 Bf5 13 Bc4 Na5 14 Bd3 Rd8 15 Nd4 Nac6 16 Nxf5? Nxf5 17
Bg3 Nxg3 18 hxg3 Ne5 19 Bb5+ Qxb5 0-1.
The game won by the other author of the gambit went as follows. J.
Benzinger - Heinrich von Hennig Duisburg II, 1929 1 d4 d5 2 c4
e6 3 Nc3 c5 The Tarrasch Defence, but with a twist in mind... 4
cxd5 cxd4!? Instead of the standard recapture. 5 Qxd4
White can also give check here. We will look at that later. 5 dxe6?
sets the trap 5...dxc3? 6 exf7+ Ke7 7 fxg8N+ Rxg8 8 Bg5+ 1-0
(D.Fidlov-Albert Maier, USA corr 1959) but Black gets a good
game with 5...Bxe6! as has been proved on most of the few
occasions that White has tried that fifth move.
5...Nc6
A characteristic move in this counter-gambit - Black gains a tempo
thanks to the pin on the d-file. In the earliest game in my database,
Frank Marshall-Howard, Sylvan Beach 1904, Black played
5...Nf6? and White won quickly (he replied 6 e4).
6 Qd1 exd5 7 Qxd5
White almost invariably accepts the gambit. One of my opponents
did choose 7 Nf3 here (similar to the 7 e3 seen in Viakhirev-
Chepurnov, St Petersburg 1912!) but I won comfortably. If he
leaves the pawn untouched, hoping to exploit the weak isolated d-
pawn characteristic of the Tarrasch Defence, the tempi lost with
the queen are bound to benefit Black.
Euwe played 7 Nxd5 in a game played in Holland in 1920 but this
move is inaccurate and Black can equalise with 7...Nf6 e.g. 8 Nc3
Be6 9 Bd2 Qb6 10 e3 Rd8 11 Bb5 Bb4 12 Nge2 0-0 13 Bxc6 bxc6
14 0-0 Bxc3 15 bxc3 Ne4 16 Nd4 Nxd2 17 Qxd2 c5 18 Rab1 Qd6
19 Rb7 cxd4 20 cxd4 Qa6 21 Rb2 Rb8 22 Rxb8 Rxb8 23 d5 Qd6
24 e4 Bd7 25 h3 a6 26 Rc1 Qe5 27 Re1 h6 28 f4 Qb2 29 Re2
Qxd2 30 Rxd2 Rb4 31 Re2 Bb5 32 Re3 Rb2 33 e5 Rxa2 34 f5 Rd2
35 d6 f6 36 exf6 gxf6 37 Re6 Rd1+ 38 Kh2 Kf7 39 Re7+ Kf8 40
Re6 Kg7 41 Re7+ Kg8 42 Re6 a5 43 Rxf6 a4 44 Rg6+ Kh7 45 Re6
Rd2 46 Kg3 a3 47 Re7+ Kg8 48 Ra7 a2 49 f6 Bc4 50 Ra3 Rxd6 0-
1 T.Hradeczky - K.Szeles, Hungary 1972.
If instead White plays 8 Nxf6+ Black even got the better of it by
8... Qxf6 9 a3 Bc5 (as recommended by Demuth & Konikowski, in
"Fernschach" 1987) 10 Nf3 0-0 11 e3 Bb6 12 Be2 Rd8 13 Bd2
Bg4 Alv.Garcia- Ana Russek, Guarapuava 1991. Euwe's opponent
played 7...Be6!? when 8 Nc3 Nf6 would transpose to the
Hungarian game, but the Dutch player (A.Koning) chose 8...Qxd1+
and lost. If White meets 7...Be6 by 8 e4 then 8...Bxd5 (8...Nf6!?) 9
Qxd5 Qxd5 10 exd5 Bb4+ 11 Bd2 Bxd2+ 12 Kxd2 0-0-0 was
given as equal in New In Chess Yearbook 36.
(From the last diagram) 7...Be6?! The main move is 7...Bd7; see
below. In this case, Black heads directly for the endgame. White
sometimes plays 5 Qa4+ to rule out this idea but it is not
particularly dangerous anyway and we shall look at the modern
main lines shortly. 8 Qxd8+ Rxd8 9 e3 White can also play 9 g3 or
9 Bd2. 9...Nb4 10 Bb5+ Ke7 11 Ba4?! This allows Black to get a
good game. Smyslov preferred 11 Kf1 here in a well-known game
against Estrin and 11 Ke2 in another one against Aramanovich.
11...Bc4! 12 Nge2 b5 13 Bd1 Nd3+ 14 Kf1 b4 15 Ne4 (15 Na4?
Nf4) 15...f5 (15...Nf6!? 16 Nxf6 Nf4! was suggested in NIC
Yearbook 36.) 16 N4g3 g6 17 Bc2 Bg7! 18 Bxd3 Bxd3 19 f3 Rc8
20 Ke1 Nf6 21 Nf4 Bc2 22 Bd2 a5 23 Rc1 Nd7 24 Nd5+ Ke6 25
Nf4+ Kf7 26 b3 Bb2 27 Kf2 Bxc1 28 Rxc1 Ne5 29 h4 Rhd8 30
Be1 Nd3+ 31 Nxd3 Bxd3 32 Ra1 Rc2+ 33 Kg1 Rdc8 0-1.
Now let us look at the main line of the gambit. The following
diagram was reached in about 40% of the 1000+ games that with
the Henning-Schara that I found in my main database.
This position (which arose in the Kuznetsov-Lerner game given
above) can be reached by two move-orders a) 1 d4 d5 2 c4 e6 3
Nc3 c5 4 cxd5 cxd4 5 Qxd4 Nc6 6 Qd1 exd5 7 Qxd5 Bd7 8 e3 Nf6
9 Qd1 Bc5 10 Nf3 Qe7 b) 1 d4 d5 2 c4 e6 3 Nc3 c5 4 cxd5 cxd4 5
Qa4+ Bd7 6 Qxd4 exd5 7 Qxd5 Nc6 8 e3 Nf6 9 Qd1 Bc5 10 Nf3
Qe7
Note that this the same position arises after Black's 7th move in
each case, although both sides have made some moves in a
different order depending on whether White gives the check on a4
or not. His queen gets back home to d1 after making four moves in
either case: these wasted tempi are crucial to Black's
compensation. (He can try to put his queen somewhere else but
after 9 Qb3 Be6! Black gets good compensation for his pawn.)
I think it is time to see what "the world" knows about this gambit.
The new "Nunn's Chess Openings" is usually quite a good first
source for seeing what top theoreticians think White should play to
refute the gambit. The first encouraging thing is that Graham
Burgess (who wrote this section) doesn't find a refutation - the
main line just ends in a small advantage for White ("plus over
equals") which White usually gets in other lines of the Queen's
Gambit anyway.
In the 7...Be6 line, Burgess follows the Smyslov idea of 11 Kf1
mentioned above. In the main line, which they give via the 5 Qa4+
route, Burgess recommends 11 Be2
giving these variations:
a) 11...0-0 (hoping for positional compensation) is reckoned
inferior on account of 12 0-0 Rfd8 13 a3 Rac8 (Ravinsky's old
move 13...Bf5 hasn't been doing so well lately either.) 24 b4 Bg4
15 Qb3 Bd6 16 Bb2 with clear advantage to White (plus over
minus) in Nissi-Aulaskari, Finnish Corr Ch 1993. b) 11...0-0-0 12
0-0 g5 is row 1 on page 391, leading to two possibilities for White.
b1) 13 b4 Bxb4 14 Bb2 (14 Qb3 Be6 15 Bc4 Rd3 unclear) and now
14...Kb8!? and ECO's 14...Rhg8 are suggested while 14...g4 is met
by 15 Nd4 with transposition to line b2.
b2) 13 Nd4 g4 (13...Qe5!? is offered as an option with no further
comment.) 14 b4 (or 14 Bb5!?) 14...Bxb4 15 Qb3 Nxd4 16 exd4
Bc6 17 Nb5 with the slight advantage to White assessment already
mentioned.
In this line, 15 Bb2 is also considered15...h5 16 Ncb5 (16 Rb1!?)
16...Kb8 17 Qa4 a6 18 Nxc6+ Bxc6 19 Bxf6 Qe4!= (See
Diagram). Actually until Black's 19th move this is Polugaevsky-
I.Zaitsev, Alma-Ata 1969, and Black's brilliant 19th was a
suggestion by the Romanian master Samarian who made a great
study of the Tarrasch Defence.
After 19...Qe4! Black gets a tremendous attack for the piece
sacrificed. V.Bronznik - P.Cech, Prague 1993, continued 20 f3
Qxe3+ 21 Kh1 Qxe2 (21...Rd2!?) 22 Qxb4 Rd2! (Kholmov's
improvement on 22...gxf3 which was Samarian's main line.) 23
Qf4+ Ka8 24 Nc7+ Ka7! (Schiller only considers 24...Kb8 here.)
25 Bd4+ b6 26 Bf2 gxf3 27 g3 Rhd8 and now White collapsed by
28 Rab1? Qxf2! 29 Nb5+ Ka8 0-1. Bronznik should have tried 28
Rae1 when after 28...Rd1 29 Kg1 Rxe1 30 Rxe1 Rd1 31 Qb4
Rxe1+ 32 Qxe1 Kb7 the eventual result will be a drawn opposite
coloured bishop ending.
Note in both lines b1 and b2, that White does not play passively. In
order to regain the initiative, he gives back the gambit pawn by the
advance b2-b4, to deflect a black piece and open the b-file for a
counterattack against the black king. White used to play 11 a3 to
prepare b4 (this was the main line in ECO D, 2nd edition) but the
move is rather slow. In that variation Black can consider castling
kingside or can follow the usual plan and meet the eventual b2-b4
by ...Bc5-b6.
Unfortunately, although Black has a lot of tactical opportunities
that can lead to victory against inferior defence, there is as yet no
clear improvement for Black in Burgess's main line with 13 Nd4
g4 14 b4 Bxb4 15 Qb3.
One possibility for Black is to delay castling and play 11...g5!? 12
0-0 g4 13 Nd4 h5 which has had some successes. For example, the
1967 USSR correspondence game Lisov v. Shkurovich Hazin
continued 14 Bb5 h4 15 Bxc6 bxc6 16 e4? Qe5 17 Nf5 Rd8 18
Qc2 g3 19 h3 gxf2+ 20 Kh1 Nh5 and White resigned because of
the threat 21...Qxf5! 22 exf5 Ng3+ and Black mates on move 25
with a knight underpromotion! Of course White can improve on
this; one possibility is to play 12 Nd4 when if 12...g4 13 Nxc6
Bxc6 14 Bb5 as given in ECO (following Bagirov-Kurdryashov,
USSR 1969). Black's best may be 12...0-0-0 returning to the
normal lines with 11...0-0-0.
So let us try harder to revive Black's chances in line b2 above.
After 1 d4 d5 2 c4 e6 3 Nc3 c5 4 cxd5 cxd4 5 Qa4+ Bd7 6 Qxd4
exd5 7 Qxd5 Nc6 8 e3 Nf6 9 Qd1 Bc5 10 Nf3 Qe7 11 Be2 0-0-0
12 0-0 g5 13 Nd4 g4 14 b4 Bxb4 15 Qb3 Nxd4 16 exd4 let's play
16...Be6! (instead of 16...Bc6) (See Diagram)
In Schiller's monograph, on page 11, the following over-optimistic
assessment appears: "Black's advanced development and initiative
are worth more than a pawn, so White gives up a pawn in an effort
to equalize". Tozer-Schiller, Peterborough 1986, went 17 d5?!
Nxd5 18 Nxd5 Bxd5 19 Bc4 Bc6! and Black went on to win in the
ending.
Unfortunately Schiller doesn't say what Black should do against
the calmer 17 Qb2. It's not as if it is an unknown or new move.
This was in fact the old main line when the model game Portisch-
Velimirovic, Rio de Janeiro 1979 (mentioned in ECO), went 17
Qb2 Nd5 18 Nb5 Kb8 19 Bc4 Ba5 20 Bd2 Bxd2 21 Qxd2 Qf6 22
Qa5 a6 23 Bxd5 Bxd5 24 Qc7+ Ka8 25 a4 Rc8 26 Qg3 and here
26...Qc6!? would have been = according to analysis published at
the time.
Instead of 18 Nb5, NCO cites 18 Nxd5 Rxd5 19 Bf4 Bd6 20 Bxd6
Qxd6 21 Rfb1 b6 (21...Qd7? 22 Ba6!) 22 a4 with a clear advantage
in Prudnikova-Voiska, Azov women's interzonal 1990, from
Informator 50.
So which is right? Schiller's gung-ho view of Black's chances,
Burgess's dismal view of his prospects or something in between?
Look at the diagram again. Black is level on material now. White's
d-pawn is as weak as Black's g-pawn if not more so. Black is
slightly better developed too; the problems are the safety of his
king and the possible awkward pin by Bg5. If this can be solved, or
if he can give up the g-pawn to revive his attack (with hopes of
collecting on d4 later) then the line should be playable again.
It's impossible to find a better move than 17...Nd5 but after 18
Nxd5 instead of the rook recapture Black can offer the g-pawn by
18...Bxd5! e.g. 19 Bxg4+ (or first 19 Bf4 Bd6 20 Bxg4+
transposing) 19...Kb8 20 Bxg4+ Kb8 21 Rfe1 Qf6 (instead of the
Informator line 21...Qh4 22 Bxd6+ Rxd6 23 Qa3 Rg6 24 Bf3).
Except for being a pawn down, Black's position is fine; White's
queenside threats are history. Now an email game continued 22
Bxd6+ Qxd6 23 Bh3 (a suspicious move which, as soon transpires,
doesn't succeed in securing g2) 23...Rhg8 24 Rab1 Bc6 25 Rbd1
Qd5 and suddenly Black's pressure focused on g2 has brought a
tangible result. White collapsed with 26 f4 Qf3 27 Rc1 Bd5 28 Rc3
Qxf4 29 Rce3 Bc6 30 Rd3 Rde8 31 Rc1 Re2 32 Qd2 Qf2+ 0-1
M.Rice- H.Daurelle, IECG 1997.
This line may not be the ideal solution for Black but it does show
that writing off Black in the main line of the Hennig-Schara is just
as misleading as making overoptimistic claims for his chances.
There isn't space in a column like this to analyse every critical line
in such detail so I am going to do two things. One is to present here
with notes some little-known games, mostly where Black wins; the
other is to make available for download on my website a collection
of over 200 Black wins in the gambit. The games are available in
PGN and in both new and old ChessBase formats at
http//www.chessmail.com/freegames.html and they will be there
until the next Kibitzer column is posted, if not longer.
If there is sufficient reader interest in this topic, I may look at more
lines in the gambit either next month or later in the year.
The games posted at the Chess Mail website don't include any
correspondence games, since they will be among those made
available on our forthcoming database CD in August, but I give a
few here with light notes.
Laks - Aramanovich Czechoslovakia-USSR corr, 1956
1 d4 d5 2 c4 e6 3 Nc3 c5 4 cxd5 cxd4 5 Qxd4 Nc6 6 Qd1 exd5 7
Qxd5 Bd7! 8 Nf3 Nf6 9 Qd1 Bc5 10 e3 Qe7 11 Bd2 Rd8 12 Be2 0-
0
The kingside castling option is useful sometimes, as here where
White was planning to mobilise his queenside. White's moves
don't make so much sense when the king runs away.
13 0-0 Bf5 14 a3 Ng4 15 Qa4 Nce5 16 Rad1 Bd7 17 Qc2 Nxf3+
18 Bxf3 Qh4 19 h3 Ne5 20 Bxb7 Bxh3 21 gxh3
Not 21 Qe4 Bg4! 22 Qxe5 Bd6 23 g3 Bxe5 24 gxh4 Bxd1 25 Rxd1
Bxc3 26 bxc3 Rd7 and in view of the coming ...Rfd8 White will
lose a bishop and remain the exchange down - 27 Bc6 Rd6-+.
21...Qxh3 22 Bg2? This loses at once. The critical
line was 22 Qe4 Rd6 23 Ne2 which is unclear according to CC-
grandmaster Hermann Heemsoth, in Fernschach.
22...Nf3+! 23 Bxf3 Bd6 0-1.
McLardy - Timperley corr (England), 1965
1 d4 d5 2 c4 e6 3 Nc3 c5 4 cxd5 cxd4 5 Qa4+ Bd7 6 Qxd4 exd5 7
Qxd5 Nc6 8 e3 Nf6 9 Qb3 Bc5 10 Nf3 Qe7 11 Be2 0-0-0 12 Bd2
Bf5 13 0-0 g5! 14 Nd4 Nxd4 15 exd4 Rxd4 16 Bxg5?
Accepting the pawn is a blunder. White should have played 16 Be3
Rd7 (16...Rb4 17 Nd5) 17 Bxc5 Qxc5 18 Rac1.
16...Rg8 17 Bxf6 Qxf6 18 Bf3 b6 19 Nb5 Rd3 20 Nxa7+ Kb8 21
Nc6+ Kc7 22 Qa4 Rxf3 23 Nb4
Again Black crashes through: 23...Rxg2+!! 24 Kxg2 Rxf2+! 25
Rxf2 Be4+ 0-1
Black wins usually, but not always, come on the kingside. The
following game is reminiscent of the famous "Appointment at
Samara" story where a man meets Death in the market. "Is your
friend not with you today?" asks Death; "No, he had a dream that
he would have an accident so he went to Samara to avoid it." "That
is strange," answers Death, "for tonight I have an appointment in
Samara". The white king here has a premonition of disaster on the
kingside at move 17, so flees to the other wing and meets Death
there!
K.Strand - H. Sabel Norway-Finland corr. 1990
1 d4 d5 2 c4 e6 3 Nc3 c5 4 cxd5 cxd4 5 Qa4+ Bd7 6 Qxd4 exd5 7
Qxd5 Nc6 8 e3 Nf6 9 Qb3 Bc5 10 Nf3 Qe7 11 a3 0-0-0 12 Qc2
Kb8 13 Be2 g514 b4 g4 15 Nh4!? A Belyavsky suggestion.
15...Bb6 16 Bb2 h5 17 0-0-0? 17 0-0 would indeed have been
safer. 17...Rc8 18 Nf5 Bxf5 19 Qxf5 a5 20 b5 Nb4! 21 Kb1 Rc5 22
Qf4+ Bc7 Black is winning for if 23 Qd4 Be5 24 Qd2 Bxc3 25
Bxc3 Ne4. 23 axb4 Bxf4 24 bxc5 Be5 25 Na4 Bxb2 26 Nxb2 Ne4
27 Rc1 Nxf2 28 Rhe1 Rc8 29 e4 Rxc5 30 Rxc5 Qxc5 31 Bc4 Qb4
0-1.
I had better show one White win. Black doesn't have things all his
own way and cannot afford to play inferior variations.
Dr Vojin Savic- Matyas Berta Belgrade-30 jubilee corr, 1979
1 d4 d5 2 c4 e6 3 Nc3 c5 4 cxd5 cxd4 5 Qa4+ Bd7 6 Qxd4 exd5 7
Qxd5 Nc6 8 Nf3 Nf6 9 Qd1 Bc5 10 e3 Qe7 11 Be2 0-0-0 12 0-0 g5
13 b4 Be6?! Declining the pawn does not stop White from
developing an attack. 14 Qa4 Bxb4 15 Nb5 a6 16 Bb2! Bd7 Also
16...axb5 17 Bxb5 gives complications favouring White. 17 Rac1
Kb8 18 Nc7! Kxc7 If 18...Nd4 to attack the queen, then19 Nxa6+
bxa6 (19...Ka8 20 Nc7+ and 21 Qa8 mate) 20 Qxa6 Nxe2+ 21 Qxe2
and White wins eventually. 19 Bxf6 Qxf6 20 Qxb4 Rdg8?! Better
20...Kb8 despite 21 Qb6. 21 Rfd1 Rg6 22 Rxd7+! Kxd7 23 Qxb7+
Ke8 24 Nd4 Nd8 25 Rc8 Qd6 26 Qe4+ Qe7 27 Nc6! Rxc6
28 Qxc6+ Qd7 29 Qxa6 Kf8 30 Qh6+ Ke8 31 Rb8 Qd5 32 Qf6 Rg8
33 Bb5+ Kf8 34 Rxd8+ 1-0
Finally, one more elegant Black win.
Glikshtein - Boris Shkurovich Hazin corr, 1970 1 d4 d5 2 c4 e6 3
Nc3 c5 4 cxd5 cxd4 5 Qa4+ Bd7 6 Qxd4 exd5 7 Qxd5 Nc6 8 e3
Nf6 9 Qd1 Bc5 10 Nf3 Qe7 11 Bc4 An unusual and not very
relevant move. 11...0-0-0 12 0-0 Bg4 13 Qe2 Ne5 14 Bb3 a6 15 e4
Rd3! 16 Nd5 Nxd5 17 Bxd5 Rxd5!? 18 Qc2 If 18 exd5 Nxf3+ 19
gxf3 Qxe2. 18...Rd7 19
b4 Or 19 Nxe5 Qxe5 20 b4 Qxa1. 19...Bxf3 20 gxf3 Rc7 21 bxc5
Nxf3+ 22 Kg2 Qf6 23 Be3 Nh4+ 24 Kg3 Qf3+! 25 Kxh4 Re8 26
Rg1 Rxe4+ 27 Qxe4 Qxe4+ 28 Kh3 f5 29 Raf1 f4 30 Bd2 Qf3+ 31
Kh4 Rc6 0-1.
That was rather a nice attacking game, I thought, but in his
autobiography, Shkurovich Hazin regrets the opportunity he
missed at move 16 to win in even more spectacular fashion. In the
diagram position, he says he should have played 17...Rxf3!. The
variations given by Shkurovich Hazin are a) 18 gxf3 Nxf3+ 19
Kg2 (19 Kh1 Qe5) 19...Nh4+. b) 18 Qc2 Qc7 and now b1) 19 gxf3
Nxf3+; b2) 19 b4 Bd4 20 Qxc7+ Kxc7 21 gxf3 Nxf3+ 22 Kg2
Bxa1 23 Bf4+ Be5-+. b3) 19 Bg5 Rh3!! by analogy
with the famous combination in Pillsbury-Lasker, Petersburg 1896.
20 Bf4 Nf3+ 21 gxf3 (21 Kh1 Rxh2+) 21...Qxf4! 22 Qxc5+ Kb8
and now b21) 23 Rfc1 Qxh2+ 24 Kf1 Qh1+ 25 Ke2 Qxf3+ 26 Kd2
Qe2 mate; or b22) 23 Rfd1 Qxh2+ 24 Kf1 Qh1+ 25 Ke2 Qxf3+ 26
Kd2 Qe2+ 27 Kc1 Qxd1 mate; or b23) 23 Rfe1 Qxh2+ 24 Kf1
Bxf3 and mates on h1.