The Kibitzer 
by Tim Harding

Is The Urusov Gambit Sound?

Having dealt last month with preliminary digressions from the
main move order, it is time now to examine the position arising
from

1 e4 e5 2 Bc4 Nf6 3 d4 exd4 4 Nf3 (or 2 Nf3 Nf6 3 d4 exd4 4 Bc4)
4...Nxe4 5 Qxd4 Nf6 6 Nc3 Be7 7 Bg5 (See Diagram)

Black has two (very different) principal defenses to the gambit:
7...c6 followed by...d5 to occupy the centre and, in some cases,
castle queenside; 7...Nc6 followed by...d6, trying to catch up on
development to neutralise the white initiative.

Black can of course castle at once by 7...0-0 but this generally
transposes to lines below. By committing his king so early, Black
loses flexibility and makes it easier for White to decide on his plan.

The line 7...c6 8 0 0 0 d5 was introduced in last month's Kibitzer.
Even at move 9, castling may be premature for Black. Last month
we saw the consultation game Schlechter versus Neustadtl and
Tietz, in which Black defended by 9 Rhe1 0-0 10 Qh4 h6?
overlooking a powerful combination, 11 Bxd5!. Black does have
other tenth moves here, but they have been little explored:

a) 10...Bf5 seems good for White; it is covered in the final game of
this article.

b) 10...Nbd7 is inferior as after 11 Bd3 Black cannot transpose
back to main lines because his queen's bishop is hemmed in. Now
11...h6?! is risky because of 12 Bxh6 gxh6 (12...Ne4? 13 Qg4), a
trick which would not be possible if Black had not castled. After
13 Qxh6 Re8 14 Ng5 (14 Re3) 14...Nf8 15 Re3 White has a
dangerous attack; even if it is not quite clear; I doubt if you would
like to be the defender here! So after 10...Nbd7 11 Bd3 the move
11...g6 has been tried: 12 Ne5 Nc5! Mieses-von Holzhausen,
Duisburg 1929. The best reply seems to be 12 Re2! when White is
clearly better, according to Keres (12 Nd4!? was another Keres
suggestion), e.g. 12...Re8 13 Rde1 Ne4 14 Nxe4 dxe4 15 Rxe4 was
the end of my old notes, but ECO has a pretty line credited to GM
Larry Evans continuing 15...f6 (15...Bxg5+ 16 Nxg5+-) 16 Bc4+
Kg7 17 Qxh7+! Kxh7 18 Rh4+ Kg7 19 Bh6+ Kh7 20 Bf8 mate.

c) 10...Be6 11 Bd3 h6 (11...g6 12 Nd4) is a line which has now
been refuted. Virtually all the games in my database were played
by computers, so evidently this is a line that programmers have
been putting into the openings books of their creations, primarily
trying to catch out other computers! White has to decide where to
put the queen. White is virtually forced to sacrifice the bishop by
12 Bxh6 and since 12...gxh6 13 Qxh6 is hopeless for Black, the
critical reply is 12...Ne4. Old analysis (e.g. by Keres and the late
David Hooper) concentrated on 13 Qf4 but it now seems that 13
Qh5! does the business for White. Then 13...Nxc3? would lose
rapidly to the thematic 14 Bxg7! so there are two main lines:

c1) 13...gxh6 14 Bxe4 Nd7 15 Bf5 Nf6 16 Qh3 Bxf5 17 Qxf5 and
White has a favourable position (safer king, no weaknesses)
without any material disadvantage; and

c2) 13...g6 14 Qe5 Bf6 15 Qf4 with tremendous complications that
readers can enjoy analysing for themselves, e.g. 15...Nxc3 16 Rxe6
fxe6 17 Qg4 Nxa2+ 18 Kb1 Rf7 19 Kxa2 Kh8 20 Qxg6 Qg8
(20...Qa5+ 21 Kb1 Qb4 22 Bg7+! Rxg7 23 Qe8+ and wins) 21
Qh5 and White should soon regain the sacrificed exchange with a
very good game.

In view of lines like this, Black normally defers castling to keep
White guessing about where the king will go.

From the first diagram, play is more likely to continue 7...c6 8
0 0 0 d5 9 Rhe1 (9 Qh4 almost always transposes, e.g....Be6 10
Bd3 Nbd7 11 Nd4 Nc5 12 Rhe1 reaching the next diagram
position) 9...Be6 (Not 9...dxc4?? 10 Qxd8 mate!) 10 Qh4 (or 10
Bd3 Nbd7 transposing) 10...Nbd7 11 Bd3 Nc5 12 Nd4 which
brings about another position for analysis. (See Diagram)

In my 1970s book on the Bishop's Opening, I quoted Keres as
writing that White now has a threatening position but it is not clear
whether he can force more than the return of his pawn.

I gave this position to Chessbase 7.0 to try its new Opening Report
feature.

It found 33 games with this position in my very large
(approximately 1.5 million game) reference database and the
program's analysis included the following points:

* White scores well (65%);

* Short draws are rare (none under 20 moves);

* Draws are of average length (circa 38 moves); and

* Wins by both sides are shorter than average.

Doing this kind of analysis, it is always important to remember
that it is only "lies, damned lies and statistics". What ChessBase is
doing here is just a modern version of Ulvestad's "Chess Charts"
of the 1940s. The program can only work reasonably well if it is
analysing a large and representative database, with games by
strong players. If a game is duplicated in the database, it distorts
the results. Games by weak players or decided by blunders (or on
time against the run of play) are also highly misleading.

Even when your database is both large and clean   so that the
analysis can be at its most revealing   you have to remember that
one strong move can be a counter-example that blows a hole in any
sample based on previous play. What is most convenient about this
Opening Report feature of Chessbase 7 is that it gives access to the
games in your database in a convenient way.

So far as moves from this position are concerned, there is little
choice, since 12...h6 carries no threat and is powerfully answered
by 13 Nf5!. Black must try to relieve the pressure by exchanges
and 12...Nfd7 is unreliable. As the ChessBase report correctly
informed me, Black "scores miserably". Tartakower's move here
was 13 f4 but Estrin's plan 13 Bxe7 Qxe7 14 Qxe7+ Kxe7 15 f4! is
much stronger and after 15...Nxd3+ 16 Rxde3 (Larsen even
suggested 16 cxd3!? once.) Black has tried unsuccessfully: 

a) 16...g6 in the well-known miniature Neishtadt-Volkovich,
USSR 1958. 17 g4 Nc5 18 Rde3 Kd6 19 b4 Ne4 20 Nxe4+ dxe4
21 Nxe6 fxe6 22 Rxe4 1-0. Black resigned as he cannot defend his
e-pawn (22...Rae8 23 f5); and

b) 16...Nc5 was seen in Hausner-Weber, 1992, which went 17
Rde3 g6 18 b4 Na6 19 Nxe6 fxe6 20 Rxe6+ Kd8 21 Re7 with the
better ending for White 21...d4 22 Na4 b6 23 a3 Nc7 24 Rg7 Ne8
25 Rge7 Nd6 26 R1e6 Re8 27 Rxe8+ Nxe8 28 Rxc6 and the extra
pawn was duly cashed in. 

12...Kd7!? has been seen. (See Diagram)

At first sight the move looks rather like desperation, even if there is
no immediate refutation. However, the point is that ...h6 next move
will really threaten...hxg5 because the h-pawn is not pinned (the
rook being defended by the queen) and this will force White to
retreat his advanced pieces or enter murky complications. If you
are looking for a defence for Black this is worth investigating
further, to judge from this game which you should analyse for
yourself

L.Lepre - U, Nyffenegger, corr, 1992 1 e4 e5 2 Bc4 Nf6 3 d4 exd4
4 Nf3 Nxe4 5 Qxd4 Nf6 6 Bg5 Be7 7 Nc3 c6 8 0-0-0 d5 9 Rhe1
Be6 10 Qh4 Nbd7 11 Bd3 Nc5 12 Nd4 Kd7!? 13 f4 (Maybe White
should prefer 13 Nxe6 or 13 Bf5 h6 14 Bd2 but I see nothing
forcing.) 13...h6 14 Bxf6 Bxf6 15 Qf2 Nxd3+ 16 Rxd3 g6 17 Ne4
b6 (This prevents Nc5+, after which the black king is surprisingly
invulnerable in the centre.) 18 f5 gxf5 19 Nxf5 Rg8 20 c4 Rg4 21
g3 Be5 22 h3 Rxe4 23 Rxe4 Qg5+ 24 Ne3 Bxg3 25 Qf3 Rd8 26
Kc2 Ke8 27 cxd5 cxd5 28 Nxd5 Rxd5 29 Qxg3 Rc5+ 30 Kd1
Qc1+ 31 Ke2 Rc2+ 32 Kf3 Qf1+ 33 Ke3 Re2+ 34 Kd4 Qf6+ 35
Qe5 Qd8+ 36 Kc3 Qc8+ 0-1.

Generally 12...Ng8 is considered the critical defence. (See
Diagram)

Now the double exchange on e7 is ineffective because at move 14
Black takes with the knight. It is possible to exchange only the
bishops and then play (13 Bxe7 Qxe7) 14 Qg3 g6 15 Nce2 with
compensation, but instead 13 f4 (not mentioned in ECO) is
probably best, when:

a) 13...Kf8?! 14 b4 Nxd3+ 15 Rxd3 with advantage to White
(C.Torre-Tholfsen, New York 1924) was a line from my 1970s
book. An old "MCO" continued 15...Qd6 here which is not entirely
clear.

b) In my book I said that Black can defend by 13...Bxg5 14 fxg5
Ne7 (as suggested by Pachman and approved by Larsen. This
remains untested. White could then capture on h7 (intending to
continue b4 and, when the knight retreats, Rxe6 followed by huge
complications) but the self-pin of the bishop against the white
queen is unattractive to say the least. Instead of 14...Ne7 Black
may be able to get away with 14...h6 as 15 Nxe6 Nxe6 is
unconvincing. I saw 16 Nxd5 in an old game Giertz-Kornetzky but
it looks highly unsound! However, 15 Bg6!? is interesting with a
long forced variation leading to a position that is hard to
assess15...Kd7 16 Bxf7 Bxf7 17 Qg4+ Kc7 18 Qf4+ Kc8 (18...Kb6
19 Nf3!) 19 Qxf7 Qxg5+ 20 Kb1 with compensation for the pawn,
but is it just enough or can White stand better?

Even the young Karpov underestimated the dangers of the gambit.
G.Timoschenko-Karpov, USSR Junior Ch, Moscow 1969, went 1
e4 e5 2 Bc4 Nf6 3 d4 exd4 4 Nf3 Nxe4 5 Qxd4 Nf6 6 Bg5 Be7 7
Nc3 c6 8 0-0-0 d5 9 Rhe1 Be6 10 Qh4 Nbd7 11 Bd3 and now
11...c5!? 12 Ne5 Nxe5 13 Rxe5 d4 14 f4 Nd7 15 Bb5 Bxg5 16
fxg5 Qc7 17 Bxd7+ Kxd7 18 Qe4 Qc6? 19 Rxc5 and White won in
54 moves. The "Encyclopaedia of Chess Openings" gives this as
its main line for the Urusov, saying that White has compensation
for the pawn after his18th move; however, there seem to have been
no other willing takers for Karpov's 11th move.

Going back to the beginning, the alternative defensive line 7...Nc6
leads to a different kind of struggle. I would tend to prefer this
move as Black because it gets a piece into play whereas 7...c6
impedes development. After 8 Qh4 we get to this position: (See
Diagram)

The first thing to realise here is that while 8...d5 looks wrong
(since the d-pawn lacks protection) this has never actually been
proved. After 9 0-0-0 Be6 10 Rhe1 there are these examples: 

a) 10...h6 11 Bxf6 Bxf6 12 Qh5 Bxc3? (Black may only be a little
worse if he avoids this.) 13 Rxe6+ Kf8 14 Rxd5 etc. (Tereschenko-
Rotlevi, St Petersburg 1909);

b) 10...Qd7 11 Bb5 0-0 12 Bd3 h6 13 Bxh6 gxh6 14 Qxh6 Bf5 15
Ne5 Nxe5 16 Rxe5 Ng4 17 Bxf5 Qxf5 18 Qxf8+ Bxf8 19 Rxf5
Bh6+ 20 Kb1 1-0 Zarske-Hoffmann, Zurich open, 1993.

c) 10...0-0! 11 Bd3 h6 is really far from clear. In my 1970s
Bishop's Opening book I incorrectly gave now 12 Rxe6 fxe6 (not
12...hxg5? 13 Nxg5 Nh5 17 Rxe7 and White does win) 13 Bxh6
"for if then 13...gxh6? 14 Qg3+ Kh8 15 Qg6 gives White a
winning attack". Rubbish! as GM Bent Larsen soon pointed out to
me. The correct assessment is equal, probably an immediate draw.
In addition, Black can consider 13...Nb4.

Also 12 Bxh6 does not work because of 12...Ne4, so after 11...h6
the game Hmelnitsky-Eventov, USSR 1956, went 12 Kb1 Qd7? 13
Bxh6 Ne4 14 Bg5 Bxg5 15 Nxg5 with a good game for White.
However at move 12 Black has various possible improvements
such as 12...Ne8 (Euwe) and 12...Re8 (Wallinger-Beutel, corr
1991) which need proper testing.

Usually, however, Black has played 7...Nc6 with the idea of
continuing 8 Qh4 d6 9 0-0-0 Be6 10 Bd3 reaching the next
diagram. (See Diagram)

Now if 10...h6 Neishtadt won a game in a simultaneous by 11
Bxh6 but this should only yield a draw after 11...gxh6 (11...Ng4 12
Bg5) 12 Qxh6 Ne5 13 Nxe5 dxe5 14 Qg5+ Kh8 15 Bf5
(threatening Rd3) 15...Qe8.

White should therefore play 11 Rhe1 Qd7 (Euwe) 12 Bb5 0-0-0
which, back in the 1970s Larsen told me was unclear. However,
Black can easily get massacred in practice:

E.Svensson-H.Tolksdorf European master class corr. 1973 1 e4 e5
2 Bc4 Nf6 3 d4 exd4 4 Nf3 Nxe4 5 Qxd4 Nf6 6 Nc3 Be7 7 Bg5
Nc6 8 Qh4 d6 9 0-0-0 Be6 10 Bd3 h6 11 Rhe1 Qd7 12 Bb5 0-0-0
13 Qa4 Nd5? 14 Rxd5! Bxg5+ (Or 14...hxg5 15 Nd4 Bxd5 16
Nxd5 Bf6 17 Bxc6 1-0 J.Uschold-M.Wallinger, West German corr
1985.) 15 Rxg5 hxg5 16 Nd4 Rdg8 17 Nxc6 bxc6 18 Bxc6 Qe7 19
Qxa7 1-0.

From the diagram, the main line is 10...Qd7 which allows the pin
11 Bb5. Now Black has to give up the idea of queenside castling
because the white queen can switch wings in a mirror of the kind
of attacks that take place on the kingside in other variations: 11...0-
0-0? 12 Qa4 a6 13 Bxa6 bxa6 14 Qxa6+ Kb8 but now not 15 Nb5
(Keres) 15...Nd5; however, 15 Be3! seems quite promising.

So after 10...Qd7 11 Bb5 comes 11...0-0 and now 12 Ne5 (one of
the points of Bb5 in these lines) has been known since a 1924
British postal game Griffith-MacDonald.

However 12...Qc8 13 Nxc6 bxc6 is an improvement. Then 14
Bxc6 (14 Bd3!?) when:

a) 14...Rb8 15 Ne4 Qd8 16 Rd3 Kh8 17 Nxf6 Bxf6 (17...gxf6?? 18
Be4+-1-0 Matrisch-Ninkovic, corr 1973) 18 Bxf6 with a slight
advantage to White;

b) 14...h6 15 Bd2 [15 f4 hxg5 16 fxg5 Nh7!] 15...Rb8 is reckoned
to be equal, following Neishtadt-Burlyaev, Candidate masters
tournament, Moscow, 1958. 16 Qa4 Ng4 17 Be1 soon led to a
draw. However, Estrin and Glazkov in their book "3 Double King
Pawn Openings" (Chess Enterprises, 1982) suggested 17 Rdf1!
Ne5 (17...Bf6 18 f4!) 18 Nd5 here. 

In fact, they preferred 12 Nd4 to 12 Ne5, partly because against the
Keres suggestion 12...h6 they had prepared the sacrifice 13 Bxh6
gxh6 14 Qxh6 Ng4 15 Qh5. So Black should play 12...a6 following
Mieses-Rubinstein, Breslau 1912, when after 13 Bd3 Black's safest
course would have been 13...h6 14 Bxh6 (14 Ne4 Nd5! or 14 Nxc6
bxc6 15 Bxh6 Ng4) 14...gxh6 15 Qxh6 Nxd4 16 Qg5+ Kh8 17
Qh6+ Kg8 18 Qg5+ with perpetual check (Estrin and Glazkov).

Summary

Whether we take the stem position after 7 Nc3 or major
subvariations from a later point, the attraction of the Urusov is
clear -

* White scores well (65% or better) whatever Black's 7th move;

* White wins are shorter than average;

* Draws are rare.

On top of this, consider that the gambit generally involves the kind
of open tactical play which doesn't require you to be a
grandmaster. Positional subtleties revolve around assessing
whether White has sufficient dynamic compensation, not on
whether an endgame that could arise will be decisive. The gambit
can also confuse the judgment of programs like Fritz. All in all, it
is a good bet below master level but remember: Preparation pays!

Until you get into the details, it is really hard to believe that the
Urusov Gambit can be sound against correct defence. The first
impression is that Black has made no major structural concession
and should be able to catch up on development without falling into
a trap. That may be true at the highest level (where the gambit is
never played!) but at the level where most mortal chess players
operate it has a good chance of success. 

Against a computer, it is probably not a good bet as they tend to be
ingenious defenders, but private games against computers can help
to prepare you for games with human opponents.

Here is one last game, sent to me by the winner after he read the
previous Kibitzer column.

Juergen Zarske (Switzerland)-H. Schneider corr. 1993

1 e4 e5 2 Bc4 Nf6 3 d4 exd4 4 Nf3 Nxe4 5 Qxd4 Nf6 6 Bg5 Be7 7
Nc3 c6 8 0-0-0 d5 9 Rhe1 0-0 10 Qh4 Bf5 

Unusual, but there is a precedent.

11 Nd4

Another correspondence game with the same player as Black
went11 g4!? Bd7 12 Bd3 h6 13 Bxh6 Ne4 14 g5 gxh6 15 Rg1
Nxc3 16 Bh7+ Kxh7 17 Qxh6+ Kg8 18 g6 1-0 R.Mercurio-
H.Schneider, corr. 1996.

11... Bg6 12 Bd3 Qd7 

12...h6 13 Bxg6 hxg5 14 Qxg5 fxg6 15 Ne6 Ne4 16 Qxg6 Bg5+ 17
f4 gave interesting complications in Caro-Janowski, Berlin 1897
(drawn in 81 moves).

13 f4 h6 

Asking for trouble! Now we get a classic Urusov Gambit sacrificial
attack.

14 Bxh6 gxh6 

Not 14...Bxd3 15 Bxg7! Kxg7 16 Rxe7 and Black either gets a lost
ending (after 16...Qg4+) or is mated after 16...Qxe7 17 Rxd3.

15 Bxg6 fxg6 16 Qxh6 Kf7 (See Diagram)

It is hard for Black to choose among the many possibilities here, as
all are at least slightly worse for him. Now 17 Nf3 looks obvious
but Zarske prefers another approach.

17 Ne6!? Rg8 18 g4 Qd6 19 Ne4

Exploiting the pin on the d-file, and not for the last time in this
game. 

19...Nxe4 20 Rxe4 Bf8 21 Ng5+ Kf6 22 Qh4 Qd7 23 Ne6+ Kf7 24
Ng5+ Kf6 25 Re5! 

After the repetition, White tries a new tack. 

25...Na6 26 Ne4+ Kg7 27 Qf6+ Kh7 28 Ng5+ Kh6 29 Ne6 Kh7 30
Nxf8+ and Black resigned since 30...Rxf8 is met by 31 Re7+ with
mate or win of the queen. (1-0)

Copyright 1998 Tim Harding. All Rights Reserved.