The Kibitzer
by Tim Harding

Has The Marshall Attack Been Refuted On The Internet?

The Marshall Counter-Attack in the Ruy Lopez (Spanish Opening) is
facing its biggest crisis in many years - not because of a discovery by a
leading GM or famous theoretician but because of an amateur expert's
home analysis posted in a Usenet newsgroup back in March. 

Six weeks intensive research and analysis have failed to uncover any
hole in the question posed by Daniel Quigley on
rec.games.chess.analysis

(See Diagram) Why in the diagram position does White not continue 19
axb5!! This very strong move just is not in the books, including (alas)
the ones the ones that I was involved in writing.

The diagram position arises after the well-known sequence 1 e4 e5 2
Nf3 Nc6 3 Bb5 a6 4 Ba4 Nf6 5 0-0 Be7 6 Re1 b5 7 Bb3 0-0 8 c3 d5 9
exd5 Nxd5 10 Nxe5 Nxe5 11 Rxe5 c6 12 d4 Bd6 13 Re1 Qh4 14 g3 Qh3
15 Be3 Bg4 16 Qd3 Rae8 17 Nd2 Re6 18 a4 f5.

Black has sacrificed a pawn for some positional compensation and the
initiative. The originator of the gambit, American grandmaster Frank
Marshall, first played the gambit with 11...Nf6 in the famous game he
lost to Capablanca (New York 1918) but 11...c6 was also one of
Marshall's ideas. 

Although 11...Bb7 has also attracted interest, 11...c6 is still reckoned to
be the critical move and the whole line to 17...Re6 became popular when
Spassky started to play it in the early 1960s, en route to the World
Championship. That move was Spassky's patent, replacing direct crude
attacks with 16...f5 or 17...f5 that had not given good results.

If you look this line up in books and databases, you will see that the
move sequence 18 a4 f5 19 Qf1 Qh5 has been taken for granted as the
main line by players and theoreticians alike for over 35 years. Black
does have alternatives at move 18 but the majority of space in books like
the "Encyclopaedia of Chess Openings" (ECO) is devoted to this line.

Many games have gone 18 Qf1 Qh5 19 a4, usually leading to positions
considered in theory books under 18 a4 f5 19 Qf1 Qh5 but Quigley's
discovery strongly suggests that 18 Qf1 is an inferior move and that 18
a4 must be met by a move other than 18...f5.

The logic behind 18 a4 is that opening a second front on the queenside is
the key to White's counterplay and should not be delayed.

In the position after 18 a4, three moves are known for Black: a)
18...bxa4; b) 18...Qh5 c) 18...f5.

This article is almost exclusively concerned with the third of these, but a
few remarks on the others are called for.

18...bxa4 was the move played by Spassky against Novopashin in 1962,
one of the key games with the Spassky attack, and it includes a lot of
traps. However, in recent years the pawn capture has been unpopular,
perhaps because it seems to lose a tempo. The main line in the new ECO
claims some advantage for White after 19 Rxa4 f5 20 Qf1 Qh5 21 c4 and
it may not be possible to overturn that long-standing assessment.

18...Qh5 was Spassky's second idea in the line, introduced in his 1965
candidates match against Tal. The move has a reputation of being
drawish.

These points probably explain why 18...f5 19 Qf1 Qh5 has come to be
considered the main line, in ECO and most books.

Now we come to the point of Daniel Quigley's newsgroup posting, in
which he referred to the 1980s Batsford book by Nunn, in which my
contribution was the chapters on the minor lines (without 11...c6) and
Anti-Marshalls (8 a4 etc).

"In the late '80s, "wrote Quigley, "a master friend of mine bought the
book and we pored over those lines. I remember then suggesting the
busting move. We analyzed and analyzed but could not come to a
definitive conclusion. About a week ago I received the new ECO and I
decided to give the Gambit another look."

Instead of 19 Qf1, he suggested, why should White not reply 19 axb5!!
(See Diagram).

Most books, considering the diagram position, briefly dismiss 19 f4 and
concentrate on 19 Qf1. The move axb5 is not mentioned in John Nunn's
book, nor in most Marshall books, while in my "Spanish Ruy Lopez
Marshall" (1979) I wrote, incorrectly as it now seems, "The threat of
19...f4 followed by ...Rh6 rules out an immediate axb5 for White." 

You can read the whole debate about this at http//www.dejanews.com by
searching for Marshall Busted? which was the title of the thread.

Quigley continued, "I can conceive of only two possible replies by Black,
19...f4 and 19...axb5, both of which I shall now show are inadequate".
He more or less did that, but as we shall see below, there was one major
possibility for Black at move 20 that he did not mention.

Although some other people in reply posted games they had found, and
which I also had in my database, it was clear that none really settled the
issue. A few players had hit on the idea of 19 axb5 but none of these
were high-level master games and in many cases the follow-up was
incorrect.

For example, after 19...axb5 an Italian postal game Corradini-Di
Stefano. 1986, went 20 20 Bd1? Bxd1 21 Raxd1 f4 and Black went on
to win, but Quigley identified the correct continuation for White: 20
Bxd5 cxd5 21 Qxb5! (as in fact occurred in a game Uhlig-Guzman,
Compuserve 1996) 21...f4 22 Bxf4 and now two main lines were
considered in the newsgroup debate

a) 22...Bxf4 23 Rxe6 Bxe6 (23...Bxd2 24 Re8! intending 25 Qxd5+ +-)
24 gxf4 Qg4+ 25 Kh1 Qxf4 26 Qe2 and White has a little more work to
do, but surely the two pawns up should win, wrote Quigley.

b) 22...Rxf4 23 Rxe6 Rxf2 24 Re8+ Kf7 25 Qxd5+ Kxe8 26 Ra8+ Ke7
27 Ra7+ Bd7 28 Qe4+ was suggested by another Usenet poster. 

Quigley replied with various improvements on that line, both at move 28
and at move 24 where he pointed out "White has the possibility of 24
Ra8+ Bf8 25 Rxf8+ Rxf8 26 Re8 and 27 Rxf8+. Black has insufficient
material left to generate any substantial threats and White's three pawn
surplus will win easily."

Now let us look at the most likely reply for Black, namely 19...f4.
Doubtless this is the reason why 19 axb5 has been disregarded in the
past, for example if 20 Bf1?? Bf3-+ or 20 Bxd5? cxd5 21 Qf1 fxe3 22
Qxh3 0-1 (Prelog-Lorbek, Slovenia ch 1991) while my database has
several nasty examples of White losing after 20 bxc6? For example:
20...fxg3 (20...fxe3 and 20...Rh6 have also been successfully played.) 21
hxg3 (21 fxg3 Bxg3 22 Re2 Bf5!-+ or 22 hxg3 Qxg3+ 23 Kh1 Rf2)
21...Bxg3 22 Bf1 Bf3 0-1 Kiltti-Sammalvuo, Tampere 1995. However,
White has a much stronger possibility, indeed a forced but very likely
winning move!

Quigley advocates 20 Bxf4!, saying "Black wins a piece but loses the
game because of White's queenside counterplay." (See Diagram).

Black can capture the bishop in three different ways, but there are three
crucial factors for White in this position:

a) He threatens bxc6 to undermine the black centre and create a
dangerous passed pawn; in some circumstances, bxa6 can also be
considered.

b) The half-pin on the a2-g8 diagonal limits Black's options if the knight
captures on f4 then the e6-rook is pinned, or if the e6-rook moves than
the knight cannot.

c) White has retained the option of playing either Nf1 or Qf1 to defend
against concrete threats according to Black's actual continuation.

The first variation to be dismissed is 20...Rxf4 21 Qf1! (improving on 21
Rxe6 Rxf2 22 Re8+ Kf7 23 Bxd5+ cxd5 24 Re7+ Kf6 25 Re6+ Kf7 26
Re7+ 1/2-1/2 Gorges-Kling, cor Germany 1987) 21...Qh5 22 bxc6 Rff6
23 Rxe6 Bxe6 24 Ra5! Rh6 25 h4+- Quigley. 20...Rxe1+ 21 Rxe1 Bxf4
also gets nowhere: 22 bxc6 Be6 23 Rxe6 Qxe6 24 Qe4 1-0 Uhlig-Clare,
Compuserve cor 1995.

It should also be noted 20...g5? loses in many ways, e.g., 21 Rxe6 Bxe6
22 Bxd6 Rxf2 (22...Bf4 23 Bxe6+) 23 Kxf2 Qxh2+ 24 Ke1 and that
20...axb5? fails to 21 Rxe6 Rxf4 (21...Bxe6 22 Bxd6) 22 Rxd6.

20...Bxf4 takes a bit more work to refute: 21 Rxe6 Bxe6 22 bxa6 (Also
22 bxc6 Rb8 23 Qe4 Bxd2 24 Bxd5 Bxd5 25 Qxd5+ Kh8 26 Rxa6 Re8
27 Ra1 Qg4 28 Qd7 Qe2 29 c7 Rf8 30 Qf7 1-0 A. Uhlig-J.Guzman,
Compuserve casual cor 1996) 22...Bxd2 23 a7 and now Quigley's
original posting dealt with 23...Ra8 24 Qxd2 Bf7 25 c4 Bf6 26 Ra3
intending Bd1 to protect against ...Bg4 and then to activate the Queen.

Somebody then suggested the improvement 23...Bc7 24 Bxe6+ Qxe6 25
Qxd2 after which Black has a knight for 3 pawns. Readers may like to
examine this variation for themselves but if anybody can find a
continuation that is actually satisfactory for Black then I shall be pleased
and surprised to hear about it.

The last main line considered by Quigley is 20...Nxf4 21 Qf1!. (See
Diagram).

Not 21 Bxe6+? Bxe6 22 Qf1 Qh6 23 gxf4 Bxf4 24 Qg2 Bxd2 25 Rxe6
Qxe6 26 bxa6 Ra8 unclear.

Now Quigley analysed 21...Qh5 22 gxf4 Bxf4 23 Qg2 Bxd2 24 Re5!
when he said Black has to trade into a losing endgame and 21...Qxf1+
which also seems hopeless in the long run. There was also a Spanish
postal game from 1987, Gomez Baillo-Alhambra, that he did not know
about, with a third move, which also loses for Black: 21...Qh6 22 gxf4
Qxf4 23 Qg2 Kh8 24 Bxe6 Bxe6 25 Bf1 Bd5 26 Qg4 cxb5 27 Re7 Qf6
28 Qe2 Bf3 29 Qe6 Bxh2+ 30 Bxh2 Qg5+ 31 Kf1 Qg2+ 32 Ke1 Bd5 33
Qe3 Qg1+ 34 Bf1 Bc4 35 Re8 1-0.

There is a final possibility for Black that neither Quigley nor the other
newsgroup contributors considered, one hinted at in the quote from my
1979 book, namely to decline the piece offered by 20 Bxf4 and reply
20...Rh6!? (See Diagram).

At first I thought this might draw for Black, e.g. 21 Bxh6 (21 Bxd5+
cxd5 is probably only a transposition.) 21... Rxf2 22 Bxd5+ cxd5 23
Re8+ Kf7 24 Re7+ (24 Kxf2 Qxh2+ at least draws for Black) 24...Kf8
25 Bxg7+ Kg8 26 Kxf2 (Else there is a draw by repetition.) 26...Qxh2+
27 Ke3 (27 Ke1 Bxg3+ 28 Qxg3 Qxg3+ 29 Kf1 Bh3+ drawing.

Unfortunately 21 Bf1! Rxf4 (best?) 22 bxc6 seems to refute the idea
22...Bf3 23 Qxf3 Rxf3 24 Bxd5+ Rf7 (24...Kf8 25 Bxf3) 25 Re8+ Bf8
26 c7 Rc6 27 Bxc6 Rxc7 28 Bd5+ is hopeless for Black in the long run. 

One of the earliest sources I have dealing with the Spassky line is an
article by Mikenas in "Shakhmatny Byulleten" 6/1963. There 18 a4 is
given a ! but after 18...f5 the possibility 19 axb5 is not even mentioned. I
have found references to an article by Henkin in the "Central Chess Club
Bulletin" 1/1963 and to analysis by Boleslavsky and Suetin in Soviet
publications from 1962 which unfortunately I have been unable to
consult, although I did have access to these in the 1970s. "Chess
Archives" 1963 may also be relevant. Without checking these references
from the early 1960s I cannot be certain that Mr Quigley's idea was not
anticipated but it now seems very likely to me that the line was not
correctly analysed at that time in the USSR, and that possibly 20 Bxf4!
was not seen in those early days of the 17...Re6 variation.

My impression so far is that:

a) 19 axb5 has not been considered seriously by analysts and players; b)
It does indeed seem to refute 18...f5; c) This does not mean that the
Marshall as a whole is refuted but it is a major body-blow that could be
part of an overall refutation in due course.

I wondered how it could be that this strong line for White has been
overlooked, or at least never played. Presumably anybody who thought
about it just assumed that the move had been refuted earlier and did not
take the time to look at in depth, probably stopping after seeing that
White would have to give up a piece.

Looking again at the sources still available to me, I notice two salient
points:

a) that Spassky never played 18...f5, so he was possibly aware all along
of the problem with that move order but never told anybody. 

b) Most of the early games cited in books as arising via 18 a4 f5 19 Qf1
Qh5 did not in fact take that course; the move order was 18 Qf1 Qh5 and
only then 19 a4 f5!

In practice, the move order given as the main line in all the books,
including mine, seems hardly ever to have occurred in GM play in the
early years of the variation - although an exception is Tal-Geller, USSR
Ch 1975 (Informator 20/322). That shows that if a flaw in 18...f5 had
been detected by some players it was not widely known. Those two were
after all among the top GMs in the world at the time. Their game
continued 19 Qf1 Qh5 and ended in a draw.

An article by A.C. van der Tak in "Schaakbulletin" in 1976 (between my
two books) took 18 a4 f5 19 Qf1 Qh5 as a standard starting point; it
seems that everybody from then on either did not consider 19 axb5 or
just assumed it must have been refuted.

To answer the question in the title, Quigley's discovery is not in itself a
refutation as Black can answer 18 a4 by 18...bxa4 or 18...Qh5 or diverge
at an earlier stage, such as Michael Adams' move 17...Qh5. However, I
should like to see 17...Re6 18 a4 f5 19 axb5 tested in a grandmaster
game or two, and then I think the truth would soon be revealed. 

Maybe OTB masters and GMs are too busy to read newsgroups, as there
seem to have been no games in this line since Quigley's idea was first
published, but some masters may drop by to kibitz Kibitzer. So any
reader who plays White in the Marshall should definitely try this idea
now and score a few points with it before it becomes better known!