The following snappy game was played in the Allegheny Futurity tournament in
August.  It is an example of a kind of game that one sometimes sees at the
highest levels, in which the outcome depends on a "theoretical discussion" in
the opening.  While Martinak's defeat may have been partly caused by a
tactical error at move 23, the real problem that his "book" was not as up to
date as Eidemiller's "book".  Characteristically, Mark had specifically
prepared for this game the night before by brushing up.  Not an easy opponent
to face!

White: Mark Eidemiller
Black: Tom Martinak
Allegheny Futurity, August 1996
Queen's Gambit Accepted

 1 d4    d5
 2 c4    dxc4
 3 Nf3   Nf6
 4 e3    Bg4

   A variation of particular interest to me.  The first time I played Kimball
   Nedved, in the early 1970's, I lost with the White pieces in this variation.
   Not long afterwards, I started playing it with the Black side.  I won some
   nice games, but eventually gave it up in favor of the relative safety of the
   4 ... e6 line.

 5 Bxc4  e6
 6 Nc3

   Black answers 6 Qb3 best by giving a gambit:  6 ... Bxf3 7 gxf3 Nbd7!
   8 Qxb7 c5 with surprisingly strong compensation.

 6 ...   Nbd7

   An important alternative is 6 ... a6, planning 7 ... Nc6.  White can allow
   this, or he can play 7 h3 Bh5 8 g4 Bg6 9 Ne5 Nbd7 10 Nxg6 hxg6 11 Bf1! and
   12 Bg2.

 7 h3    Bh5
 8 O-O   Bd6

   Safer is 8 ... Be7.  A famous game in this variation, Larsen-Spassky,
   Santa Monica 1966, continued 9 e4 O-O 10 Be3 Bg6 11 Bd3 c6.  Black must play
   passively, but is well prepared to defend, in a position resembling some
   variations of the Slav Defense.

 9 e4

   The threat of 10 e5 starts a long forced sequence.

 9 ...   e5

   diagram

      r2qk2r
      pppn1ppp
      3b1n2
      4p2b
      2BPP3
      2N2N1P
      PP3PP1
      R1BQ1RK1

10 g4

   The tame alternative 10 Be2 gives White approximately nothing.

10 ...   Bg6
11 dxe5  Nxe5
12 Nxe5  Bxd5
13 f4

   Now White threatens to trap the bishop on g6.

13 ...   Qd4+
14 Qxd4  Bxd4+
15 Kh2   Bxc3

   Not 15 ... Bxe4 16 Nxe4 Nxe4 17 Re1.  Also 15 ... h6 or 15 ... h5 would save
   the bishop, but White would get an overwhelming position.

16 bxc3  Bxe4
17 g5    Bd5

   If 17 ... Nd7 18 Re1 wins.  Martinak writes, "I was familiar with this
   position.  When I first started playing the QGA I considered this a key
   position.  This was a long time ago (I remember analyzing with Jeff Gabel).
   At that point the line was 18 gxf6 and the book evaluation was better for
   White.  However we found several improvements and thought Black was better.
   I was not familiar with 18 Re1+ and I don't remember if we ever analyzed it."

18 Re1+  Kd7

   18 ... Kf8 occurred in Ftacnik-Matulovic, Vrsac 1981.  Now 19 gxf6 Bxc4
   20 f5!?, threatening 21 Rg1 and 22 Ba3+, as recommended in MCO-13, appears
   dangerous.  Martinak was not familiar with this analysis, but rejected
   18 ... Kf8 anyway.

19 Bd3   Ne8

   19 ... Rae8 (or 19 ... Rhe8) only makes matters worse after 20 Rd1 Ng8
   20 c4 Bf3 (or 21 ... Bc6 22 Bf5+ Ke7 23 Ba3 mate) 21 Be2+.

20 c4    Bf3

   Not 20 ... Be6 21 f5.  Also, 20 ... Bc6 appears to close off an important
   escape route:  21 Ba3 Nd6 22 c5! Nb5 (or 22 ... Bb5 23 Rad1) 23 Bf5+ Kd8
   24 Rad1+, winning.

21 Re3   Bh5
22 Bf5+  Kc6
23 Be4+

   diagram

      r3n2r
      ppp2ppp
      2k5
      6Pb
      2P1BP2
      4R2P
      P6K
      R1B5

23 ...   Kd7

   Martinak writes, "After the game we couldn't find a clear win after
   23 ... Kc5, but it looks like it must be there.  However I had to do it, as
   in the game I lose quickly."  Unfortunately I cannot settle this, as I have
   not had time to analyze 23 ... Kc5 either.  I would encourage readers to
   check this out.  Indeed, it looks dangerous for Black to have his king on
   the fourth rank, but in chess tactics, anything goes.

24 Rd3+  Ke6
25 f5+   Ke5
26 Bxb7  Nd6

   Or 26 ... Rb8 27 Bb2+ Kxf5 28 Rf1+ Kxg5 29 Rg3+, etc.

27 Rd5+  Ke4
28 Rxd6+ Resigns