Computer Chess Club Archives


Search

Terms

Messages

Subject: Glaurung in Mainz: Part 1 (very long)

Author: Tord Romstad

Date: 15:28:25 08/13/05


I came back from the FRC world championship earlier tonight.  In the
hope that somebody will find it entertaining, I post this diary of my
experiences from Mainz.  I hope that many of the other participants in
Mainz and Reykjavík will do the same.

This is part 1 of the diary, which covers the first four rounds.  I'll
post part 2, covering the last three rounds, some time tomorrow.

Day 0 (2005-08-10):

After a long and tiring journey, I finally arrive in Mainz on
Wednesday night, and manage to find the hotel and the playing hall
without any particular adventures.  To my disappointment, the other
programmers have already left, but Mark Vogelgesang (who did a
flawless job in organising the tournament) is there to wish me welcome
and give me one of the Windows XP computers supplied by the hardware
sponsor.  I am relieved to see that the computer accepts my USB memory
stick, and that installation of Glaurung and the Shredder Classic GUI
is unproblematic.  After spending a couple of minutes scratching my
head while trying to find out how to change the main engine in the
Shredder GUI, I finally discover the trick and verify that Glaurung
works correctly.  I was a bit worried that my total lack of experience
with Windows in general and the Shredder GUI in particular would be
a problem, but so far everything seems OK.

Day 1 (2005-08-11):

In the elevator on my way to breakfast, I finally meet the first
programmer: Jaime (Ayito).  During breakfast we are joined by several
others, including Anastasios (AICE), Franck (Pharaon) and Richard
(Baron).  Everybody turns out to be even nicer in person than online.
Breakfast is followed by a short player's meeting discussing the
tournament rules, and shortly after the pairings for round 1 are
presented.  Glaurung is seeded unexpectedly low (number 11 of 19), and
is paired against second seeded List in the first round.  I don't
complain, because white against one of the favorites is never a bad
way to start a tournament.

Round 1: Glaurung - List
[D]bqnrkbnr/pppppppp/8/8/8/8/PPPPPPPP/BQNRKBNR w HDhd - 0 1

List has always been a somewhat mysterious and enigmatic engine to me,
because the author is not active in any of the online computer chess
forums.  It was therefore a great pleasure to meet Fritz Reul in
person and have the opportunity to ask a few questions about his
engine.  Fritz is a modest, soft-spoken, friendly and intelligent
young man.  For those who are interested in news about List, I can
tell you that the program has been completely rewritten since version
5.12.  Fritz is not sure how big the improvement in strength is.
There are currently no plans of a new public version.

Like most of the favorites, List did not use a PC from the hardware
sponsor, but some fast Athlon computer (I don't remember exactly).
List is absurdly fast, searching about 2.5 million nodes/second in the
middle game, IIRC.

1. e3 b5 2. b4 e6 3. Nf3 Bxf3 4. gxf3 Nge7 5. Rg1 Ng6 6. f4 f6 7. a3
Nh4 9. Bh5+ Ke7 10. Ne2 Nf3+ 11. Bxf3 Qxf3 12. d3 Qh3 13. c4 Qxh2
14. cxb5
[D]2nr1b1r/p1ppk1pp/4pp2/1P6/1P3P2/P2PP3/4NP1q/BQ1RK1R1 b GD - 0 14
Both programs are very happy about the outcome of the opening, and
believe they have a clear advantage.  Glaurung is optimistic because
of its superior mobility, List probably prefers black because of
white's somewhat compromised pawn structure and lack of a good place
for the king.  I suspect that List's evaluation is closer to the
truth.

14... Nb6 15. Rc1 Rc8 16. Qb3 Kf7 17. f5 Bd6 18. fxe6+ dxe6 19. e4 Qh3
20. Rc3 Qh5 21. Rc6 Rce8 22. Qc3 Nd7 23. Nd4 Qh2 24. Nf3 Qh6 25. d4
Nb6 26. Qc2 Bf4 27. a4 Re7 28. a5 Nc8 29. Rc5 Bd6 30. Rc4 Qh5 31. Qb3
Bf4 32. Rc5 Qh6 33. Rg4 Bd6 34. Rc4 Rd7 35. e5 fxe5 36. Rh4 Qg6
[D]2n4r/p1pr1kpp/3bp1q1/PP2p3/1PRP3R/1Q3N2/5P2/B3K3 w - - 0 37
37. Kf1!?
I expected 37. dxe5, which was Glaurung's preferred move until it
changed its mind at the last moment.  At this stage of the game I was
rather pessimistic.  It looks like black will soon be able to complete
an artificial kingside castling, and white's king will be in great
danger.  Glaurung, however, was happier than ever before.  The score
after Kf1 was +0.89.  I didn't understand the move Kf1 at all, and I
still don't.  In just a few moves white develops a deadly attack, but
I have no idea whether this is because Kf1 is a strong move or because
black makes some mistake at this stage.

37... Rhd8 38. d5 Kg8 39. Rcg4 Qf5 40. Rg5 Qf6 41. dxe6 Re7 42. Qd3
[D]2nr2k1/p1p1r1pp/3bPq2/PP2p1R1/1P5R/3Q1N2/5P2/B4K2 b d - 0 42
White is winning.  The rest is simple technique.
e4 43. Bxf6 exd3 44. Rhg4 g6 45. Rd5 Rde8 46. Bxe7 Nxe7 47. Rxd3 Rb8
48. b6 axb6 49. a6 b5 50. Nd4 Ra8 51. Nxb5 Rxa6 52. Rc4 Kg7 53. Nxc7
Bxc7 54. Rxc7 Rxe6 55. b5 Kf6 56. Rdd7 Nf5 57. Rxh7 Nd4 58. Rhd7 Nf5
59. Rc6 1-0

This first game was in my opinion Glaurung's most interesting game in
the whole tournament.  I hestitate to say that it was the best game,
for the simple reason that I don't understand the game at all.  I have
no idea whether Glaurung's position was really sound throughout the
game (the eval was always positive), and where List's decisive mistake
is.  It would be very interesting if some strong player could give a
few comments about this game.

At any rate, the game was highly entertaining, and a very
characteristic example of Glaurung's style.  A great start of the
tournament!

Round 2: Ant960 - Glaurung
Hans Secalle, the author of Ant, is a strong chess player and an even
stronger shogi player.  I learned a lot from his insightful comments
during the game.  Ant is not a very famous program, but it managed a
draw (and even came close to winning) against Sjeng in the first
round, and I expected a hard fight.

Ant secured a small edge in the opening, and matters got worse when
Glaurung played the dubious 23... g5?!.  Glaurung's weak pawn on b6
remained an annoying problem for a very long time, and I felt very
unhappy about the position.  Fortunately, Ant never managed to find
anything concrete.  Eventually, the game simplified into a
drawish-looking endgame where both sides were in time trouble and
probably made numerous mistakes.  After various adventures Glaurung
ended up a pawn up, but just a few moves later Ant found a tablebase
draw.  It is possible that Glaurung could have made white's defence
more difficult if it had tablebases itself, but I think the result
would still have been a draw.

[D]bqrnnbkr/pppppppp/8/8/8/8/PPPPPPPP/BQRNNBKR w HChc - 0 1
1. e3 b6 2. Bb5 Nc6 3. b4 e6 4. Bc3 Be7 5. O-O O-O 6. Qb3 Nd6 7. Be2
Ne4 8. Bf3 Nxc3 9. dxc3 a5 10. b5 Ne5 11. Bxa8 Qxa8 12. Nb2 c5 13. c4
d6 14. Ned3 Ng6 15. Rce1 Bf6 16. Na4 Qb7 17. Re2 d5 18. f3 Rfd8 19. a3
Rd7 20. c3 dxc4 21. Qxc4 Rcd8 22. Ndb2 Ne5 23. Qb3 g5 24. Qc2 Qc7
25. f4 gxf4 26. exf4 Ng4 27. h3 Nh6 28. Nc4 Nf5 29. Ref2 Rd3 30. Rf3
Rxf3 31. Rxf3 Nd6 32. Nxd6 Rxd6 33. Kh2 Rd5 34. Nb2 Bg7 35. Nc4 Qb8
36. Qe4 Qd8 37. Rg3 Kh8 38. Qf3 Bf6 39. Qe2 Qb8 40. Qe4 Bh4 41. Rf3
Qc7 42. Re3 Bf6 43. a4 Bg7 44. Qf3 Kg8 45. Qg4 h5 46. Qf3 Kf8 47. g3
Bf6 48. Qf1 Qd8 49. h4 Kg7 50. Qe2 Kh7 51. Kh3 Kh6 52. Rf3 Kg7 53. Qe4
Rd1 54. Re3 Rd5 55. Kh2 Qb8 56. Qg2 Qc7 57. Kh3 Kh6 58. Qe4 Qd8
59. Qf3 Qb8 60. Qf1 Qc7 61. Rf3 Qb7 62. Kh2 Kg6 63. Qg2 Qc7 64. Kh3
Kh7 65. Qe2 Kh6 66. Ne3 Rd8 67. f5 Qe5 68. fxe6 Qxe6+ 69. Rf5 Kg7
70. Qxh5 Rd3 71. Ng4 Bxc3 72. Rg5+ Kf8 73. Qh7 Rd1 74. Qg8+ Ke7
75. Qa8 Rd8 76. Qf3 Bd4 77. Qf4 Kf8 78. h5 Qd6 79. Qxd6+ Rxd6 80. h6
c4 81. h7 Bg7 82. Kg2 c3 83. Rh5 Bh8 84. Ne3 Rd2+ 85. Kf3 Kg7 86. g4
Rd3 87. Ke4 Rd1 88. Nf5+ Kf6 89. Rh6+ Kg5 90. Rc6 Ra1 91. Ne3 Rxa4+
92. Kf3 Rf4+ 93. Ke2 Re4 94. Kf3 Re8 95. Nf5 f6 96. Rxc3 Ra8 97. Rc1
a4 98. Ra1 a3 99. Ra2 Ra5 100. Ne3 Kg6 101. Nc4 Rxb5 102. Rxa3 Kxh7
103. Ra7+ Kg6 104. Nd6 Rb3+ 105. Kf4 Bg7 106. Nf5 Bh6+ 107. Ke4 Bf8
108. Ra8 Rb4+ 109. Kf3 Kf7 110. Rc8 Bc5 111. Nh6+ Ke7 112. Rc6 Rb1
113. Nf5+ Kf7 114. Ke4 Rb4+ 115. Kf3 Ra4 116. Nh4 Bd4 117. Nf5 b5
118. Nxd4 Rxd4 119. Rc5 Rb4 120. g5 fxg5 121. Rxg5 Kf6 1/2-1/2


Round 3: Glaurung - Jonny
Glaurung made a dubious trade of rook for knight and a couple of pawns
in the opening.  Even at that early stage of the game, I was pretty
sure what would happen.  In such positions, Glaurung is too happy to
trade down pieces and simplify, eventually reaching an endgame with
K+N+pawns vs K+R+pawns.  My experience is that the rook almost always
wins such endgames, even when the side with the knight has two pawns
more (unless there are dangerous passed pawns, of course).  Glaurung
still doesn't know this, though.

The game followed exactly the pattern I expected, and Jonny went on to
a well-deserved, simple and effortless win.  I am not very
disappointed about losing against Jonny (which is a very strong
program, and had much faster hardware than Glaurung), but I wish I
could have offered more resistance.

[D]brnqnbkr/pppppppp/8/8/8/8/PPPPPPPP/BRNQNBKR w HBhb - 0 1
1. b4 d5 2. e3 e6 3. Bd3 b6 4. O-O Be7 5. c4 h5 6. Ne2 Nf6 7. cxd5
Nxd5 8. b5 h4 9. h3 Qd6 10. Bb2 Bf6 11. Nd4 Nf4 12. Bc2 Ne7 13. Qg4
Be5 14. Nb3 Bxb2 15. Rxb2 Qa3 16. Qxf4 Qxb2 17. Qxc7 Re8 18. Qxa7 Nd5
19. Nd4 Qb4 20. Nef3 Rc8 21. Qa6 Qc5 22. Rc1 g6 23. Nb3 Qf8 24. Be4
Rxc1+ 25. Nxc1 Nc7 26. Qxb6 Bxe4 27. Qxc7 Bxf3 28. gxf3 Rh5 29. Ne2
Rg5+ 30. Kf1 Rxb5 31. Nc3 Rb2 32. Ke2 Qa8 33. Qd7 Qa5 34. Qd3 Kf8
35. a4 Ke7 36. Ne4 Qb6 37. f4 Qb4 38. Kf3 f5 39. Nc3 Qd6 40. Ke2 Qxd3+
41. Kxd3 Kd7 42. Ke2 Kc6 43. Kd3 Kc5 44. Ke2 Rb4 45. Kd3 Kc6 46. f3
Kb6 47. Kc2 Ka5 48. Kd3 Rb6 49. Kc4 Rb7 50. Nb5 Kxa4 51. Nd4 Rb2
52. Kc3 Rb6 53. Kc4 Ka3 54. Kc3 Ka2 55. Kc2 Ra6 56. Kd3 Kb1 57. Ke2
Kc1 58. Nb3+ Kc2 59. Nd4+ Kb2 60. Kd3 Rb6 61. e4 Kc1 62. Kc3 Kd1
63. d3 Ke1 64. Kc4 Kf2 65. Kc5 Ra6 66. Kb5 Ra3 67. Nxe6 Rxd3 68. e5
Rxf3 69. Ng5 Re3 70. Kc5 {-7.12/18 8s} 0-1

Round 4: Patzer - Glaurung
Patzer is one of the truly "classic" chess programs which have been
around since long before I could write a single line of C code.  It
was also one of the first programs to support more than one CPU.
Unlike most other programs from the same era, Patzer is still being
developed and takes part in tournaments.  The author, Roland Pfister,
is a very nice guy and a true gentleman.  I was just beginning to look
forward to a good chat and an interesting game when Patzer suddenly
made an illegal move at move 10:

[D]bbrnkrnq/pppppppp/8/8/8/8/PPPPPPPP/BBRNKRNQ w FCfc - 0 1
1. g3 c5 2. b3 e6 3. f4 f5 4. c4 Nf7 5. Qf3 Nf6 6. Bxf6 gxf6 7. Nc3
Qg7 8. a3 Bd6 9. e4 b6

In this position, white played Ke1xc8.  From Patzer's PV, it was
obvious that it really intended to castle queenside, and that there
was just a bug in the UCI interface code.  Because both players wanted
to play on, we asked arbiter Ingo Althöfer (who did an excellent job
throughout the tournament) for permission to manually set up the
position after 10. O-O-O and continue the game, but unfortunately Ingo
saw no other solution than declaring the game to be a win for black.

There was some consolation in the fact that we managed to sit down and
find and fix the bug shortly after the game, and that Patzer had no
further such problems in the rest of the tournament.

Round 4 is the last round of the first day, and afterwards most of the
participants go out to enjoy a nice dinner at a Greek restaurant.
Thanks to Daniel for suggesting the place, and to Anastasios for
expert translation and good advice in choice of food.  Lots of
interesting technical computer chess discussions are heard around the
table.

Tord



This page took 0.13 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.