Computer Chess Club Archives




Subject: Glaurung in Mainz: Part 2 (even longer)

Author: Tord Romstad

Date: 04:03:55 08/14/05

Day 2 (2005-08-12):

We start playing earlier this day, because there is no player's
meeting.  Dramatic events occur not only on the boards, but also
outside -- there is a terrible rainfall over Mainz, and suddenly the
roof starts leaking and we have a big waterfall inside the playing
hall, just 10-15 metres away from our computers.

Glaurung faces a really tough opponent in the first round of the day:
Sjeng running on a fast dual CPU computer.

Round 5: Glaurung - Sjeng

Gian-Carlo Pascutto is one of those I have known since a long time
from the CCC and the Winboard Forum, but I was very happy to finally
meet him in person.  I was considerably less happy about what was
going on on the board.

[D]rkqbbnrn/pppppppp/8/8/8/8/PPPPPPPP/RKQBBNRN w GAga - 0 1

1. d4 d6 2. Nhg3 Nhg6 3. Bc3 e5 4. dxe5 dxe5 5. e3 Be7 6. e4

My English vocabulary does not suffice to express my utter disgust and
disbelief when watching Glaurung's opening play.  White's moves 3
through 6 are among the silliest I have ever seen on a chess board.

6... Ne6 7. Ne3 Bd7 8. Nd5 Bg5 9. Bd2 Bh4 10. Bh5 c6 11. Ne3 Nef4
12. Qd1 Qc7 13. Bb4 O-O-O

[D]2kr2r1/ppqb1ppp/2p3n1/4p2B/1B2Pn1b/4N1N1/PPP2PPP/RK1Q2R1 w GA - 0 14
Only 13 moves are played, but the position already looks close to lost
for white.  Glaurung has somehow managed to reach an almost
symmetrical position where it is several tempi behind, despite playing
with the white pieces.  But the comedy is not over yet.  Glaurung
finds a way to make the position even worse:

14. Bd6?

Brilliant.  White forces the black queen to the excellent square a5,
where it makes it even more difficult for white to complete his
development by castling.  Simultaneously, Glaurung creates a nice
self-pin along the d file.

14... Qa5 15. Rf1 Be6 16. Ngf5 Be7 7. Bxg6 Nxg6 18. Nxe7+ Nxe7 19. Qd3
1c5 20. Rd1 Rd7 21. Qa3 Qxa3 22. bxa3 Rgd8 23. Bxe7 Rxe7 24. Rxd8+
Kxd8 25. O-O-O+ Rd7 26. Nd5 f5 27. Nc3 Rxd1+ 28. Kxd1 f4

[D]3k4/pp4pp/4b3/2p1p3/4Pp2/P1N5/P1P2PPP/3K4 w - - 0 29
Glaurung has somehow managed to simplify to an endgame which looks
almost playable.  I suspect it was helped by some inaccurate black
moves on the way.  It cannot really have been necessary for black to
allow the simplifications.

I was still convinced that the position was won for black, though.
White's queenside pawn structure is ugly, and the bishop should be
stronger than a knight in this endgame with pawns on both wings.
Considering the fact that Glaurung is also known as the leading
endgame clown among the currently active chess engines, I was sure the
game was as good as over.  To my surprise, Glaurung found a clever way
to block the position and restrict the mobility of black's bishop:

29. Na4! b6 30. Nb2!

Exploiting the weakness of the e5 pawn.  White threatens Nd3 followed
by Nxe5.  Black has the choice between capturing on a2 and giving
white a passed pawn on e4, or defending the e5 pawn with the king and
allowing white to block the pawn structure by playing c2-c4.  Sjeng
chose the latter option:

30... Ke7 31. c4 g5 32. Kd2 a6 33. Kc3 Kd6 34. Na4 b5 35. Nb6 Kc6
36. Nd5 h5 37. Ne7+ Kd6 38. Nd5 h4 39. a4 b4+ 40. Kd3 a5 41. Nf6 Ke7
42. Nd5+ Kf7 43. Nc7 Bc8 44. Kc2 Ke7 45. Nd5+ Kd6 46. Nf6 Bb7 47. f3
Ke6 48. Ng4

[D]8/1b6/4k3/p1p1p1p1/PpP1PpNp/5P2/P1K3PP/8 b - - 0 48
Sjeng has been rather cooperative in blocking the pawn structure.  By
now the position looks very drawish.

48... Bc6 49. Kb3 Bd7 50. Nh6 Ke7 51. h3 Be6 52. Ng4 Bxg4
53. hxg4

[D]8/4k3/8/p1p1p1p1/PpP1PpPp/1K3P2/P5P1/8 b - - 0 53
Now the draw is completely obvious, but Ingo asks us to play on for a
while to make sure both programs understand the position well enough
not to make any blunders.  Nothing interesting happens, of course.
This was the last game of the round to finish, and the most amusing
thing about the game was the big crowd of spectators around our table
towards the end of the game, watching this very exciting pawn endgame.

53... Kd6 54. Kb2 Kc7 55. a3 Kb8 56. Kc1 Kb7 57. Kc2 Kc6 58. Kb2
Kc7 59. Ka2 Kb8 60. Kb2 Kc8 61. Kb3 Kd7 62. Kb2 Kc6 63. Ka2 Kd6
64. Kb3 Ke7 65. Ka2 Ke6 66. Kb2 Kd7 67. Ka1 Kc8 68. Kb1 Kd8 69. Kb2
Ke8 70. Kc2 Kf8 71. Kc1 Ke7 72. Kc2 Ke8 73. Kb3 Kf7 74. Kc2 Ke7
75. Kb2 1/2-1/2

Round 6: Pharaon - Glaurung

When looking at the table of results form a tournament like the one in
Mainz, it is too easy to just see the winner and overlook that there
are also some impressive performances a few places further down on the
list.  In Mainz, Pharaon deserves to be mentioned.  After the first
three rounds, it shared the first place with Spike, with a perfect
score.  It lost against Spike and Shredder (which are formidable
opponents, of course) in the next two rounds, but ended on a strong
7th place with 4/7 despite facing very tough opposition throughout the
tournament.  Unlike most of the strong programs ahead of it, Pharaon
played on the rather slow Pentium IV computers provided by the
hardware sponsor.

The game between Pharaon and Glaurung was rather strange.  Pharaon
played very passively in the centre.  An interesting curiosity is that
all white's four central pawns (c2, d2, e2 and f2) remained on their
original squares until move 29!  Glaurung's opening play wasn't really
much better.  It developed its pieces very clumsily, and the weak
light squares around the black king made me rather worried for a
while.  As it turned out, Pharaon wasn't able to exploit the
weaknesses, and Glaurung's pawn centre decided the game.

[D]brnknbqr/pppppppp/8/8/8/8/PPPPPPPP/BRNKNBQR w HBhb - 0 1

1. g3 e6 2. b3 Ne7  3. Bg2 Nc6 4. Nf3  b6 5. Nd3 O-O-O 6. b4 f6 7. b5
Na5 8. Bc3 Nb7 9. Nfe1 Ned6 10. Nf4 Kb8 11. Ned3 e5 12. Bd5 Nf7
13. Nh5 Nc5  14. Bxa8 Kxa8 15. Qg2+  d5 16. Nb4 Nh6 17. Nc6 Re8 18.
h3 Qe6 19. Bb4 Rg8 20. g4 g6 21. Nxf6 Qxf6  22. g5 Qd6 23. gxh6 a5 24.
Ba3 Qe6 25. Rf1 Ne4 26. Bxf8 Rgxf8 27. Rb3 Rf4 28. Re3 Ref8

[D]k4r2/2p4p/1pN1q1pP/pP1pp3/4nr2/4R2P/P1PPPPQ1/3K1R2 w - - 0 29
Now white finally pushes the f pawn:

29. f3 d4 30. Rb3 Nc5 31. Rb2 Qf6 32. Rb1 Qd6 33. Qg1 Re8  34. Rb2 Qf8
35. Qg2 Qf6 36. Rb1 e4 37. Qg3 Na4 38. Re1 e3

[D]k3r3/2p4p/1pN2qpP/pP6/n2p1r2/4pPQP/P1PPP3/1R1KR3 w - - 0 39
Black must be winning.

39. h4 exd2 40. Kxd2 Nc3 41. a3 Rxh4  42. Qxc7 Nxb1+ 43. Kc1  Qf4+ 44.
Qxf4 Rxf4 45. Kxb1 Rxf3 46. Rd1 Rf5  47. a4 Re4 48. e3 Rxe3 49. Rxd4
Rf7 50. Kb2  Re2 51. Rd6 Re8 52. Nd8 Rfe7 53. Nc6 Rb7  54. c4 Rf7 55.
c5 bxc5 56. Nxa5 Rf2+ 57. Kc3 Re3+ 58. Kc4 Rc2+  59. Kd5 Rd3+ 60. Kc6
Rxd6+ 61. Kxd6 c4 62. Kc5 g5 63. Nc6 g4 64. Ne5 g3 65. Kb4 Rf2 66. Ka5
c3  67. Nd3 c2 68. Kb4 g2 69. b6 g1=Q 70. b7+  Kxb7 71. Kc3 c1=Q+ 72.
Nxc1 Qxc1+ 73. Kd3 Qd2+ 74. Ke4 Rf4+ 75. Ke5 Kc6  76. a5 Qe3# 0-1

Round 7: Glaurung - Spike

Before the last round, the winner of the tournament already seemed to
be decided.  Spike was leading by a full point ahead of Jonny, and
even if Spike lot and Jonny won in the last round, a small miracle
would be needed for Jonny to overtake Spike in Bucholz points (the sum
of the scores of all opponents).  Nevertheless, the modest and
sympathetic Spike authors refused to accept any congratulations before
the last game was played.  The task of helping Jonny achieve the
impossible fell on Glaurung.

[D]bbqrknnr/pppppppp/8/8/8/8/PPPPPPPP/BBQRKNNR w HDhd - 0 1
1. c4 b6 2. Nf3 Nf6 3. b4 Ne6 4. d4 d5 5. c5 c6 6. e3

[D]bbqrk2r/p3pppp/1pp1nn2/2Pp4/1P1P4/4PN2/P4PPP/BBQRKN1R b HDhd - 0 6
As usual, Glaurung is happy in the opening, and gives white a +0.2
score.  The reasons are probably white's space advantage and the black
knight on e6 blocking the e7 pawn.  Personally I think black is a bit
better.  White's a1 bishop will be difficult to activate, while black
has the simple Ba8-b7-a6 maneuver to activate his bad bishop.  Spike
proceeds to execute exactly this maneuvre.

6... Bb7 7. Bd3 bxc5 8. bxc5 Ba6 9. Qc2 O-O 10. N1d2 Rfe8 11. Bf5

An unexpected and somewhat strange move.  Shouldn't white castle

11... h6?

Playing a move like this against Glaurung is the computer chess
equivalent of waving a red rag to a bull.  This is not necessarily a
bad thing; Glaurung often self-destructs and loses without a fight
when trying to attack.  In the current game, however, it looks like
white's attack is really dangerous.

12. Rb1 Bb5 13. a4 Ba6

[D]1bqrr1k1/p3ppp1/b1p1nn1p/2Pp1B2/P2P4/4PN2/2QN1PPP/BR2K2R w H - 0 14

14. g4!

During the game, I thought this attack was too optimistic.  There is
no easy way to bring the bishop on a1 and the rook on b1 into the
attack, and if the attack fails white's king will probably be in great
danger.  In retrospect, I think white's attack really works.  Serious
threats against the black king appear with frightening speed.  Perhaps
black's 11th move was even the decisive mistake.

14... Nd7 15. Bxe6 fxe6 16. g5 hxg5 17. Rg1

The score is already +1.65, and if I recall correctly Spike displayed
something similar.  White is probably winning.  Glaurung continues to
conduct the attack in an efficient and precise way, but the attack is
too brutal to be described as beautiful.  I suppose it must still be
characterised as spectator friendly - The audience always likes to see

17... Nf8 18. Nxg5 Bc7 19. h4  e5 20. dxe5 e6 21. h5 Rd7 22. f4 Bd8
23. h6 Kh8 24. Ndf3 gxh6 25. Qh2 Nh7 26. Qxh6 Rg7 27. Kf2 Qd7

[D]3br2k/p2q2rn/b1p1p2Q/2PpP1N1/P4P2/4PN2/5K2/BR4R1 w - - 0 28

28. f5

At last, even the bishop and rook on the queenside are ready to
participate in the attack.

28... d4 29. f6 dxe3+ 30. Kg3 Reg8 31. Rbd1 Bd3 32. Kf4

[D]3b2rk/p2q2rn/2p1pP1Q/2P1P1N1/P4K2/3bpN2/8/B2R2R1 b - - 0 32
Possibly the most picturesque position from Glaurung's games in

32... e2  33. fxg7+ Qxg7 34. Qxg7+  Rxg7 35. Rxd3 Bxg5+ 36. Nxg5
Nxg5 37. Rdg3  e1=R 38. Rxe1 Nh7  39. Rxg7 Kxg7 40. a5 Nf8 41. Rb1
Kh6 42. a6 Ng6+ 43. Ke4 1-0

The end result for Glaurung was 5/7 (+4,=2,-1), and somehow it even
managed to sneak ahead of Shredder by 0.5 Buchholz points to achieve
third place.  This is of course a tremendous success, especially
considering my inferior hardware.  But if we look behind the numbers,
it is clear that luck was a very big factor.  One of the wins (against
Patzer) was because of an illegal move, so we could argue that
Glaurung's true score was only 4/7.  I was also very lucky with the
colors:  White against Spike, Jonny, List and Sjeng, and black against
Pharaon, Patzer and Ant.  Four white games out of seven (or six, if we
omit the game against Patzer), and there is no doubt that the average
strength of the opponents was higher in the white games.  It should be
added that the wins against List, Pharaon and Spike were not entirely
convincing, and that the draws against Sjeng and Ant came from
inferior positions.  Glaurung really didn't deserve the third place,
but in such a short tournament it is always possible to be lucky.

After the prize ceremony, the champions unfortunately have to leave
us, but most of the other programmers and groupies go out for dinner
again, this time at a Mexican restaurant.  The food is not bad, but
the service is awful and the music too loud.  Of course, this does not
prevent us from having a really good time.  The technical discussions
do not quite reach the same level of quality as the day before;
without a doubt the reason is a dramatically higher consumption of
alcohol.  Nevertheless, Tony's idea of lossy compression of endgame
bitbases catches my attention.

After returning to the hotel, some of us decide to stay in the hotel
bar for a few drinks.  Around 1:30, I finally say good night to the
others and return to my room, tired but happy.

This was my first live tournament, and I enjoyed it even more than I
expected.  Meeting the other programmers in person was great.
Suddenly I have many new friends, and invitations to several European

Congratulations to Volker and Ralf for a very impressive performance
and a well deserved victory, and also to Johannes for finishing with
the same score.  Spike and Jonny dominated the tournament, and clearly
deserved the first place.

Mark Vogelgesang and Ingo Althöfer did a fantastic job in organising
the tournament.  Despite Mark's lack of experience with this type of
events, almost everything went flawlessly (the only exception was the
lack of food and water in the playing hall the first day).  I hope
they will be able to repeat the tournament next year, preferably with
more rounds, better Internet coverage, more participants and faster
sponsored hardware.

Finally, I would like to thank Fabien Letouzey for his invaluable help
in my preparations for Mainz.  Without his help, I would not have been
able to test my own engine.  In a very short time, he added Chess960
support to PolyGlot, and helped me in creating a Chess960 enabled
xboard.  He also kindly sent me a private version of Fruit with
Chess960 support, giving me an excellent sparring partner in the last
few weeks before Mainz.  I am not sure whether Fabien reads CCC while
in Reykjavík.  In case he doesn't, I hope some of the other people
there could send him my greetings, and say that the third place in
Mainz belongs to him as much as me.


This page took 0.02 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.