Author: Dieter Buerssner
Date: 08:10:06 02/26/01
I hope you won't mind, that I send the same post to CCC and the WB forum.
My impressions from Paderborn. This was the first computer chess tournament I
played ever. It was a very nice and relaxing atmosphere. It was very
interesting, to personally meet all the programmers and some spectators, that I
knew from email or fora. Everybody was really friendly. All operators have shown
their screens to the opponents. Also, there was much chatting during the games.
The programmers didn't hide (some) ideas used in their programs, and even
explained some interesting approaches.
For me as a programmer, it was a very good experience, to manually operate the
engine at a long time control. I detected some misfeatures, that I probably
would not have seen easily without this experience. I.e. Yace does not have an
"easy move" - meaning, that it will think the same amount of time for an obvious
recapture. When you sit there, and see a fail low for such an obvious recapture
just before the "normal" thinking time is over, and then you have to wait for
another 20 minutes, until the program eventually decides to move ...
I think, the tournament was organized extremely well by Ulf Lorenz. A fast
Pentium III 933 Mhz computer with 256 MB of RAM running Windows 2000 was
provided to all the participants by the organizer, SMP capable programs got a
dual PIII 933 with 512 MB RAM.
In the tournament, I played with Yace version 0.99.09. In 7 of the 9 games I
used a book with *many* hand tuned lines by Carlos Pesce. These lines were on
top of a 40 ply book from a game collection of 350000 games contributed by
Mogens Larsen. In two games, I used a very wide book with a small "Crafty-like"
start.pgn (the start.pgn, that is in the publically available distribution of
Yace), also by Mogens Larson. I played with the text interface of Yace.
Some comments about the games of Yace. I won't show the games. They can be found
at http://www.uni-paderborn.de/~IPCCC/
The first game was against Patzer by Roland Pfister. I think, the decisive
position was the following:
[D] R4b2/p4pkp/1p4p1/8/1n1P4/5NBP/r4PP1/6K1 w - -
Here Patzer played Rxf8, after which the score of Yace jumped from 0.1 to 1.38.
The next game was against P.Conners by Ulf Lorenz. Yace played with the white
pieces. For this game, I decided to make Yace very selective (more risky pruning
rules). Allready after 26 moves, a very interesting rook ending was on the
board, where Yace was one pawn down. The rook of Yace was in a very bad position
at a2 with white pawns at a3 and b2. A long walk by the king was needed to free
the rook. Then a position with two white pawns on the queen side and 3 black
pawns on the king side evolved. It was rather exciting to watch. To me, it was
not clear for a long time, who would win. Only later I (and Yace) could see,
that, while white can promote first, that black was winning. Black could sac its
rook for the white b pawn, but the black pawns on 3rd rank would win against the
rook. As often, one tempo decided this. P.Conners could see the win first,
probably because of superior knowledge about the connected passers. One crucial
position of the game, where Yace could still have managed a draw was allready
discussed in CCC.
In the game against SOS by Rudolf Huber, Yace came out of book with -0.4 as
black. Also here there was soon a rook ending on the board, where Yace was one
pawn down, but the score was never less than -0.9. After SOS moved 45. e7 the
following position was on the board:
[D] 8/p2RP3/2r5/5pp1/2kp4/P3PK2/8/8 b - -
Here Yace has shown draw score. I was still nervous, because now SOS showed a
high score for white. But soon it seemed clear, that the position really was a
draw.
Against Matador by Stefan Knappe, Yace played white again. Matador was out of
book very early. When Yace was out of book at it 8th move, it allready showed a
score of 0.57, and was in a very good position. After white's 14th move
[D] rnb2rk1/pp1pp1bp/1q4p1/1N1P2B1/8/8/PP2BPPP/R2Q1RK1 b - -
Matador played 14...Bxb2. I think from this point, the game was won for white.
It was very refreshing to discuss with the strong chess player Stefan Knappe.
Against Zchess by Frank Zibi Yace played black. It was in book for 17 moves.
While Yace came
out of book with a small positive score, the position looked even better to me
than the score.
After white's 30th move
[D] 3r2kb/pb2p2p/1p3ppB/4P3/1R1PqP2/8/P4QPP/5NK1 b - -
Yace played 30...g5. At first, I did not like this move at all. But perhaps it
was not too bad, because the white bishop could not escape its prison on h6 for
the whole game.
After white's 32nd move
[D] 3r2kb/pb2p2p/1p3p1B/4P1p1/1R1P1P2/3q2NP/P4QP1/6K1 b - -
Yace played Qc3. I think, this was a very weak move, that gave black's advantage
away. Later Yace could regain an advantage and win the game.
In the game against Gandalf, that was operated by its book author Dan Wulff,
Yace played black. I think, it came out of book very good after 18 moves in a
French opening. But Gandalf could equalize the game and even get advantage with
a pawn majority on the queen-side. The spectators even commented, that this was
an easy win for Gandalf, but I am not sure. This was the position after white's
33rd move
[D] 8/3nkp1p/4p2P/p2pP3/5PN1/2K5/PP6/8 b - -
Here Yace played f5, which I think was a nice move, that was not expected by
Gandalf. The idea was to force the trade of knights, and to be able to draw the
game. In CCC, there was suggested, that Nh2 would still win the game for
Gandalf, which I think is correct. Perhaps, I was lucky here, that Yace would
search to considerably higher depth in this phase of the game, than Gandalf.
The game against Shredder was very interesting. I think Yace came out of book in
a reasonable position after 12 moves. Shredder was in book for one move longer.
One interesting point in the game was after the 20th move
[D] 2r1r1k1/pb1n1pp1/1p1p3p/2q4n/2P5/P3PP2/1P1NBBQP/2KR2R1 w - -
Here Yace failed high soon with Ne4. Later the score dropped again. At the last
iteration searched the score dropped a little bit more than 0.2, but
(unfortunately?) less than 0.25, so Yace did not decide to allocate more time.
At the next iteration Yace would have switched back to e4. I think, Ne4 was not
a good move, although the very good chess player Ossi Weiner commented during
the game, that it was a good move. Interestingly, after the game Stefan
Meyer-Kahlen analyzed the position with Shredder, and Shredder switched to Ne4
after very long thinking, so my judgement of this move may be wrong.
Another interesting position was after the queens where traded
[D] 3r4/6k1/2P3pp/p4pn1/Pp1P4/1P1R4/K1B4P/8 b - -
At this point, Yace has shown a drawing score, while Shredder allready showed a
score close to 1 in its favor (from memory). I think, at this point the game was
lost for Yace. It is very interesting how well Shredder analyzed the position. I
have no doubt, that it "knew", that the majority on the king side will win
against the two connected passers in the center.
The game against Gromit, operated by its authors Frank Schneider and Skai Skibbe
was the worst game of Yace. When I checked, this game was not available at the
home page of the IPCCC, so I give it here with the scores of Yace.
[Event "IPCCC 2001"]
[Site "Paderborn"]
[Date "2001.02.24"]
[Round "8"]
[White "Gromit"]
[Black "Yace"]
[Result "1-0"]
1. e4 e6 2. d4 d5 3. Nd2 Nc6
4. Ngf3 Nf6 5. e5 Nd7 6. Be2 f6
7. exf6 Qxf6 8. Nf1 Bd6 9. Ne3 O-O
10. O-O Nb6 11. Ng4 Qe7 12. Re1 Bd7
13. c3 h6 {-0.30} 14. Bd3 Be8 {-0.24} 15. Bc2 Bh5 {-0.43}
16. Qd3 Bxg4 {-0.63} 17. Qh7+ Kf7 {-0.63} 18. Bg5 hxg5 {-0.61}
19. Qg6+ Kg8 {-0.61} 20. Nxg5 Qxg5 {-0.56} 21. Qxg5 Bf5 {-0.53}
22. Rac1 Rf6 {-0.42} 23. Bxf5 Rxf5 {-0.45} 24. Qe3 e5 {-0.42}
25. dxe5 Nxe5 {-0.40} 26. f3 c6 {-0.42} 27. b3 Nbd7 {-0.30}
28. Kh1 Nc5 {-0.30} 29. Rcd1 Raf8 {-0.27} 30. b4 Nc4 {0.00}
31. Qe2 Na4 {0.14} 32. Qe6+ R5f7 {0.20} 33. Rd4 Be7 {0.30}
34. b5 Bf6 {0.29} 35. bxc6 Bxd4 {0.29} 36. cxd4 bxc6 {0.23}
37. Qxc6 Nab6 {0.23} 38. Qe6 Rd8 {0.25} 39. h3 Na8 {0.28}
40. Qa6 Nc7 {0.26} 41. Qxa7 Ne6 {0.34} 42. Qa4 Nxd4 {0.36}
43. Kg1 Nf5 {0.35} 44. Re8+ Rf8 {0.40} 45. Rxd8 Rxd8 {0.47}
46. Qa7 Kh7 {0.45} 47. Kf2 Nfe3 {0.43} 48. a4 Rd6 {0.43}
49. Qf7 Rg6 {0.37} 50. g3 Rf6 {0.34} 51. Qa7 Nd1+ {0.48}
52. Kf1 Re6 {0.46} 53. Qd4 Kg8 {1.23} 54. f4 Nce3+ {1.35}
55. Kg1 Rg6 {1.36} 56. g4 Rh6 {0.00} 57. Kh2 Nf2 {1.45}
58. Kg3 Ne4+ {1.29} 59. Kh2 Nd2 {1.14} 60. Kg3 Ne4+ {1.39}
61. Kh2 Nd2 {1.06} 62. Kg3 Ndf1+ {1.21} 63. Kf2 Rxh3 {1.10}
64. g5 Rh2+ {1.02} 65. Kf3 Kh8 {1.02} 66. Qe5 Kh7 {0.00}
67. Qe6 Ra2 {0.00} 68. a5 Rxa5 {0.00} 69. f5 d4 {-0.81}
70. Qg6+ Kh8 {-0.81} 71. f6 gxf6 {-0.84} 72. gxf6 Nf5 {-0.84}
73. Qe8+ Kh7 74. f7 Ng7 {-16.95} 1-0
I think, here Yace came out of book in a very difficult position:
[D] r4rk1/pppbq1pp/1nnbp3/3p4/3P2N1/2P2N2/PP2BPPP/R1BQR1K1 b - -
Just before the end of the search time after exactly 7 minutes, Yace switched to
the horrible move 13...h6. But Gromit gave Yace another chance and traded into a
position with very unbalanced material. During most of the game Gromit had a
queen and a pawn vs. 3 minor pieces for Yace. I actually believed, that Yace
would be able to win the game.
The game allready went through repetitions, but Yace still tried to win. At the
end it lost ...
During this game, I also detected an evaluation bug of Yace for pawnless
endgames, which were possible. I have the rule built into Yace, that without
pawns, a material advantage of a minor piece is not enough to win, and give it a
drawish score. As shown here, this is obviously wrong.
At the end of the game, where Yace showed the score of about -0.8, it has seen a
main line with 2 queens vs. 1 Q and 2 Ns.
The last game of Yace was against Holmes by Andreas Herrmann. Again Yace was
longer in book than Holmes. Few moves after Yace left the book, Yace had
allready a very strong position. In this game, Yace searched considerably deeper
than Holmes, and had not much difficulties to win. But I have no doubt, that
Andreas will be able to improve his program soon.
[Event "IPCCC Paderborn 2001"]
[Site "Paderborn"]
[Date "2001.02.25"]
[Round "9"]
[White "Yace"]
[Black "Holmes"]
[Result "1-0"]
1. Nf3 c5 2. c4 Nf6 3. Nc3 d5
4. cxd5 Nxd5 5. d4 Nxc3 6. bxc3 g6
7. e3 Bg7 8. Bd3 O-O 9. O-O Qc7
10. Ba3 Nd7 11. e4 e6 12. Rb1 {0.03} Rd8
13. Bc1 {0.20} Rb8 14. Bg5 {0.39} Re8 15. Qd2 {0.44} c4
16. Be2 {0.51} Ra8 17. Bh6 {0.72} Bxh6 18. Qxh6 {1.05} a5
19. Nd2 {1.23} Nb6 20. e5 {1.82} f5 21. exf6 {1.79} Nd5
22. Bxc4 {1.95} Nxf6 23. Rfe1 {1.94} Ng4 24. Qh4 {2.46} h5
25. Bd3 {2.86} Kg7 26. Nf3 {3.10} Qe7 27. Rb5 {3.41} e5
28. h3 {3.87} Qxh4 29. Nxh4 {3.86} Nxf2 30. Kxf2 {5.66} g5
31. Nf3 {5.99} g4 32. hxg4 {6.20} hxg4 33. Nxe5 {6.35} Rh8
34. Ng6 {9.48} Rf8+ 35. Nxf8 {11.39} Kxf8 36. Bc4 {11.66} g3+
37. Kxg3 {17.30} Ra6 38. Rg5 {18.92} Be6 39. Bxe6 {Mat08} Rxe6
40. Rxe6 {Mat06} Kf7 41. Rge5 {Mat05} b6 42. Kg4 {Mat04} Kg7
43. Re8 {Mat03} a4 44. Kg5 {Mat02} a3 45. R5e7# {Mat01} 1-0
I was very happy with the result of Yace in my first tournament. I like to thank
the organizers, the fellow programmers and the few spectators - some of which I
have known from email - for the very nice atmosphere. My thanks also goes to
Mogens Larsen, for creating the game collection on which the Yace book is based.
He also created the binary book for me, a task which I could not do personally,
because of hardware limits. I want to emphasize my thanks to Carlos Pesce. He
has invested a lot of time for creating a tournament book for Yace, which he
continiously improved and tested. He also has sent my very much feedback about
the playing style of Yace, showed me many weak points and helped me extremely to
improve Yace. Also many other people sent feedback, that helped.
Thanks, to all of them.
Regards,
Dieter
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.