

Thanks to the organizers for another successful event and to the guys on our team for their hard work. Our team was:
Operator: Hans van der Zijden
Hardware: Lukas Cimiotti
Opening book: Jiří Dufek
Engine: Vasik Rajlich
Lukas has continued working on his cluster, which now contains 9 Nehalem 8-core machines running at speeds ranging from 2.93 to 3.33 GHz. For the first time, we were fully ready several days before the tournament and had no major problems during the event. This project has been a lot of work but also a great succes and quite some fun. The next steps are to clean up a few things and to configure and optimize some of the algorithm's parameters. We also hope that there will be some freestyle events in the next couple of months so that we can showcase what we have done.
Jiří again handled our opening book. He's adapted himself very quickly to the unique requirements of the situation: outbook the Deep Sjeng team, the main tournament threat for already some time and especially now, and avoid drawish lines with both colors in the remaining games. Jiří has tackled the drawing issue in a moderate way, manually eliminating long well-treaded variations but still playing mostly mainstream openings. He also changed his testing methods to include a bigger variety of opening books and engines, to better-simulate the tournament conditions. Our opening play was again a big success - the two primary goals were clearly reached and as a nice bonus two games were practically won by the book on the last day.
Hans was again our operator and was joined on the last day by Jeroen Noomen, who made a visit to the tournament as his computer chess sabbatical winds down.
Despite Rybka playing reasonably well this time, the tournament was again exciting as Deep Sjeng ripped through the rest of the field and finished half a point behind us and two points clear of third place. Tournament play comes down to hardware & parallelization, book preparation, and single-CPU engine strength, and the Deep Sjeng team seems to have checked all three boxes and will likely remain a force for the near future.
The tournament web site is here: http://www.computerschaak.nl/
The tournament forum thread is here: http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=12928
Jiri's opening report can be found here.
All games can be found in our updated Rybka Cluster Base
Gian-Carlo's thread about Deep Sjeng can be found here: http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=12969
Vas
I'll post a perspective from our POV in another thread.
Amazing tournament and fun games to watch.
Congrats on your team's result.
Do you think there's a chance of a 12 game long time control match between Rybka Cluster and Sjeng Cluster? Maybe with the right sponsor?
>Do you think there's a chance of a 12 game long time control match between Rybka Cluster and Sjeng Cluster? Maybe with the right >sponsor?
Nice! Now that i'll be missing for some time can you repeat it every single day until it's done? :-P
> Do you think there's a chance of a 12 game long time control match between Rybka Cluster and Sjeng Cluster? Maybe with the right sponsor?
It could be. We consider all offers :)
Vas
When is Rybka 4 due? :-P :-D :-)
Vas
Vas
Vas
go go Vas!
I have still Rybka 2.3.2a (don't buy R3, because of bug's (and hoped of bugfixes, but this is not happen)) and therefore i will buy R4 of course because of the expected strength.
sorry for my english
> don't buy R3, because of bug's
Rybka 3 could have had twice the bugs that it had, and she would still have been better than Rybka 2.3.2a, by far.
Vas
Vas
PS: Where is the great BB?
> We'll have an announcement pretty shortly. However - sorry for the tease - it won't be quite what you expect.
Well lets see, Hiarcs is due to launch fairly soon I believe. And what I won't expect is to see a RYBARCS or a HIABKA :-)
>We'll have an announcement pretty shortly.
I need help, my English is so bad! What´s more immediately: pretty shortly or very soooooooon? :-)
Vas
> The number of 'o's is obviously crucial.
Indeed, I used "sooon" and had to do it under 24 hours!
This has been patiently waiting since the days of Rybka Beta.
I changed too testing phase – this time I tested only little on Playchess, but much more I tested against different engines and different openings books (which are freely available on the internet) which offers wide spectrum of openings for testing. It was tested in over 700 games (in 10 periods, every period took 2-3 days) when Rybka 3 scored approx. 76%. After finishing every period I checked all games and prepared innovation in lines, where I wasn’t happy.
Round 1: Hermann–Rybka
I didn’t predict that my “secret” opening will be unhide in the first round. Tango (3..Nc6!? – and Palliser book about it) worked very well. Last book move was 10..Nd7
1. Nf3 Nf6 2. c4 e6 3. d4 Nc6 4. a3 d6 5. Nc3 g6 6. e4 Bg7 7. h3 O-O 8. Bg5 h6 9. Be3 Re8 10. Bd3 Nd7
[D]
With score about +0.36 for white after 11.0-0. What is more important – on the board will be kind of King’s Indian defence, which engines plays pretty bad. Only 5 moves later score was lightly positive for black and after 20....bxa5 score jumped was very close to +1 for Rybka. Games was finished with very nice attack by Rybka.
Round 3: Sjeng-Rybka
I was very surprised with this pairings, because this game was decisive for all tournament.
1.e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 a6 5. Nc3 Qc7 6. Be2 b5 7. O-O Bb7 8. Re1 Nc6 9. Nxc6 dxc6 10. e5 Rd8 11. Bd3 c5 12. Qe2 c4 13. Be4 Bxe4 14. Qxe4 Ne7 15. Be3 b4 (last book move, somewhere about 0.00)
[D]
Its interesting that GCP told me, that Sjeng was out of book after 10..Rd8, because I think that Erdo knows this line up to 14.Qxe4
In this game there were three openings milestones: first is 12.Qe2 – there are others possibilities as 12.Qh5 , 12.Qg4 or 12. Bg5. I think that after all black hold equality but he need some precise moves. Second one will be hidden for this moment and will be uncovered lately :-). Third one is 16.Nd1 (16.Ne2 was played before).
Very interesting about this line is how different engines hadle this variation – and answer is: simply bad. Engines very often overestimated white position which leads lately to catastrophe. Sjeng played not so his best and after piece sacrifice for only 2 pawns and few checks games was over.
Round 4: Rybka-Hansdampf
In this game Hansdampg played Nimzo-indian and played very well. Last book move was 10.g3
1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. f3 d5 5. a3 Bxc3+ 6. bxc3 c5 7. cxd5 Nxd5 8.dxc5 f5 9. Qc2 f4 10. g3
[D]
I must say that I was surprised, that Rybka didn’t choose after 10..Nc6 nor 11.Qd3 neither 11.e4 but 11.Nh3!?. After this position became quite complicated but after 16.Nd6 (+0.34) Rybka started like it:
[D]
Black greedy took rook (16..Bxb1) and immediately score jumped to +1.16 after 17.Qxd5. After this black position was worse and worse and black resign at move 42
Round 5: Rybka-Pandix
There is not much what say – after opening Rybka took exchange and won game in endgame.
1.d4 Nf6 2. c4 e6 3. Nf3 d5 4. g3 dxc4 5. Bg2 Bb4+ 6. Bd2 a5 7. O-O O-O 8. Bg5 (last book move)
[D]
Pandix played 8..Nbd7 which looks as inaccuracy (8..Nc6 is more common) and after 9. Qc1 h6 10. Bxf6 Nxf6 11. Qxc4 b6 12. a3 Be7 13. Ne5 Ba6 14. Qc2 Qxd4 15. Nc6 Qc5 16. Nxe7+ Qxe7 17. Bxa8 Rxa8
[D]
Rybka won exchange and safely converted advantage to win.
Round 6: Ktulu-Rybka
Ktulu is known me as very good engame player and dangerous engine (probably Sjeng team can talk about it for Sundays game).
1.e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 Nc6 5. Nc3 Qc7 6. Be3 a6 7. Qd2 Nf6 8.O-O-O Be7 9. f3 b5 10. Kb1 Ne5 11. g4 O-O 12. g5 Nh5 13. f4 Ng4 14. e5 b4 15.Ne4 Bb7 16. Rg1 (16.Bg2!?) 16..Bxe4 (last book move, score about +0.00 after 17.Rxg4 d6)
[D]
This line is very complicated and maybe 8..Be7 will in the future replace now more popular (and long time mentioned as only move) 8..Bb4. After 17.Rxg4 we have very lively and interesting position on the board. However Ktulu didn’t understand his position and lost game pretty quickly and resigned with piece up but under matting attack.
[D]
0-1
Round 7: Rybka-Baron
Baron played very well in tournament and very well against us.
1. d4 d5 2. c4 dxc4 3. e3 e5 4. Bxc4 exd4 5. exd4 Bd6 6. Nf3 Nf6 7. O-O O-O 8.
h3 Nc6 9. Nc3 h6 10. Qc2 (last book move, 10..Na5 11.Bd3 around +0.3)
[D]
Baron had prepared 3 moves more in the book. Arising position resembles Russian defence and give small advantage for white. During all game Baron defended very well (with help of moves which I dislike – 17..f5 for example) and in spite of score climbed up to +0.65 for Rybka during game, on the board remain few pieces and game ended in a draw. My congratulions to Baron team (and Arturo Ochoa for his book)
Round 8: Joker-Rybka
Joker played all tournament without book and this time he hadn’t any chance against Rybka.
1.e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 Nc6 5. Nb5 d6 6. Bf4 e5 7. Bg5? Qxg5 8.Nc7+ Kd8 9. Nxa8 Nf6 10. Nc3 Nd4 -/+
[D]
Rybka very quickly over +1 and white could resign every time, but he played until move 56.
Round 9: Rybka-The King
1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. g3 dxc4 5. Bg2 c5 6. O-O Nc6 7. Qa4 cxd4 8.
Nxd4 Qxd4 9. Bxc6+ Bd7 10. Rd1 Qxd1+ 11. Qxd1 Bxc6 12. Nd2 b5 13. a4 Be7 14.
axb5 Bxb5 15. Nxc4 (about +1)
[D]
During this game I was pretty nervous. The King chose line, which leads to small material advantage with drawing chance for black. I think,thatthis position is quit difficult for engine-only games and The King played weakly endgame than middlegame in this tournament.
The King missed after 15… O-O 16. b3 Rfd8 17. Qc2 Rdc8 18. Ba3 Bxa3 19. Rxa3 h6 20.
Qb2
[D]
possibility exchange remaining queens pawns with 20..Bxc4 21.bxc4 Rxc4 22.Rxa7 or 22.Qb7 and 23.Rxa7 – when its not clear how to simple enlarge white advantage (of course – kingside pawn storm is only possibility). After that black position quickly collapsed mainly with help of 34..a4? – after this move passed b-pawn was unstoppable.
Summary – I think that opening was quite succesfull for Rybka. We didn’t get worse position after opening and Rybka had every chance show her playing strength. New openings works very well and suited Rybka’s playing style. Of course, there are moments, which are important for future development of book – games with Hansdampf and Baron come on mind and question if is Catalan playable for win after game with The King :-). Last but not least i thanks and congratulate to Lukas and Vasik for their beast and fourth Leiden-win in a row :-). Congratulions too second Sjeng team and third Baron team too for excellent tournament.
I have a question that interests me greatly, as I spent by chance a couple of months ago a long time analysing the opening that occurred in the Sjeng game.
My conclusion was that 15. a3 (instead of 15. Be3) should lead to a draw (stops that b4 plan for black). So I discarded this line from my book for white or black.
Had a draw been reached in this game, it would have meant the tournament would be decided on tie-breaks, possibly.
Can you please share with us if you analysed 15. a3 and wasn't worried about it, assuming you wanted a win against Sjeng?
Should be very interesting to see how you planned Black's plan for that scenario.
Thanks again - this was a very interesting post to read by you.
>Can you please share with us if you analysed 15. a3 and wasn't worried about it, assuming you wanted a win against Sjeng?
>Should be very interesting to see how you planned Black's plan for that scenario.
From his comments I presume he understood that the engines overestimate the white postion and that hence a draw is unlikely as white will be playing for a win (and overstretching as a result).
Which is pretty much what happened.
When I analysed that opening it was for a potential 'advanced chess' game, and I guess there that holds differently.
Thanks for taking the time to answer this.
Thx GCP
Thank you too.
I have a question that interests me greatly, as I spent by chance a couple of months ago a long time analysing the opening that occurred in the Sjeng game.
My conclusion was that 15. a3 (instead of 15. Be3) should lead to a draw (stops that b4 plan for black). So I discarded this line from my book for white or black.
Had a draw been reached in this game, it would have meant the tournament would be decided on tie-breaks, possibly.
I don't remember exactly, but i think that i didnt analyzed 15.a3 at all for following reasons - 12.Qg4/Qh5 are more dangerous (and analysis were very time consuming), after 14..Ne7 is black simply OK (equal).
I remember old story, in which Petrosian told, that is better play for win from equal position than from bad. After first 2 games of Sjeng i thought that it was good idea.
If we drew against Sjeng, then Sjeng would be clearly winner - but this would be another tournament.
We can continue over messages our theoretical discussion :-)
>My congratulions to Baron team (and Harvey for his book)
It was alleged the book came from Arturo Ochoa. Is this not correct?
on Leiden web is writenn : Erdogan Gunes (book), but if there is mistake i will change my report. Maybe there is our misunderstanding.
http://www.csvn.nl/index.php?option=com_content&view=article&id=387%3Aonk2009-leiden-list-of-participants&catid=51%3Atoernooien&Itemid=28&lang=en
Btw: on the same page there is info, that i am from Poland. :-) I am from Czech Republis.
Jiri
>Ktulu is known me as very good engame player
Please, have a look to the game Joker vs. Ktulu! :-)
We speak about a serious computer chess tournament and we see this in 2009 (I´m really pissed off):
[Event "29th Dutch Open CCC"]
[Site "Leiden"]
[Date "2009.10.17"]
[Round "4"]
[White "Joker"]
[Black "Ktulu"]
[Result "1/2-1/2"]
[Annotator "1. +0.12"]
[PlyCount "240"]
[EventDate "2009.??.??"]
[TimeControl "4800"]
1. e4 {0.12/14 1} c6 2. d4 {0.33/14 50} d5 3. e5 {0.29/14 41} Bf5 4. Be2 {0.45/14 40} e6 5. h4 {0.22/14 2} Be7 6. Nf3 {0.17/14 1} Nh6 7. Bxh6 {0.45/13 37} gxh6 8. Nbd2 {0.32/14 1} Nd7 9. h5 {0.31/15 3} Rg8 10. O-O 0.19/15 1} Qb6 11. Nb3 {0.36/15 1} O-O-O 12. a4 {0.25/14 1} a5 13. Qd2 {0.23/15 1} Qb4 14. Qxb4 {0.44/15 1} axb4 15. Ne1 {0.36/16 1} c5 16. dxc5 {0.67/15 1} Nxe5 17. a5 {0.47/15 1} Kc7 18. Nd3 {0.32/13 5} Nc6 19. Ra4 {0.34/14 3} Bf6 20. Rb1 {0.23/14 1} Be4 21. Bf1 {0.27/15 2} Rg5 22. a6 {0.07/14 30} bxa6 23. Rxa6 {0.41/14 1} Rb8 24. Ne1 {0.38/14 1} Rxh5 25. Be2 {0.27/14 1} Rg5 26. Bd3 {0.19/14 1} h5 27. Na5 {-0.20/14 1} Nxa5 28. Rxa5 {-0.73/15 1} Bd4 29. Kf1 {-0.43/15 1} h4 30. Ra7+ {-1.27/15 1} Rb7 31. Rxb7+ {-1.58/17 1} Kxb7 32. Bxe4 {-1.87/18 1} dxe4 33. c3 {-2.04/18 43} Bxc5 34. cxb4 {-1.98/19 55} Bxb4 35. Nc2 {-1.74/17 54} Bd2 36. Rd1 {-1.75/17 48} Rd5 37. Ke2 {-1.80/17 28} Bf4 38. b3 {-1.90/18 1} Rxd1 39. Kxd1 {-2.19/19 38} Kc6 40. Ke2 {-2.77/19 37} f5 41. Kf1 {-2.91/18 30} Kc5 42. Ke2 {-3.15/18 31} h5 43. Ke1 {-3.13/18 30} Bd6 44. Ke2 {-3.29/20 55} f4 45. Kf1 {-3.36/20 24} Kd5 46. Kg1 {-4.16/21 1} Bc5 47. Kf1 {-4.04/21 27} e5 48. Ke2 {-4.23/22 33} f3+ 49. gxf3 {-3.15/22 22} h3 50. fxe4+ {-5.37/23 41} Kxe4 51. Kf1 {-7.41/21 5} h2 52. Kg2 {-9.16/24 28} Kd3 53. Ne1+ {-10.89/24 57} Ke2 54. Nc2 {-10.94/25 46} e4 55. Kxh2 {-10.93/1 1} Kd2 56. b4 {-13.42/20 1} Bd6+ 57. Kg1 {-14.19/1 1} Kxc2 58. b5 {-14.19/18 1} Bc5 59. Kf1 {-14.65/18 52} h4 60. Ke2 {-14.37/17 20} Kc3 61. Kf1 {-14.37/16 8} Kd2 62. Kg2 {-23.05/17 50} h3+ 63. Kxh3 {-14.46/17 28} Bxf2 64. Kg2 {-100.11/17 17} Bb6 65. Kh3 {-100.09/17 38} e3 66. Kh4 {-100.08/16 46} e2 67. Kg5 {-100.07/13 7} e1=Q 68. Kf6 {-100.06/12 1.3} Bd8+ 69. Kf5 {-100.06/11 10} Qe8 70. Kf4 {-100.05/10 1.5} Qxb5 71. Kf3 {-100.04/8 0.2} Qb4 72. Kg3 {-100.04/8 0.2} Qe4 73. Kh3 {-100.03/6 0.1} Qf3+ 74. Kh2 {-100.03/6 0.1} Kc3 75. Kg1 {-100.03/6 0.1} Kb2 76. Kh2 {-100.03/6 0.1} Kc2 77. Kg1 {-100.03/6 0.1} Kb2 78. Kh2 {-100.03/6 0.1} Ba5 79. Kg1 {-100.03/6 0.1} Bb6+ 80. Kh2 {-100.02/4 0.1} Qf1 81. Kg3 {-100.04/8 0.6} Qf5 82. Kg2 {-100.03/6 0.1} Qg4+ 83. Kh2 {-100.02/4 0.1} Kc2 84. Kh1 {-100.01/2 0.1} Kb2 85. Kh2 {-100.02/4 0.1} Kc1 86. Kh1 {-100.01/2 0.1} Kc2 87. Kh2 {-100.02/4 0.1} Bc7+ 88. Kh1 {-100.02/4 0.1} Qe2 89. Kg1 {-100.02/4 0.1} Kb2 90. Kh1 {-100.01/2 0.1} Kc1 91. Kg1 {-100.02/4 0.1} Kd2 92. Kh1 {-100.01/2 0.1} Kc2 93. Kg1 {-100.02/4 0.1} Bd8 94. Kh1 {-100.02/4 0.1} Bf6 95. Kg1 {-100.02/4 0.1} Bd4+ 96. Kh1 {-100.02/4 0.1} Qf3+ 97. Kh2 {-100.02/4 0.1} Bf6 98. Kg1 {-100.03/6 0.1} Qe2 99. Kh1 {-100.02/4 0.1} Bc3 100. Kg1 {-100.02/4 0.1} Kd1 101. Kh1 {-100.02/4 0.1} Be5 102. Kg1 {-100.02/4 0.1} Bh2+ 103. Kh1 {-100.03/6 0.1} Bc7 104. Kg1 {-100.02/40.1} Kc1 105. Kh1 {-100.01/2 0.1} Kb2 106. Kg1 {-100.02/4 0.1} Be5 107. Kh1 {-100.01/2 0.1} Kc2 108. Kg1 {-100.02/4 0.1} Kd2 109. Kh1 {-100.01/2 0.1} Qf3+ 110. Kg1 {-100.02/4 0.1} Bd4+ 111. Kh2 {-100.02/4 0.1} Bf6 112. Kg1 {-100.02/40.1} Qh3 113. Kf2 {-100.01/2 0.1} Kc2 114. Ke2 {-100.03/6 0.1} Bd4 115. Ke1 {-100.02/4 0.1} Qh2 116. Kf1 {-100.01/2 0.1} Qg3 117. Ke2 {-100.01/2 0.1} Qg2+ 118. Ke1 {-100.01/2 0.1} Bf2+ 119. Ke2 {-100.02/4 0.1} Bc5+ 120. Ke1 {-100.01/2 0.1} Kb2 {Draw by fifty move rule} 1/2-1/2
After move 70. from Black (Ktulu) we have this position:
And Black (Ktulu) can´t win in the next 50 moves! No excuse! Not to win KQB vs K can´t be excused! I vote for disqualifing Ktulu! When I see this from Rybka in a tournament because of EGTB bug, I would vote for disqualifing Rybka. There is no place for engines with <1000 Elo in computer chess tournaments! Computer chess becomes absurd! Ktulu makes place 4! out of nine! Every serious chess player only shaking the head. And when he hears Ktulu has Elo performance greater than 2800 he knows, what he ever has known about computer chess!
BULLSHIT!!
Just plain sad.
edit: what's even worse is that it's a mate in 4!
In fact Hans Secelle, who operated Ktulu in this tournament, wanted to withdraw after this game. But as that would the tournaments with two byes each round, he continued.
Richard.


Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill