Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka @ Paderborn 17th IPCCC
- - By Vasik Rajlich (Silver) [hu] Date 2007-12-30 13:56
The 17th Paderborn IPCCC has finished and Rybka has finished in a shared first, together with Hiarcs and ahead of Cluster Toga.

The full Rybka team was:

Operation: Hans van der Zijden
Hardware: Lukas Cimiotti
Opening book: Jeroen Noomen, Dagh Nielsen
Engine: Vasik Rajlich, Larry Kaufman

I'd like to congratulate the Hiarcs team for their performance. I'd also like to thank Ulf Lorenz for pulling the tournament together and the Rybka team members for all of their hard work.

Our information center about the tournament is here
All games from the tournament can be downloaded (***)

The latest Rybka version is actually looking quite good, but she had her share of problems in this event. One bright spot for us was two book wins with black - has that ever been done before?

Here is a brief rundown of the games:

Rd 1, Gandalf - Rybka 0-1: A long theoretical line led to a position where Rybka could (and did) go into a slightly better endgame in which white had a weak d-pawn and passive bishop. Black tied white down, activated her rooks and pieces, and white's position collapsed.

Rd 2, Rybka - Hiarcs 1/2-1/2: An offbeat opening led to a solid position which was somewhat better for white. Rybka quickly dissipated it with a sequence of inaccurate moves, as she overlooked the strength of 14. .. Bxf3! This is a difficult search issue and Hiarcs deserves credit for handling it correctly. The resulting position was quite solid and dry - the pawns were fixed and neither side could do anything.

Rd 3, Jonny - Rybka 0-1: The first black book win. Jonny was out of book early in a seemingly uncritical position, but blundered with the losing 9. Nxb5? The refutation of this move is quite deep, I'm not sure that Jonny can be faulted for this error.

Rd 4, Rybka - Shredder 1-0: Sandro and Jeroen resumed their decade-long book wars in another ultra-sharp Sicilian. I like white in this early middlegame, although Rybka's play was unconvincing. She took the positional route, sticking a bone in black's throat on h6, but black soon whipped up counterplay on the queenside. I'm not sure what the truth is, but Rybka defused the attack with an exchange sacrifice, centralized her knights, and her queenside pawns gradually settled the game.

Rd 5, Cluster Toga - Rybka 1-0: An ultra-sharp Sicilian line led to a fairly stable position where neither side had any clear way to make progress. Two human opponents would likely have drawn fairly quickly, but, by the margin of a couple of centipawns, Rybka decided to pour gas on the fire with 30. .. Qh3 and especially 32. .. Qxh4, the latter of which appears to be losing. This type of position may require significant changes to handle properly. It is analyzed in more detail here.

Rd 6, Rybka - IsiChess 1-0: After a quiet Spanish opening, black quite logically struck in the center with 15. .. d5 and 17. .. f6. In this case, the center is apparently a target more than a strength - Rybka already reports big scores for herself after 21. Ng3. In round 5, we saw a position where computers have trouble relative to humans. Here, the opposite is the case - I'm not quite sure why black has problems here, but I have little doubt that Rybka's 20-ply search is correct.

Rd 7, Diep - Rybka 0-1: The second black book win. The opening variation is really crazy, I won't try to guess what is happening. Black sacrificed an exchange, but for very promising compensation, and by the time our book ended on move 21, black's knights, bishops, pawns and queen controlled the board.
Parent - - By nine castles (**) [us] Date 2007-12-30 23:09
Thanks for the comments! Vas, can you say where Rybka was out of book in the Hiarcs game? Thanks
Parent - By Vasik Rajlich (Silver) [hu] Date 2008-01-03 15:49
Re1 was the first engine move, I believe on move 11.

Vas
Parent - - By Venator (Silver) [nl] Date 2007-12-31 10:37
Hi all,

Sunday evening December 30th Hans van der Zijden and I have returned from Paderborn. The IPCCC tournament
was excellently organised, the Heinz Nixdorf Museum Forum really is a nice venue.

Thanks to Lukas for using his awesome machine, we only had a minor problem during our first round game, but after
that was fixed the whole tournament ran flawlessly. Also thanks to Hans for operating Rybka once again and of
course to Larry and Vas for the new Rybka version.

As Vas already made short comments on the games, I try to focus on the openings in the IPCCC and some other
aspects of the games:

Round 1, Gandalf - Rybka 0-1: Before the tournament I did some work on the Semi Slav, so it was nice to see this played right in the first round. Gandalf went 5.Bg5 h6 6.Bxf6, a positional approach. 6.Bh4 - the Moscow gambit - can be really murky. White's setup with 11.Rc1, 12.Re1 and 13.Qe2 followed by e4 and d5 looks rather harmless to me. After the none standard 16... Bxc3! black already has a slight edge. Of course Gandalf could have defended better, but it simply thought white was very much OK with the passer on d5. I think that 28.Rcd3? was really bad and from that point Rybka quickly increased her advantage and neatly outplayed Gandalf in the ending.

Round 2, Rybka - Hiarcs 1/2-1/2: A very risky opening line by Hiarcs payed off, as Rybka didn't play it very well. It is essential to flick in h2-h3! at some point, after which black has problems in activating his pieces. 11.Re1, 12.Na4 and 13.Qb3 was based on the trick 13... Bxf3 14.Bxf3 Rxd4 15.Be3 Rd8 16.Nb6!, but Hiarcs avoided it and after 14... Bxf3 and 15... e5! black has equality. After 42 moves a position arised in which white nor black could do something sensible and as Hans and I were suffering from being ill, we decided to call it a day and offer a draw.

Round 3, Jonny - Rybka 0-1: Another Semi Slav with 5.Bg5 and 6.Bxf6, where Jonny opted for 7.Qc2 instead of 7.e3. Its book ended one move too early - at move 8 - and Jonny went for the material gain with 9.Nxb5?, where 9.a4 is the standard move. Rybka played the refutation from book and 12... Bd7 was the last book move. After 14... Bxf3! - another none trivial bishop for knight exchange like in game 1 - followed by 15... Qf5 white is already lost. Still, it was nice how Rybka wrapped it up. 23... bxa3! with the point 24... Qb8+! and 25... Nxd4! is difficult to spot for a human.

Round 4, Rybka - Shredder 1-0: No more toying around with the Najdorf, this time the Shredder team chose a sharp Rauzer system. I am not so sure, but it might be that 11.h4 is more exact than 11.Kb1. In any case, Shredders reaction with 16... d5?! looks very doubtful to me and after that white is fine. White's attack was progressing much faster and Shredder's eval dropped to -0.80. Alas, Rybka made a typical mistake with 22.h6? closing the position and handing over the initiative to black. 26... b3! was a fine pawn sac, but somehow Rybka hold on with the very important defensive move 30.Bh3! to prevent Bf5 coming. After the exchange sacrifice for two pawns black looks better and after showing 0.00 for many moves Rybka's eval dropped to -0.37 after 45.Qe4 was played. Rybka expected 45... Qb6! but Shredder went 45... Rc7? after which 46.Nb4! equalised. In the end Shredder pressed too hard for a win, avoided several drawing opportunities and finally overstepped the limit with the bad move 66... Ba3+? A tough and exciting game.

Round 5, Cluster Toga - Rybka 1-0: The 12... Nh5 was prepared for Paderborn, but Cluster Toga reacted excellently with 13.Kb1! transposing to a line I very much feared. When I knew Toga had such a good book playing this line, I never would have chosen this. Alas, it was too late and after 19.h4 a typical position arises where white has two strong bishops and positional pressure for the pawn deficit. I would say this is two results only, white wins or black manages a draw. After the somewhat strange 25.c3 black must have equality, but Rybka misplayed the position by going for the kingside pawns, giving white a hugely strong passed a-pawn. After that I had no hope Rybka could save the game, although somebody reported black did very badly at move 62-64 and could have put up a tough resistance. 6 piece TB only showed -1.5 in those positions, so maybe somebody can figure that out. Of course I have to give credit to Toga too, as it played an excellent game and outplayed Rybka after the unfortunate queen sortie. Still I was very irritated at myself, throwing away our 1 point advantage.

Round 6, Rybka - IsiChess 1-0: Isichess played 10... Nd4?! against the Anti-Marshall, which is not really good. Still, Rybka's 14.a4 and especially 17.Be3 and 18.Qc2 puzzled me. After the bad 19... Bd6? Rybka immediately showed high scores for herself, but nobody understood why! After 25.Ne4! it became clear black is in big trouble and Rybka had no problems in wrapping up the game quickly.

Round 7, Diep - Rybka 0-1: A strange incident at the beginning of the game. We were already 15 minutes late, when Diep's author announced he was going to postpone the beginning of the game, as he wanted to download a new book. This was dismissed by the organisation and Diep's clock was started. At that moment Vincent played 1.d4 and pushed our clock, while Diep itself hadn't played any move yet. Of course an operator can only execute moves played by the engine, so this was completely against the rules. After some hesitation another book was loaded and we saw another Slav Defence, this time the Meran. 15.Qf3 and 16.Qh3 is a rather new idea, but I already found out the refutation when preparing this at home. 17... hxg5 is a logical and strong exchange sacrifice and after the last book move 20... Qh5 white is already in big trouble. Rybka's eval went high very quickly and the nicest position of the game IMO is after 32... Nd5. Incredible how the black pieces control the complete white position! After 39... Bc7 Diep's eval showed more than -5, so Vincent resigned.

In the end we ended on 5.5 points, just like Hiarcs. Hiarcs won because of the better Bochholz score, something we already knew before the last round was started: a simple calculation showed that a Hiarcs victory was enough to clinch victory. Of course I'd like to congratulate the Hiarcs team for this splendid achievement! Mark, Harvey, Sebastian, well done!

Next year there will be another Paderborn IPCCC, but sadly enough that will be the last one. I enjoyed being there, I had a nice time, although the first 2 days were rather hard because of the flu. I didn't see much of Paderborn, chosing to watch the PDC darts and K1 final in Tokyo instead. While the K1 final was won by Semmy Schild (Dutch, for the 3rd time in a row, somehow we Dutch have great fighters!), the PDC darts was a bit of a deception with Barney losing in the 1/8 finals already. IMO they should not program the IPCCC during the worldchampionships of darts!

Well, in any case I hope you liked our live coverage and of course don't be too depressed because of '2nd spot only'. See you next time!

Jeroen
Parent - By Harvey Williamson (*****) Date 2007-12-31 11:03
Hi Jeroen,

Thanks for a nice report. After your game with Shredder I decided to make our book play h4 instead of Kb1, just so we played a different line. - I am not sure which is best. We both won :)

Happy New Year,

Harvey
Parent - By Nick (*****) [gb] Date 2007-12-31 11:34
Thanks for the fun games and reports Jeroen and Vas.   Congratulations to the HIARCS team also :)
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2007-12-31 13:40
Yesterday i talked to Clemens Keck on Playchess. He told me, he helped Toga with the book and with some tests.
Clemens is very successful on Playchess in engines games and I assume his book is mainly based on Rybka games.
I think, his work has helped Toga a lot.

Best regards and a happy new year
Lukas
Parent - By FWCC (***) [us] Date 2007-12-31 18:02
Thanks for the fine report Jeroen!Keep up the brilliant book work!
Parent - - By pawformation (*) [us] Date 2007-12-31 18:48
So Vas,

Were there any specific moves that you can point to and say, "this is due to the recent changes in search/evaluation"?
Parent - By Vasik Rajlich (Silver) [hu] Date 2008-01-03 15:50
I'm sure there are many, but I haven't compared the two versions like this.

Vas
Parent - By utzinger (*) [ch] Date 2008-01-03 14:20
Hi Vas
Many thanks for this nice report. It helps very much to
understand more deeply the games.
Kind regards
Kurt
Parent - - By thierrycatalan (**) [fr] Date 2008-01-03 14:52
Only 7 round is perhaps not enough to see which is the best engine ... 9 round or 11 rond will be better ...

It is also damage that the book's openning  are so important , so weight,  in the  games and in the result of competition ....

perhaps Paderborn is no more a competition between chess engines ...but between chess books opennings ??

so it will be nice that all the engine have the same book.. a little book with neutral positions. ??
Parent - By Vempele (Silver) [fi] Date 2008-01-03 15:30

> Only 7 round is perhaps not enough to see which is the best engine ... 9 round or 11 rond will be better ...


9 rounds (or some multiple of it) would have been best, but 11 is just wrong with only 10 participants.

>perhaps Paderborn is no more a competition between chess engines ...but between chess books opennings ??


Chess entities.
Parent - By Vasik Rajlich (Silver) [hu] Date 2008-01-03 15:52

> perhaps Paderborn is no more a competition between chess engines ...but between chess books opennings ??


It's a sport, the goal is to win. Science is for the rating lists.

Vas
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka @ Paderborn 17th IPCCC

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill