Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / WCCC 2007 report from Jeroen's point of view
- - By Venator (Silver) Date 2007-06-19 12:21
Hi all,

First of all a big thank you to all the members of the Rybka team! The WCCC was running smoothly and with the new Rybka version on fast hardware we finally made it to win the most important title of the year. Also I'd like to thank all the Rybka fans for their support and help, it is always a great pleasure to see so many people being interested in the Rybka games!

Now some comments on all the games from my point of view:

Round 1: The King - Rybka 0-1
Like in the Dutch Open 2006 The King chose a quiet English line. Johan de Koning was afraid we might repeat this game, but that didn't happen. Anyway, I was quite happy to see this, a board full of pieces, all 8 pawns still present and a strategical battle coming up. Until 19... Rfd8 there was not much going on yet, but The King went f2-f4-f5? to gain space and that gave white a lot of weaknesses. A bit later Rybka punched with 24... b5! and a bit later with 30... d5! The King gave its queen for 2 rooks, but the black queen was 'boss' and after that exchange the game didn't last long.

Round 2: MicroMax - Rybka 0-1
MicroMax is really amazing, it consists of a mere two A4 pages of code! Incredible that such a small program can play chess. It plays without book, though, so I didn't count on a heavy theoretical battle in this game. MicroMax allowed Rybka to weaken its position considerably and after that black quickly won material and a bit later the game.

Round 3: Rybka - Zappa draw
Already in round 3 the big clash between Rybka and Zappa took place. Erdo has been experimenting with the French lately, so Zappa's 1.e4 e6 was not really a surprise. More of a surprise to me was '5.Nxf6+ Zappa out of book'. That is rather early and I don't like the black setup very much. Zappa took back 5... gxf6 and Rybka's last book move was 9.Re1. A bit later Zappa sacrificed a pawn with 11... Rg8, after which white simply must be better. The rest of the story has been told by Vas already. What striked me most in this game was the difference in evaluations between the top programs after 29.gxh3: most of them think it is equal or even better for black! Really amazing. Rybka proved right, outplayed Zap and reached a winning ending. But alas, it still ended in a draw.

Round 4: Loop - Rybka draw
A game that still puzzles me. 3.Bb5+ indicated that white wasn't interested in a theoretical dual, prefering a quiet position instead. Instead of the immediate equalizer 3... Bd7 I chose the slightly less strong 3... Nc6!? to keep more play in the position and to avoid drawing continuations. 9... g6 was end of book for both, as I didn't have 10.Nd4 in the book. Nevertheless this looks like easy play for black. After 17... Nd6 I already started of thinking of a black edge. The black knight keeps the d-file closed, so all is set for a minority attack on the queen's side by b5, a5 etc. Black even has a f5 plan in reserve. Rybka chose to exchange all pieces instead, leading to a slightly inferior ending, that was drawn a bit later. Note the weak move 34.a5?! that spoils the queenside majority. Still, I think this was a missed chance. Black could have easily pushed white with the queenside plan, while there is not much white can do.

Round 5: Rybka - Jonny 1-0
Jonny went for the Marshall, which was declined by Rybka´s 8.h3. After 9.d3 Jonny was out of book, again rather early for such a difficult variation. In the early rounds I really got the feeling opponents didn´t want to make many book moves against Rybka, which I don´t mind BTW :-). Anyway, I think 10... h6!? might be best, some moves later Jonny allowed 12.Bg5 and from that point the game started to go downhill for black. 15.d4! was very strong and after that Rybka took complete control of the kingside by sacrificing its b-pawn. 28.f3! gave a huge fail high and the game ended soon after that.

Round 6: IsiChess - Rybka 0-1
Another 'take Rybka quickly out of book' effort, that simply backfired. 6.Be2 against the Najdorf is solid, but certainly not critical. I was well prepared for it. After 8... Nc6 Isi was on its own and it started to place pieces in awkward positions. Rybka took the initiative and already after 15... d5! there is something smelly about white's position. It didn't take long for black to cash in, first by winning an exchange and somewhat later by penetrating into Isi's position.

Round 7: Rybka - The Baron 1-0
The first two games between these two were not easy for Rybka: in the Dutch Open 2006 Rybka needed a bad Baron blunder to win and in the ICT Leiden tournament earlier this year Rybka even spoiled a winning rook ending. This time it went much more smoothly. The 3.e5 and 4.Nf3 line against the Caro-Kann gives white more space and easy development, while black still has to make sure where to put his pieces and his king. 9.dxc5! was played from book and The Baron tried to build up a wall with some exchanges and 15... g6. The right plan is of course g4! and f5! and I was happy to see Rybka finding this plan quickly. 20... Qxc8? was the losing mistake, it was absolutely necessary to take Nxc8 and follow up with Bd8, to get rid of the annoying a5 knight. White is still clearly better, of course. In the game Rybka could keep its knight on a5, which delivered the final blow 30.Nxb7+! with a winning position.

Round 8: Deep Sjeng - Rybka 0-1
Again no messing around with sharp stuff, instead a positional QP opening arose with the slightly passive 9.Bd2 by white. After that I see absolutely no problems for black, so does Rybka. While black was improving its position move by move, Sjeng really had no clue what to do. Qc2-f5-h3 looks strange, where a normal approach would be to try pressurising the black d-pawn. After the weakening 22.f4 the position already became critical for white, a few moves later it was already lost. Rybka's pieces entered the white position one by one, winning the queen for rook and bishop. Another fluid win.

Round 9: Rybka - Diep 1-0
A cushing victory in only 29 moves, that is very rare in computer games these days. I counted on a Najdorf by Vincent heavily, but no, 2... e6 was played. The opening went fine for white and 15.f5!! uncorked some great fireworks. Rybka didn't bother to defend against the threat Nd2, putting the e5 pawn en prise in the process. Diep went for the material, but ended up in deep trouble after missing 18.Ncxb5! The remainder of the game is simply beautiful, Rybka being a rook down and winning everything back with heavy interest. Superb game, I loved it!

Round 10: Gridchess - Rybka 0-1
Another 6.Be2 Najdorf with the topical 9.Kh1, but 9... Nc6 immediately threw Gridchess out of book. Rybka's last book move was 11... exf4!? counting on 12.Rxf4, which was still in book. I did quite some analysis on this line, trying to find ways to keep as much play in the position as possible. I think this worked very much OK here. With all pieces on the board Rybka feels black is OK and Grid exchanged its e4-pawn for black's b-pawn. Queens were exchanged a bit later, with still a level position. 26... d5! was a great Rybka move, which seemed to throw Grid a bit off guard. 28.Rff1? was a big mistake, Rybka's eval went up immediately and never looked back. It seems programs have difficulties to evaluate the resulting N+2P vs rook positions properly. Clearly Rybka evaluates this much much better. After 46... Rxg4 white might as well have resigned, but decided to take the bait until mate at move 93.

Round 11: Rybka - Shredder 1-0
The story before this game is quite amusing and I will tell you this soon. Now I can tell you that I have been carefully keeping this tricky line in reserve. When I saw the pairings I knew the 11th round was the one to place the 'bomb'. Actually I was 90% sure the Poisoned Pawn with 10.e5 would be played today, my opponent probably spending a lot of time on 6.Be3. After 14.Rd1 Shredder was on its own already and I knew the game would be 1-0. For a computer program it is simply impossible to find its way through the myriads of traps and tricky lines, with such limited time available. No program will consider the exchange sacrifice 20... Nc6 from Anand-Van Wely (Corus 2007), why give up the exchange when there is no white win to be seen!? 22... fxe4 was already losing, 22... Nc6 from Junior-Fritz is a bit better and at least holds out longer. Still, it is weird Fritz played this losing line :-). Last Rybka book move was 25.Bg4, but this was actually a bad move! Curse my slow Dual core machine, 25.Bb4! kills black instantly, as somebody reported in the Rybka forum. In any case, now the game lasted a bit longer, but the result was never in doubt. A sweet revenge for the loss in Turin against the ancient Urusov gambit.

All in all I am happy with Rybka's and my performance. When you look at the games, most of them were over before move 30. Rybka simply outclassed the opposition and she is a worthy 2007 champion. Great work by Vas and Larry!
Parent - - By Vasik Rajlich (Silver) Date 2007-06-19 13:56
Thanks Jeroen for the comments.

In theory we should look at that Loop game at some point. For now I can forgive Rybka for this game :)

Parent - - By Venator (Silver) Date 2007-06-19 14:18
Yeah, no problem. All in all Rybka is also a bit human :-).

BTW, there might be a simple explanation why Shredder was out of book so quickly: perhaps it had
the Motylev-Anand and Anand-Van Wely games in its book and the 'minimum games' set to 3. In that
case there will be no GUI priority for a book move. In such a case the book author has to activate the
book moves himself. On the other hand 10.e5 is a very critical move and I must confess that I cursed
Junior for playing this a few days earlier against Fritz:-). Luckily it didn't spoil the fun!

Best regards, Jeroen
Parent - - By Felix Kling (Gold) Date 2007-06-19 14:46
If Rybka would win all the games it would look somehow arrogant, making a draw makes her more likeable ;)
Parent - - By Nelson Hernandez (Gold) Date 2007-06-19 18:22
The key sentence in Jeroen's report is this.  It stands out head and shoulders above the others:

"When you look at the games, most of them were over before move 30."

Indeed it was so.  Which ought to open some people's eyes!  To make an analogy, the Rybka engine is a knockout punch.  Spectacular, and what is shown in the highlight film.  But Jeroen's book is a persistent and punishing left jab which sets up that finishing blow.  In my view, that is what needs to be studied and emulated.  If you gave some of the Amsterdam competitors Jeroen's book they just might have won instead!

Rybka is a great engine, no doubt about it, but it isn't invincible in every single game.  But opponents who need to run through this opening gauntlet before they meet the Rybka engine are already so softened up that they haven't got a chance.  My conclusion is that the best response to 1.e4 is something that gets Jeroen out of the game very early and then beats down Rybka with a very deep trap line.  In other words, you have to Jeroen them before they Jeroen you!  Otherwise you'll be really badly Jeroened!
Parent - By Banned for Life (Gold) Date 2007-06-19 18:28
It would seem like common sense to avoid modern theory when going up against team Rybka in these tournaments, but most of the other teams are slow learners. It seems pretty obvious that Shredder's victory against Rybka last year in an offbeat opening is a much better model for success than trying to go toe to toe with Jeroen.
Parent - By Felix Kling (Gold) Date 2007-06-19 18:39
Ok, but as Jeroen also said, Rybka had equal positions with black sometimes and then Rybka showed that her eval is simply better than the eval of other engines. If you would play without an opening book you would get results similar to the rating lists like CCRL, CEGT, ... then you could stop playing those tournaments.
Parent - By turbojuice1122 (Gold) Date 2007-06-19 19:03
Amusingly, even the two draws were wins before move 30 :-)
Parent - - By Uri Blass (*****) Date 2007-06-20 09:42
I disagree

It is easier to make a book when you use the best engine.
The opponents have no good choice because if they go to quiet lines they can probably expect rybka to beat them thanks to the fact that rybka is a better engine and the book writer may need to write book moves only for lines that rybka is relatively not strong by herself so he can concentrate on less lines when opponents with more weaknesses need to concentrate on more lines that the program is weak so they can do less preperation in the same time.

I do not think that other engines could win with Jeroen's book and I do not remember Rebel winning tournaments when Jeroen was the author of Rebel's book.

Parent - By Banned for Life (Gold) Date 2007-06-20 13:47
A great book is a very important aspect of these tournaments, but its certainly not the only important aspect. With Jeroen on another team, I could see Rybka scoring 10-15 % less and another team scoring that much more. Rybka would still be the dominant engine, but would lose more short tournaments. Its not really a black and white issue.

Parent - - By Venator (Silver) Date 2007-06-20 15:23
Not winning tournaments with Pro Deo, but still a shared 1st place in the Dutch Open a couple of years ago
and a 2nd place behind Shredder in the Leiden ICT tournament 2 years ago.
Parent - - By garyf919 (**) Date 2007-06-20 18:12
Ah, I sent this question to the wrong party below.


Will there be a new Rybka book release coming some time
in the near future? What are your plans for future book versions?

It seems that since 2.3.2xx are the last releases before release 3,
that there would be one more final book release to go with 2.3.2?
Parent - By Venator (Silver) Date 2007-06-20 18:37
Hi Gary,

I don't know yet! I have started with RybkaIII, but we have no release plans for it in the
near future. But of course it will come one day :-).

Best regards, Jeroen
Parent - - By Vasik Rajlich (Silver) Date 2007-06-21 14:24
It's not that easy to just get out of main-line theory and still have a good position. In fact, this is what Erdo has been trying against us - it got him killed in Leiden, and barely a draw in Amsterdam.

Parent - By lkaufman (*****) Date 2007-06-21 14:40
Well, it's very easy when you have White (for example, IM Eric Prie regularly plays 1d4 d5 2a3 with excellent results) , it's only difficult with Black.
Parent - By Venator (Silver) Date 2007-06-21 15:01
'You'll be Jeroened'. I like that one :-)
Parent - - By JMand (*) Date 2007-06-19 17:10 Edited 2007-06-19 17:12
Round 4: Loop - Rybka draw
... Instead of the immediate equalizer 3... Bd7 I chose the slightly less strong 3... Nc6!? to keep more play in the position and to avoid drawing continuations...

Is that a rule, that the human operator can choose moves during the opening ? Anyway great job !
Parent - - By Fulcrum2000 (****) Date 2007-06-19 17:28
No, he means he prepared the book to play Nc6 when encountering this position.
Parent - - By JMand (*) Date 2007-06-19 18:40
ok okay. Not quite the same thing.
Parent - By garyf919 (**) Date 2007-06-19 22:41

Will there be a new Rybka book release coming some time
in the near future? What are your plans for future book versions?
Parent - - By er (**) Date 2007-06-20 09:16
Rybka - Shredder 
the weak move 22.. fe4, no PC can find it.

For Rybka 232 / 32bit
on my old Pentium 4 (3 GHz)
it took 3h 46min to find the much better move 22..Nc6.

Gerd Möller
Parent - - By turbojuice1122 (Gold) Date 2007-06-20 13:09
Of course, fxe4 is suicide, but black is already completely lost by that point anyway.  It is looking like the key is 20...Nc6 as the only move that even gives black a prayer in this variation.
Parent - By FWCC (***) Date 2007-06-20 15:57
Jeroen you can build me a book anytime!Good job and congrats for a victorious performance for the whole Rybka team.I'll be waiting for my special book.
Parent - - By premraj_n (**) Date 2007-07-21 07:14
the Rybka vs Shredder B97 game needs a good analysis so as to find if black will "always" lose if it accepts the Poisoned Pawn. If someone has done please post the main line if black could win or draw.
Parent - - By Lee Ma Hong (**) Date 2007-07-21 14:03
hmmm ... if someone has found a way in Rybka-Shredder B97 line for black to win or draw instead of losing, he won't make the analysis public but keep it for future use
Parent - By premraj_n (**) Date 2007-07-21 15:22
I think few people would have found it and also i am analysing too.
I knew they would keep it secret.
Up Topic Rybka Support & Discussion / Rybka Discussion / WCCC 2007 report from Jeroen's point of view

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill