Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka @ Amsterdam 15th WCCC
1 2 Previous Next  
- - By Vasik Rajlich (Silver) [hu] Date 2007-06-18 16:03 Edited 2007-06-19 13:16
The 15th ICGA World Computer Chess Championship has finished and Rybka has won with 10/11, ahead of Zappa, Loop, Shredder and Gridchess.

The full Rybka team was:

Operator: Hans van der Zijden
Hardware: Lukas Cimiotti
Opening book: Jeroen Noomen
Engine help: Larry Kaufman
Engine: Vasik Rajlich

Rybka is a team effort, with a lot of really hard work put in by many people. This was never more evident than here. Special mention this time goes to Larry Kaufman and Jeroen Noomen.

Larry Kaufman's contributions to Rybka tend to be subtle. Here, his revamped handling of material imbalances guided Rybka's play in two games (vs Zappa and vs Gridchess), with the new formulation proving superior in both instances.

Jeroen Noomen has consistently proven himself to be the best in his craft, and this tournament underlined this in spectacular fashion. We are accustomed to white positions which range from much better to winning and to black positions which range from dead equal to clearly better. We are accustomed to top teams ducking the theoretical battles. I can't even remember the last time we got the wrong end of the theoretical debate. The new twist this time around was the stunning tournament-clinching opening book knockout of the well-respected Shredder book author Sandro Necchi in round 11.

I'd like to thank:

- The ICGA for their organization of the event
- The other teams who took the time to participate
- The members of the Rybka team for their hard work

The tournament games can be found here:

http://www.rybkachess.com/download/wccc_2007.pgn

The full information about the tournament can be found here:

http://www.grappa.univ-lille3.fr/icga/tournament.php?id=173

Below is a brief round-up of Rybka's games. Jeroen is planning more detailed comments a bit later.

Rd 1, The King - Rybka 0-1: From a quiet English opening, white accepted a clumsy kingside pawn structure by playing 17. Nce4 and later 20. f4 and 21. f5. Black was able to take over the initiave on the queenside and in the center, getting in all three pawn breaks - .. c5, .. b5, and .. d5 - and white could not hold the position.

Rd 2, Micromax - Rybka 0-1: Rybka obtained a very strong knight pair in the opening and white's position collapsed.

Rd 3, Rybka - Zappa 1/2-1/2: Quite an amazing game. The opening went well - Rybka was not confused by Zappa's rare opening sideline and got a considerable advantage, leaving black to try a desperate pawn sacrifice to complicate matters. The middlegame also went well - Rybka was able to simplify into a much better and probably winning endgame. Even the first part of the endgame went well - black's chances lay in setting up a blockade, but white was able to crack open the position with a timely d5-d6 pawn sacrifice. Amazingly, and paying tribute to Zappa's strength as well as to the resourcefulness which chess can allow, all of this was not enough. Zappa dug in and found one last hurdle for Rybka to clear - a theoretically winning but tricky KRPKNP tablebase position. Rybka was not equipped with this tablebase and could not come up with the solution within the tournament time limits.

Rd 4, Loop - Rybka 1/2-1/2: The opening seemed very promising - on move 19, Rybka had a solid position, extra central pawn, and the chance to push her pawns on both sides of the board. For unclear reasons, she declined this opportunity. Instead, she traded off all pieces and went into a nominally worse but easily drawn endgame, which was in fact drawn without any excitement.

Rd 5, Rybka - Jonny 1-0: In a seemingly quiet Spanish position, black volunarily allowed his kingside pawn structure to get shattered by playing 12. .. Nxb3. This move doesn't look terrible at first sight, but in fact it seems to lead black into serious trouble by force. White put her finger on the weaknesses around black's king and black could not defend against the threats.

Rd 6, Isichess - Rybka 0-1: In a strategic Sicilian, with both sides castling kingside, white placed his pieces slightly awkwardly - a2-a4, Bf3, Bd2, Nc1. Rybka struck in the center, took over the initiative, and white could not avoid the loss of material.

Rd 7, Rybka - The Baron 1-0: With a closed French center, black went for the thematic pawn break .. c5. In this particular position, this had the unpleasant side-effect of allowing white to play dxc5, Nd4, and f2-f4-f5. Perhaps black could have somehow contested white's idea, but the Baron acquiesced to white's plan by playing 13. .. Nxd4 and was left with a strategically lost position.

Rd 8, Sjeng - Rybka 0-1: In a balanced isolated queen pawn position, white had trouble finding a plan, placing his queen and both bishops on awkward squares. Black expanded on the queenside, provoked white into weakening his kingside, and eventually broke through with her pieces.

Rd 9, Rybka - Diep 1-0: Black's book ended abruptly on move 8 in the middle of a theoretical Sicilian minefield. I don't know the latest theoretical variations, but black could not navigate the waters and his position was lost quickly and resignable after 17 moves.

Rd 10, Gridchess - Rybka 0-1: Gridchess was the most novel of the participants here - apparently, it is Toga with some sort of a clustered parallel search. It's hard to draw any real conclusions from just this tournament. In a strategic Sicilian variation, Rybka was able to hunt down black's dark-squared bishop with her knight. The position seems hard to win, though, and Rybka does not even give herself an advantage at this point. The fireworks began when black offered (and white accepted) a very unclear pawn sacrifice. The critical continuation would have been 28. Rf5 - the position looks double-edged and I won't dare to give an assessment. Instead, Gridchess, not having access to Larry's work on material imbalances, blundered with 28. Rff1, leading to a RBB vs BNN endgame which was very pleasant for black and which white could not hold.

Rd 11, Rybka - Shredder 1-0: A dream or a nightmare for an opening book author, depending on which side he's on. Jeroen busted out a wild Sicilian line in which white starts the festivities by sacrificing 3 pawns for a sizeable lead in development but still nothing concrete. Rybka herself would favor black at this point. Later in the variation, white offered a piece to keep the attack going. I'm not sure what should happen, but the entire refutation of black's play was in the book and black could have resigned without white playing a single new move.

Vas
Parent - By Dr.X (Gold) Date 2007-06-18 16:10
Vas,

What a wonderful Trophy-Thanks Rybka Team for this great day-

FIDE-Open your eye's to the world champion chess engine "RYBKA"

siderious
Parent - - By Torstein (*) [no] Date 2007-06-18 16:28
Congratulations with a great result!

I am looking at the game Rybka - Shredder now. Any idea how far your book goes in the opening?

Torstein
Parent - - By Felix Kling (Gold) [de] Date 2007-06-18 17:12
Till move 25.

I'm positivly surprised, Chessbase is reporting about the tournament on their german website:
http://chessbase.de/nachrichten.asp?newsid=6714
Parent - - By Dr.X (Gold) Date 2007-06-18 17:43
Hi Felix,

That's not good enough!
Parent - - By Eduard (***) [de] Date 2007-06-19 09:48 Edited 2007-06-19 09:54
The Editor of the news on the english ChessBase Page online (ChessBase.com) is Frederic Friedel,
I believe.
Parent - By Dr.X (Gold) Date 2007-06-19 13:41
Eduard,

Thank you!
Parent - - By Lee Ma Hong (**) [ph] Date 2007-06-18 23:26 Edited 2007-06-18 23:31
I am looking at the game Rybka - Shredder now. Any idea how far your book goes in the opening?

You should read Jeroen's comments in the thread

http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=1364#fp

For example, Jeroen says

Rybka has around 35 to 40 book moves on this line

after 16. Be2
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2007-06-19 10:11
It was "only" 25 moves. I was watching the computer it ran on. And you can easily find out here: http://www.grappa.univ-lille3.fr/icga/round.php?tournament=173&round=11&id=2 ,the time of each move is reported. 8 - 15 sec. is the time the operator needs to pass that move on. Everything that is significantly longer isn´t in the book any more.

Lukas
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-06-19 12:26
This Rybka-Shredder game is extremely instructive, not so much for the line itself, but for what it tells us about Rybka's tournament book.  The book clearly has properties that no conventional book has.  Just follow along with the moves white makes in your own book and look at Rybka's choices.  Would your book have followed that line left to its own devices?  No way, even if you had every game Jeroen has in your own database.  And incidentally, if Rybka stayed in book 24 moves it must mean Jeroen is staying very current on theory as that full line was played for the first time (that I know of) just two months ago.

Shredder's book blunder was very subtle.  12...h6 looks plausible on the surface but it leads directly to the disaster than ensued.  A stronger reply would have been 12...Qxa2 in book terms.  (Remember, you have to judge this in book terms.  What would the book have done, not the engine.  Evaluations are nice to have but they don't tell the whole story.)

Based on these conclusions it is very clear what the rest of the computer chess industry is up against.  Not just clearly the best engine, but also the best book.  It is easy to see that Rybka will be invincible in competitive play for the foreseeable future on its own merits, greatly aided their opposition's stodgy and antiquated ideas.
Parent - By Vasik Rajlich (Silver) [hu] Date 2007-06-19 13:10
Yes, I'll second this. This is the reason Rybka defies statistics (and sometimes Mr. Murphy) in these tournaments. The Shredder game is an extreme example, but we always get a serious advantage with white and at least complete equality with black.

Vas
Parent - - By turbojuice1122 (Gold) [us] Date 2007-06-19 14:58
You ask, "Would your book have followed that line left to its own devices?"

I answer with a definite "Yes!!"  However, if you had asked the question before the Fritz-Junior game, the answer would have been "No."  After that game, I made sure that my book covered those lines, as that is a very important line.

First, Kullberg mentions that Rybka's book for that game only went 25 moves, compared with Jereon saying "35-40 moves".  Assuming that Jereon's book is similar to mine, the 35-40 moves is probably in the line played in the Fritz-Junior game, along with some reasonable sidelines.  There wasn't much reason for going very far with the foolish ...fxe4?? in the book because that basically throws the game away.  Nonetheless, I was curious as to what would happen if black takes the knight, so I added those lines into my book through move 24.  Thus, my engine would have found the resounding Bb4! at tournament time controls.  This is not to belittle Jereon and his hard work--there are a huge amount of other lines in high-level play that I'm quite sure I don't have (based on following the games), but for which he has extensive analysis--it's just that I would have gotten lucky with this line because I did a bit of analysis on it after the Junior-Fritz game.

Second, 12...Qxa2, along with many other moves that people are mentioning, would have transposed into the game.  I specifically checked out 12...Qxa2 after the Junior-Fritz game, and I remember that with a careful couple of "non-book moves", white can put black on its own in a situation where black will naturally play ...h6, thus transposing back into the studied book lines.  I would check it out, but my engine is currently busy doing other analysis.
Parent - - By Alkelele (****) Date 2007-06-19 15:08
Seriously, if people don't stop this discussion of critical lines in public soon, I will make all my analysis public and share my book on emule!
Parent - - By turbojuice1122 (Gold) [us] Date 2007-06-19 15:24
Searching eMule for Alkelele.ctg ...

Come on, now--even Jereon, who has probably performed more analysis in this than any of the rest of us, has been putting forth some lines right in front of the finals in which Vas is competing.  I personally think it would be interesting to discuss these lines in more detail here and see what kind of improvements/refutations to the discussions that people come up with this weekend.

We need a Freestyle tournament in which the variation 10.e5! of the Poisoned Pawn Sicilian is mandatory...
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-06-19 15:53
You think Dagh uses a .ctg book?!  Too funny.  At any rate, here is a perfect illustration of an opening book move that creates mayhem. 

The focus of computer chess amateurs is engines.  Among semi-pros it is hardware.  Among professionals it is opening books.  You will notice this sequence gets progressively steeper in terms of cost.  Anybody can buy or pirate an engine.  It takes moolah to buy or connections to borrow competition-level computer gear, and more moolah to stay at the competitive edge.  But it takes a tremendous commitment of time and effort to develop a professional opening repertoire, more costly than hardware by more than one order of magnitude.  Jeroen is the best at this because he's been doing it for decades.
Parent - - By Alkelele (****) Date 2007-06-19 15:56

>You think Dagh uses a .ctg book?!


I do! And I swear to it religiously! I also depend on a little bit of memory and a few game files, but then I am ready to go!
Parent - By Nelson Hernandez (Gold) [us] Date 2007-06-19 16:22
In that case I too will be looking for Alkelele.ctg on eMule! 
Parent - - By Venator (Silver) [nl] Date 2007-06-19 19:48
Come on, now--even Jereon, who has probably performed more analysis in this than any of the rest of us...

Ugh, you should know how different is the truth :-)
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-06-19 21:16
Thanks for giving us all some hope.  However just because you personally haven't performed the analysis doesn't mean a bank of computers crunching away 24 hours a day hasn't done it for you already! 

General von Moltke (or was it von Schlieffen?) said that there were four kinds of military officers, which could be subdivided into a two-by-two grid.  On one axis, the smart and the stupid.  On the other axis, the diligent and the lazy.  Smart and diligent officers, he said, were perfect for staff work.  Smart and lazy officers were naturally suited for command positions.  Stupid and lazy officers "had their uses", for instance, if you wanted nothing to happen, or had a job to do that was too mundane to assign to a smart officer (this happens a lot in the military).  But stupid and diligent officers were a peril, he said, and must be gotten rid of at all costs.

Jeroen, I suspect you may be suited for command.
Parent - - By davidwhite (***) Date 2007-06-20 06:45
       It     was definitely von Moltke. However,he postulated that the smart and active officers were best at thinking quickly on their feet and were ideally suited for battlefield command(not for staff work) and that the smart and lazy officers,
because they would always seek to obtain their objectives with the least effort and risk, were the only officers he would choose for planning strategy at the high command(General Staff).

So Jeroen should accordingly,I believe, be given his own swivel recliner at the General Staff and continue ensuring that our lines of advance remain unimpeded...no minefields or other nasty surprises !

My CONGRATULATIONS to Vas and the whole Rybka team!!!
I'm totally exhausted and I didn't do anything.

My best to everyone,
David
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-06-20 15:10
Damn you, David!  You ruined a story I've been telling the wrong way for 20+ years!  Would I be correct in concluding that you are a current or former military officer steeped in history?

Bet you can name who spoke these lines without Googling:

Keep the right wing strong.
Do not engage in battle unless you have at least 75% likelihood of victory.
Let us cross the river and rest in the shade of the trees.
I wouldn't give a hoot in hell for a man who lost and laughed.
Thank God I have done my duty.
Parent - - By davidwhite (***) Date 2007-06-20 18:49
Hi Nelson,
             I'm sorry about that but look on the bright side of it...now you'll have an even better
story to tell. I've never been in the military but majored in history in college(along with drinking)
and for the last 40 years,since graduating,have been an omnivorous reader of history.I think I've
actually reached the point where I've forgotten more than I still remember.On more than one
occasion I'd come across that observation by von Moltke and recognized that your take on it
didn't quite square with mine but ,until I'd googled it,I wasn't sure that I was remembering
correctly.
       
            Now,without Googling, I know the 1st,3rd,and 5th of your items were spoken
by von Schliefen,Stonewall Jackson,and Lord Nelson.I've seen the 2nd before but can't
remember who said it but I do recall thinking that,by my lights,a one out of four chance
of losing a battle was too high to force a battle.Who said it?
The 4th one sounds like something Patton would have said but I don't remember it.Was it
Patton? It was undoubtedly spoken by an American anyway.
            Enlighten me,my friend

Best regards,
David
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-06-20 20:48
4 out of 5, very impressive.  #2 was Napoleon, and you're right, a 25% chance of defeat is too much.  You have too many Waterloos that way.
Parent - - By Banned for Life (Gold) Date 2007-06-21 16:29
Actually, the campaign of 100 days was one of Napoleon's finest.
Parent - By Nelson Hernandez (Gold) [us] Date 2007-06-21 18:50
Until the 100th day.  You know, I once looked up the microfilmed London Times after Waterloo and the reportage was so ecstatic it made me smile.  "220 cannon captured!  21 golden eagles!  We cannot imagine how Bonaparte can make good such losses."  On the wall above my computer is a portrait of the Duke of Wellington on horseback rallying the troops during the battle.
Parent - By Venator (Silver) [nl] Date 2007-06-20 14:45
Actually I did a lot of analysis on the Poisoned Pawn. I don't know the amount of time I spent
on it, but it must be somewhere around 100-150 hours. Still, there are Freestylers of whom I
am sure they have done a lot more than I have. That was my point in replying to the previous
poster :-).

At the moment I am doing a lot of analysis on several lines, but my time is of course limited as
I have a normal job. Nevertheless I enjoy it to discover new ideas and new paths.

Anyway, Nelson, thanks for your kind postings!
Parent - - By Roland Rösler (****) [de] Date 2007-06-21 03:02
It´s amazing, it must come from a German! It reminds me to my student time, where one of my professors makes a similar comparison about humans thinking (okay, we all were smart!). On one axis the fast and the slow (thinker). On the other axis the flat and the deep (thinker). I don´t remember exactly, but I´m absolutly sure, that fast or slow wasn´t crucial, but I wouldn´t belong to the flat thinkers. He pointed out, which job will be the best for each category (for example: fast and flat: journalism; we all studied economics). Okay, this is many years ago, and the fast seem to be in advantage nowadays. But I don´t believe it!
PS: I don´t remember the job for the slow and flat thinkers :-)
Parent - - By Roland Rösler (****) [de] Date 2007-06-21 03:06
Maybe politicans.
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-06-21 03:32
Politicians need to appeal to voters in all four grid squares.  Slow and flat is the lowest common denominator, so to establish rapport with every group they avoid communicating fast and deep thoughts at all costs.  But they stick it to us fast and deep, that's for sure.
Parent - By Roland Rösler (****) [de] Date 2007-06-21 04:52
Nelson, I know it before. You are the very best! Very fast and deep!
Parent - - By Vasik Rajlich (Silver) [hu] Date 2007-06-21 14:30
Stupid and dilligent officers also have their uses - they can for example make sure that all obscure 6-man TBs are properly installed :)

Vas
Parent - By Nelson Hernandez (Gold) [us] Date 2007-06-21 18:44
OUCH.  How did you know that I was doing PRECISELY that today?!
Parent - By Dr.X (Gold) Date 2007-06-19 19:48
So, it is fair to say that Jeroen's book coupled with the  knowledge of Rybka's idiosyncratic habits were in combination lethal to Shredder et al. 
Parent - - By albitex (***) [it] Date 2007-06-18 16:57
I am sincerely happy about the result gotten by Rybka.
Vas. and its teams surely deserve this recognition.
They is always available to help all the consumers, and they answers
always to everybody.
I think that the title of CHAMPIONS OF THE WORLD has been this year
conquered by the really more program IN of the moment.
Compliment to everybody .   Alberto
Parent - By Jim Walker (***) [us] Date 2007-06-18 17:28
Congratulations to the entire team !!  Rybka is in a class by herself.  The strongest program does not always win a tournament like this but there is no doubt it happened this time.
Jim
Parent - - By Alkelele (****) Date 2007-06-18 17:26
Congrats on awesome exhibition tournament ;-)

Perhaps you can attach the games from the tournament to your post. They can be found here:

http://f23.parsimony.net/forum50826/messages/167925.htm

Kurt Utzinger has cleaned up the PGN file.
Parent - By Dr.X (Gold) Date 2007-06-18 21:04
Alkelele,

Thanks for the link to the download of the games. Very nice!

regards
robert
Parent - - By alopez [ca] Date 2007-06-18 17:34
Hello and congratulations!

Is the new Jeroen's book going to be available ?

Thanks
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-06-19 16:23
LOL!!  This is the funniest question I've read here since "when is Rybka going to go open source?"
Parent - By turbojuice1122 (Gold) [us] Date 2007-06-19 17:01
Does the answer to both questions have something to do with flying pigs? :-)
Parent - By Michael Waesch [de] Date 2007-06-19 19:30
Easy, when it has become obsolete.

Mike
Parent - - By billyraybar (***) [us] Date 2007-06-21 22:25
If i'm not mistaked rybkaII.ctg is the exact book he used in a past major tournament, so it's not ridiculous to think he will release this one too.  :)
Parent - - By turbojuice1122 (Gold) [us] Date 2007-06-22 02:19
I highly doubt that Rybka II was his best book at the time, though.  He cannot release the book that was used in the WCCC this year--I imagine it has an incredible array of novelties that need to be saved for future competitions.
Parent - - By Lee Ma Hong (**) [ph] Date 2007-06-22 03:31
yes, the 6th freestyle finals is about to start, so it is in the interest of the rybka team not to release the book until after the event.
Parent - By turbojuice1122 (Gold) [us] Date 2007-06-22 04:14
Actually, what I'm saying is that I don't think we'll EVER see the book that was used for this year's WCCC--at least, not for at least a year or so, after the secrets involved in many of the currently unused lines in that book have been divulged by others who have stumbled upon them or when the Rybka team has been forced to use them.
Parent - By plicocf (***) [br] Date 2007-06-18 18:42
In my opinion just Loop get hard to Rybka, whit an equal game.
After this what I can make?

Congratulations for Rybka team !!!

Paulo Soares
Parent - - By robertj (**) [gb] Date 2007-07-10 11:43
Congratulations! It as a pity that Rybka did not have KRPKNP tablebase. How large is the database of all so far generated tablebases? Thanks.
Parent - - By Vasik Rajlich (Silver) [hu] Date 2007-07-10 13:31
The full 6-man set is 1,200 GB.

Vas
Parent - By Lukas Cimiotti (Bronze) [de] Date 2007-07-10 17:10
I tried starting Rybka with lots of tablebases (~250 gb) on a normal 500 gb SATA drive. Fully loading engine takes 5 minutes, total memory usage is 3.28 gb, hash set to 128 mb only. With only 30 gb of tbs it takes 5 seconds and total memory usage is 813 mb. So i think, it is practically impossible to use all 1.2 terbytes of tbs. And there is a lot of worthless stuff in my collection - e.g. who needs kqrpkp ?
Parent - - By premraj_n (**) [in] Date 2007-07-21 11:15 Edited 2007-07-21 11:45
Hi Vas,
I think Rybka can improve largely in endgame knowledge and also use the 5men TBs effectively.
For example if the game have 7pieces left if Rybka tries to see how it can reach the 5piecer by playing any "forced moves" that
could win or could draw what ever is possible. Its meaningless that you keep saying everyone to have 6-men full, when the loading and
accessing of it would take a long time and maynot even help in the game.

Another point Rybka needs to implement Bishop underpromotion and opposite colour bishop knowledge on endings.
Tell me a concrete reason why you dont want to do it.The reason why i need that is Rybka's eval is wrong on many cases when i sit with
Rybka and analyse opening lines very very deep.

One more point to add I added the 6-men KBPKBP and arranged a position where the engine has to promote to bishop and Rybka didnot
have in its variations to promote to bishop.It was having 7options out of which only 6options it looked in and ofcourse left out variation is
bishop promotion. So even when having the Tablebases its not going to do any good.

Thank you and Best Wishes for Rybka 3,
Prem
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka @ Amsterdam 15th WCCC
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill