Rybka wins 10th ICT in Leiden
31 May 2010, 17.46 CET | Last modified: 15:22 | By Peter Doggers | Filed under: Reports | Tags: Computer chess
Rybka this weekend won the 10th edition of the International Computer Chess Tournament (ICT10) in Leiden, The Netherlands.
By Eric van Reem
In the 10th edition of the International Computer Chess Tournament (ICT10) the latest Rybka version by Vasik Rajlich scored 8 points out of 9 games. Rybka lost only one game, against number 2 Deep Sjeng by Gian-Carlo Pascutto. The Belgian program played a fine tournament and scored 7 points.
Two programs shared the last spot on the winner’s podium: Hiarcs by Mark Uniacke and Shredder by Stefan Meyer-Kahlen. Both programs scored 6 points. Rybka won the “Theo van der Storm” trophy, named after the late CSVN secretary and organizer Theo van der Storm.
On the table below you can see the final results. Remarkable: only 13 of 63 games ended in a draw!
An international tournament it was: we had participants from nine countries: Holland, Germany, UK, Hungary, Brazil, USA, Poland, Belgium and Israel.
Computer chess still attracts many (new) programmers. Last year only eight programs came to play the ICT, but this year no less than 14 (!) professionals and amateurs came to Leiden, not only to play the games, but also to discuss chess in general and chess programming in particular. These lively discussions often inspire programmers to improve their engines.
It was also nice to some new faces in Leiden, like special guest Don Dailey. He is an American computer games and chess researcher and programmer. Dailey is author of various computer chess programs starting in the 80s. He played in Leiden with his latest program Komodo.
Special guest Don Dailey
A newbie in Leiden was the Brazilian with the remarkable name Ben-Hur Carlos Vieira Langoni Junior. He debuted with his program RedQueen. He started working on his chess program just one year ago and promised to come back again! Hopefully he will score more than 0,5 point next time.
Ben-Hur Carlos Vieira Langoni Junior
The next major computer chess tournament will be the ICGA 18th world computer chess championship in Kanazawa, Japan from September 24th till October 2nd. You can find more info on www.icga.org.
In November the Dutch Computer Chess Federation (CSVN) will celebrate its 30th anniversary of the Dutch Open Championship. The exact dates will be announced soon. Please check www.csvn.nl for information about the Dutch Open and the other activities of the CSVN. On the website you can also see many pictures of the ICT 10 tournament and you can download all the games (go to downloads-games-CSVN International Computerchess tournament).
On Saturday a side-event was played for owners of dedicated chess computers. Ruud Martin won this tournament with a Revelation Rebel 5.
We would like to take the opportunity to thank tournament director/webmaster Jan Krabbenbos and arbiter Rienk Doetjes for organizing another trouble-free tournament.
All Rybka games
Game viewer by ChessTempo
Game viewer by ChessTempo
Name tags of the engines at work
A computer game in action: Hiarcs-Komodo
Amir Ban from Israel (Deep Junior)
Rybka4 and Rubik
The programmers all together
The winners, L-R: Gian-Carlo Pascutto (programmer DeepSjeng), Hans van der Zijden (operator Rybka), Harvey Williamson (operator Hiarcs) and Ernst Walet (operator Shredder)
Hans van der Zijden (operator Rybka) with the first prize
Tournament director Jan Krabbenbos
Photos © Eric van Reem
Links
 ShareThis
|  Print
|
Why neither FireBird nor StockFish participated in the tournament?
Firebird , because it is a clone and they are not allowed …
btw Rybka 4 already came out last week
For those of you wondering how come Rybka lost a game,
. well, it’s possible, 
, but in that particular game the Rybka software hit on a bug, which caused it to play a move from depth 7. The bug was fixed during the tournament.
Whatever the plural is for a collection of nerds/weirdos etc. is, it is surely apt here…
A ‘trek’ perhaps…
yes, sure a bug, probably rybka4 contains many many bugs like that…
Hm… Rybka4 on a Jaguar, would that kill chess?
Rybka proves computers still don’t know how to play chess: they have no clue how to play for a draw!
@Evgen: Examples? To my knowledge there’s one bug in Rybka 4 that can strike when doing game analysis leading to very long times till the next depth is reached. Also that particular bug was cluster specific, it’s not in Rybka 4.
@Peter: Chess engines may run slower on those super computer processors, about 30 times slower on blue gene processors according to the Convekta guys. The current cluster is already so strong that even teams of Rybka (on a normal computer)+human have huge difficulties to win a game and can be happy with a draw
See http://www.rybkachess.com/free/RybkaClusterBase.pgn , Rybka did quite well against centaurs
@Bert: you can definitely do something about it, like score trades differently, Rybka has some contempt settings for that, but apparently the other engines don’t have that yet.
Btw.: During the tournament GM Larry Kaufman, who is in the Rybka team but currently focusses on Komodo (Don and Larry are the Komodo team) and was also in Leiden got ill and is in hospital now, but it looks like the operation went well and he will be able to fly home in a couple of days.
Komodo played on only one processor, since the multi processor implementation isn’t ready yet, so the result is quite remarkable.
I disagree Bert, I find Rybka’s play very impressive (what about the mysterious move Rb8 in the Leningrader against Deep Shredder (or is my theoretical knowledge falling short here), or the king’s march Kf8-e7-d8 in the 9th game which looks like something Nimzovich would have been proud of…
You can see here how Deep Rybka 3 fairs against some of the other engines out there: http://www.computerchess.org.uk/ccrl/4040.live/rating_list_all.html
Will be interesting to see how DR4 measures up.
Whoops, it’s already out for DR4, sorry about missing that in last post: http://www.computerchess.org.uk/ccrl/404/
ceann :
i have to admit i enjoy your comments…
always to the point !
I find it interesting that so many games start with 1 e2-e4. In the latest human WCH matches it seems that the top elite players have come to the conclusion that 1 d2-d4 gives better chances to obtain an advantage as white. Elite computers seem to have a different opinion! Is there some statistics available about white winning chances between 1e4 and 1 d4 in computer games?
Frank, you are right Rybka’s play is impressive, and its opponents therefore ought to be weary when playing it, but they don’t seem to be and all fail to hold the balance even when playing with white. QED
@ Chessvibes
I think you guys should change the title and the first sentence of the report, which is misleading. Cluster Rybka won the tournament, not Rybka4.
Cluster Rybka isn’t commercial, you can’t buy it and it is a very different beast from the commercial Rybka4 that you can buy.
Whenever I read comments like the last, I wonder if those ‘friends’ would use the same unfriendly words or tone if they would meet Peter or others live. No, it’s soooo easy and satisfying to hide behind a nick and to be a wisenheimer! Paying for nothing but knowing better everything! If you know it better, make your own page and write articles!
well, the post you answered to has a valid point. the version used was Rybka 4b3 cluster and is really different from Rybka 4. The cluster support is a big part of the code and comparing it to the planned remote Rybka would be more precise. Also the tournaments are won by a team, hardware and opening book are important, so just writing “Rybka” would have been most precise.
@ Deep Mikey
Come on, nothing wrong with the article in general or Peter, I enjoyed it. I’m not trying to be an unfriendly smartass… I really like Chessvibes and their very fair, professional journalism.
Rybka 4 came out a couple of days ago and lot’s of people will probably make their decision whether to purchase it based on reports like this.
They might then be disappointed that Rybka4 – what they just bought – is not the same thing that convincingly beat the whole world. That was a special, private version.
Peace out man.
OK, since the only parts not by Eric van Reem were the header and the intro, which contained that inaccuracy, I corrected it.
From chessbase
“Note that both Rybka and Sjeng played on 128 core clusters; Junior and Hiarcs each had 12 cores, Shredder had eight.”
I don’t understand a lot of computers but isn’t that unfair as compairing the programs? It seems to me that they all need the same cores.
Well david, apparently the ICT was a “Free For All” tournament, with no restrictions on hardware used. But look at what the regulations of the coming computer world championship ICGA 2010 has to say about the issue, first reminding us of what was concluded ages ago, in 2002: “The observation was clear: all kinds of differences between microcomputers, personal computers, “normal” computers, and supercomputers were in some sense obsolete and the classification thus was considered artificial. So was the division into the classes of single processors and multiprocessors.”
This division is now apparently no longer considered only artificial, but rather necessary again, as the computer world championship 2010, held later this year, will be devided in two sections: one called the World Computer Chess Championship, a Free For All like the last ICT, and the second section called the World Chess Software Championship, played on equal hardware.
The latter tournament will be more to you liking david, but it of course remains to be seen whether Rybka will participate in it, or only in the former….
Hi guys!
Sorry about my ignorance but…
It´s only engines?
What about chess book openings? It makes some difference?
Could a chess book determine a winner?
Thanks for attention
I was competing with Hiarcs in Leiden. It was a good tournament this year. I think it is important when reporting such events that the hardware is mentioned to put the results into some kind of context. I pointed this out to the writer of the above press release. i was told in reply that 99.98% of people do not care. It seems from reading the comments above that people do care. I also started a poll to try and prove this statistic. Feel free to vote here http://hiarcs.net/forums/viewtopic.php?t=3467
The information was eventually added to the press release but as far as I can see the above article has not been amended. The report on the ChessBase website includes this line, “Note that both Rybka and Sjeng played on 128 core clusters; Junior and Hiarcs each had 12 cores, Shredder had eight.”
Congratulations to all that took part.
Best Wishes,
Harvey
… mentioning the book authors would also be needed, they have a huge impact on the results.
And don’t forget Komodo was using just one single core!
I think Chessvibes is doing a good job and expecting so much precision is a bit over the top.
Actually I want to add that it’s great that many news sited reported about that tournament at all. Usually computer chess tournaments don’t get much attention.
yes as you know I posted on your forum Felix that Komodo deserves a big mention particularly as CSVN paid for Don and Larry to be there. Of course you can mention the book authors if you want. Although nobody will know who they are. Anyone interested in computers will understand hardware.
A press release from the tournament organisers should not sound like it is a press release from the Rybka team.
>> “Note that both Rybka and Sjeng played on 128 core clusters; Junior and Hiarcs each had 12 cores, Shredder had eight.” … isn’t that unfair as compairing the programs? It seems to me that they all need the same cores.
Imo that’s exactly the main reason why these tournaments get little interest from the chess public in general.
What do I care if some chess programs is good on some supercomputer? I don’t have a supercomputer.
Some years ago noone had a multiprocessor engine – those tournaments are some kind of highend tournaments where you see the future of computer chess.
Note that opening books also have a big influence, you can’t buy those books, too. The versions of the engines used are also the latest private ones, you can’t buy those either.
Tournaments are not for finding the best engine, we have rating lists for that (CCRL, CEGT and some more). Tournaments are for fun (engine teams meet and exchange things) and for the spectators.
Well this is mainly true. However I am pretty sure we will release an engine update for H13 at some point that will have the improvements we have made so far included. Those that subscribe to the Hiarcs book updates will get the main book we used in leiden http://www.hiarcs.com/chess-opening-book.htm
How do they know that no one cheats? They led them “run remotely”, but anyone could then play as a centaur. Rybka backed out over fears of this when Junior agreed to their 100000 dollar challenge, but then when losing to Zappa they both did it, when only 10000 dollars showed up. So there is a chance of it, and the organizers typically lack suitable skills, and would just declare some persons to be OK, and others they suspect and ban. There is a double standard.
Why have “operators” at all? Doesn’t the universal chess interface (UCI) make this obsolete? Seems like a meeting for computer geeks who care a little about chess rather than the other way around. Were they actually any spectators to count??
As was said, it is common knowledge that Houdini is the best now, so that some hypothetical special cloud Rybka runs well on 128 cores is just a rooting issue. Why play 9 games when the rating lists, or at least that aren’t controlled by the likes of Williamson and Kling, play 1000s and find Houdini to be the best on your home machine, without the influence of book-cookers?
The ICGA has made itself irrelevant by all of its insistence. It will cost at least 2000 to get to Japan, just to have an “operator” sit on their butt, then the large entry fee, unless like Pascutto you declare yourself an amateur even when you are selling your product, then to hire a book specialist (100 hours x 50 is 5000 dollars), to have on-demand hardware available remotely, all of this adds up to around 10000, so only those with spendy professional allowances can obtain the pompous WCCC title by now. The average Joe just wants to know who is best in their car, not the Formula One styles. I will just take Houdini or IvanHoe and have something better than the commercial guys can produce.