Rybka 232


Rybka 2.3.2

Back to forum

Thibault de Vassal    (2007-06-14)
Rybka 2.3.2

Just after President's cup "Ultimate computer challenge" finished, Vasik Rajlich threw a new Rybka 2.3.2 in the chess engines arena... (partly explains the $100,000 challenge to Ilyumzhinov a few weeks ago)

Once more, Rybka 2.3.2 seems stronger than her predecessors with indcredible results (blitz time control) against Hiarcs 11.1, Zap chess Zanzibar, Shredder and so on... Maybe +35 elo points compared to Rybka 2.3.1 !



Thibault de Vassal    (2007-06-14 18:20:51)
Rybka 2.3.2 improvements

An interesting thread (readme file) on improvements since Rybka 2.3.1


Rybka is undoubtly the strongest chess engine, but is it now the best partner to analyze correspondence chess games and the best anti-human engine !? .. Any opinions ?

Michael Aigner    (2007-06-15 12:53:31)
It depends

Hello Thibault In my opinion is Rybka very strong in the middle game as long as you got a position where a kingside attack is not possible - otherwise you should use some other engine. My experiance is that Rybka nedds some more plys to see the danger (or opportunity to attack), than other engines do. In Endgames it is also good but check always with Fritz or Shredder if you want to avoid ending in a dead drawn ending with an Rybka evaluation of +1 Best regards Michael

Thibault de Vassal    (2007-06-22 20:30:57)
Rybka's strength at CC time controls

An interesting poll, following a discussion at TCCMB :


Let's say Rybka playing alone, running 4 days per move (quite useless IMO) on a multi-processor computer, which rating would it/she achieve at FICGS ?

In my opinion, 2200 (with some peaks to 2300) would be great already... What do you think ?

Michael Aigner    (2007-06-22 21:27:50)
Re: Rybka's strength at CC time controls

I think you are right with your Elo approximation (2200 to 2300). There is just a little problem. In the case everybody would know he/she is playing against the latest Rybka version this would be a big problem for the program. In this case Rybka should loose almost every game because everybody would know Rybkas response to any move and could build wonderful traps. Did you hear about the so called "Planetenmatch" (match of the planets) where correspondence Chess GM Arno Nickel played against several Programms of unknown identity (the alias was the name of a planet). I think the engines had 24 hours per move and no chance - a desaster for the programs. That was of course before Rybka appeared, on the other hand he won also against Hydra!

Thibault de Vassal    (2007-06-22 22:35:34)
GM Arno Nickel vs. engines

I heard about this match (that happened on the Chessfriend server if I remember well)... but it seems that actually the engines won 3,5 - 2,5 ! .. A quite surprising result. Hydra was everything but designed to play at correspondence chess time controls, so let's wait for a more interesting experience, most probably with Rybka.

Nick Burrows    (2007-06-22 23:35:27)
engine strength..

Surely if different engines beat a correspondance GM at 24 hours per move, then their est. rating on 4 days per move should be higher than 2200-2300. Especially Rybka.

Glen D. Shields    (2007-06-23 00:14:24)
Chess Engine Strength

Thibault - I've been following the TCCMB discussion. I think it's impossible to answer the question what rating Rybka can achieve under the uncontrolled circumstances we play. If Rybka were playing only against humans, it would achieve a 2600+ rating. Since it plays mostly against itself and other top engines (with little human intervention), the typical results are win a few games, lose a few games and draw a lot.

Since tournaments are mostly set up so that players face opponents with similar ratings, a 2220 rated player using Rybka enters a tournament against other 2200 players. That player wins a few games, loses a few, draws a lot and leaves the tournament at approximately 2200. We conclude from that pattern that Rybka can achieve a 2200 rating.

Conversely, a player (like Uri Blass) who enters tournaments at 2600 and plays other 2600 rated opponents using Rybka wins a few games, loses a few games and draws a lot. He leaves the tournament rated approximately 2600. We conclude for that situation Rybka is rated 2600.

IMHO, it is impossible to answer the Rybka rating question under our typical tournament circumstances.

I think an even better question than worrying about Rybka's strength is "does anyone REALLY enjoy CC anymore?" Today's CC's is a race to buy the fastest hardware and make sure SSDF's top rated programs are installed. I'm playing beginners who can't explain what "en passant" is, but by parroting Rybka they compete in top tournaments and claim to hold titles that once upon a time had to be earned through hard work. After passing through the opening, it doesn't take much effort to figure out what program your opponent is using. At that point one can predict with high probablitlty every move your opponent will make for the rest of the game. Rarely do I see a move that I can can beat. The games are boring and pedictable. Those blunders and surprises that we once wrote funny stories about are long gone. IMO so is the fun.

Sorry to sound so "pessimistic," but until these problems are addressed and the fun is restored I find it just as easy to play against my computer. I can play at my pace, chose the engine I want to play, and unless my computer crashes I no longer have to worry about DMD :-)

Thanks for such a well run place to play chess. You do a great job maintaining it.

My best,


Thibault de Vassal    (2007-06-23 06:06:29)
Chess Engine Strength

Hello Glen !

I see your point, that's quite true and a consequence is what I called at TCCMB "the extensive nature of elo rating", however rating rules are more dynamic at FICGS.. So, let's say Rybka playing the FICGS championship against players of all kinds of ratings in the round-robin cycle... Anyway 2200 is only my feeling.

I understand your views about "rybka" [correspondence] chess nowadays, even if I don't agree with it completely. I saw some of your CC games played at IECG, and it looks much more like 'good old' chess with some unusual and beautiful tactical openings than typical 'correspondence computer chess' nowadays. I do believe there will be a place in the next CC years for more weird openings like bird, king's gambit, english... Also take a look at Peter Schuster and Wladyslav Krol games here !? .. Nothing boring with them, chess engine or not :)

Also advanced chess games with fast time controls could be quite interesting to watch in future as a way to see granmaster games with chess engines avoiding blunders 'only' (ok a bit more). We don't know exactly the human part in it, but draws won't be the rule for sure.

What is "boring" at correspondence chess (not new) is that achieving a top rating take a long... very long time ! .. But this is a great challenge yet IMO.

At last, thanks for you kind words :)

Best, Thibault

Thibault de Vassal    (2007-06-26 14:28:24)
Correspondence chess / Rybka

The Rybka's correspondence chess rating discussion moved from TCCMB to Rybka forum :)


Wolfgang Utesch    (2007-06-26 16:06:08)

The discussion of ratings is very problematic. Ratings on different sites are depending on different premises. What entry level was accepted? How long did you playing there how often? How much thinking time did you spent per move? Is the basic rating you earned over years to be caused by old tournaments with postcards (maybe without any help of engines and your opponents did it the same way)? How much care did you spent ratings (i.e. Norm tournaments?!)? Are you a member of the exclusive cycle of an organisation, getting invitations to closed high-level rating tournaments? Engines (also Rybka) are playing own styles and it depends on whether you can play better or worse against their special styles (knowing their potencies and weaknesses). Old fashioned players (independent from their ratings) will have much more problems to win or to hold draw against engines than players which have positioning themselves at actual situation. In my opinion today Rybka alone with one week thinking time per every move without any other help will reach a rating of about 2.400 at FICGS SM-tournament with an average rating of 2.450. In an ICCF anniversary tournament (average rating of about 2.600) same Rybka under same conditions will reach a rating about 2.550. I for myself wouldnt play longer correspondence chess, if I would have the feeling that any engine is playing better without my command. How long will it still take? My engine handling is not in this way, that I am waiting for longer times which move is offer by the engine. I have own ideas and Im trying their possibilities, investigating positions in depth over many moves in all directions. But sometimes engines have the better ideas and I have to accept this!

Thibault de Vassal    (2007-06-26 16:28:41)

Fully agreed ! .. By the way, what performance could achieve Rybka (mp) in a CLASS M or CLASS A tournament. Finally, in average about 2200-2300 IMO.