It has an ELO of well over 3000!!!

Read about it here.

It was made by Chessbase, creator of Fritz, and Rybka.

OMFG chess is going to be completely solved soon, by the looks of it.

edit: It's out now!

It's true that rybka is phenomenal, and it has started a new era in computer chess, but this doesn't have anything to do with solving chess. Solving chess is more about stored memory of positions, like tablebases.

But it's already virtually unbeatable.

So is Kasparov on a good day. I don't think he has solved chess though. To solve chess you would need to prove that there is a forced result (win for white or black, or a draw) available from the first move no matter what the opponent plays. Although Rybka is strong a lot of that strength relative to other engines will depend on the hardware used. For instance, if I run Rybka on my wife's PC (1.5 GHz Athlon single core) I reckon that HIARCS 12 running on my galaxy modelling machine (Mac Pro with four 3 Ghz quad core Xeons) will thrash Rybka handily. However, If I was to run HIARCS on my internet machine (Mac with 400 Mhz G4) then Rybka would win every time. None of this indicates that HIARCS or Rybka has solved chess, just that one is much stronger than the other depending on hardware used.

There are two ways to solve chess. One involves producing a mathematical proof that demonstrates the win for one side or the other or shows that a draw will always result. The other involves calculating every possible move in every possible position and then showing that one side or the other can always choose a sequence of moves that leads to a particular result. One version requires N=NP (google it) and that doesn't look like being proved anytime soon. The other requires more time than the universe has already been in existence for so that won't happen anytime soon either.

There is another way to look at this actually. We can see that chess cannot be solved simply by looking at the reasons why a game can end and finding circumstances under which a game might go on forever. Under the current rules threefold repetition and 50 moves without a pawn move or capture both result in a draw if claimed. So there are circumstances under which there is no forced result and the game can last forever. That kind of game cannot be solved.

Sharukin wrote:

So is Kasparov on a good day. I don't think he has solved chess though. To solve chess you would need to prove that there is a forced result (win for white or black, or a draw) available from the first move no matter what the opponent plays. Although Rybka is strong a lot of that strength relative to other engines will depend on the hardware used. For instance, if I run Rybka on my wife's PC (1.5 GHz Athlon single core) I reckon that HIARCS 12 running on my galaxy modelling machine (Mac Pro with four 3 Ghz quad core Xeons) will thrash Rybka handily. However, If I was to run HIARCS on my internet machine (Mac with 400 Mhz G4) then Rybka would win every time. None of this indicates that HIARCS or Rybka has solved chess, just that one is much stronger than the other depending on hardware used.

There are two ways to solve chess. One involves producing a mathematical proof that demonstrates the win for one side or the other or shows that a draw will always result. The other involves calculating every possible move in every possible position and then showing that one side or the other can always choose a sequence of moves that leads to a particular result. One version requires N=NP (google it) and that doesn't look like being proved anytime soon. The other requires more time than the universe has already been in existence for so that won't happen anytime soon either.

There is another way to look at this actually. We can see that chess cannot be solved simply by looking at the reasons why a game can end and finding circumstances under which a game might go on forever. Under the current rules threefold repetition and 50 moves without a pawn move or capture both result in a draw if claimed. So there are circumstances under which there is no forced result and the game can last forever. That kind of game cannot be solved.

Sorry to say, but the first paragraph is essentially wrong, if not completely. Strength of engines are tested with the same level of hardware by independent testers, and claiming that "Rybka's strength depends on the hardware used" is not only ignorant, but disrespectful against the enormous amount of work by two International Grandmasters, Vasik Rajlich and Larry Kaufman on the evaluation and knowledge of rybka.

Here's a statistical proof: Rybka on single processor computer is stronger than Hiarcs on 4 processors.(not on one on one match, but by hundreds of games played against various engines).

http://www.husvankempen.de/nunn/40_40%20Rating%20List/40_40%20All%20Versions/rangliste.html

and on the same hardware, Rybka 3 is 258 elo points stronger than Hiarcs 12:

http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=5456

the strength of an engine is constant whatever hardware is used. What differs is the amount of time required to reach the same conclusion (ie depth of analysis)

eg my old 1.6Ghz machine needs overnight to reach 20 ply plus but when it gets there the result is the same as a multicore machine that can achieve the same result in hours or even minutes.

OK, play a match at set time controls on the hardware I suggested. Will the supposedly stronger engine running on much lesser hardware win? I think not. In fact, I know that this will not be the case since I have tried it. The Elo ratings of engines are calculated using equal hardware but the playing strength in practical terms is dependent on the hardware used. If this were not so there would be no need to ensure equality of hardware when establishing ratings.

You will notice I am not just suggesting a single processor against a four processor machine, effectively I am suggesting 16 64-bit processors running at twice the speed of the single 32-bit processor machine. If that has no effect there is no reason to choose a multiple processor version of any engine over a single processor version.

Sharukin wrote:

OK, play a match at set time controls on the hardware I suggested. Will the supposedly stronger engine running on much lesser hardware win? I think not.

yes if you set time limits then of course the fatser machine will win!!

all this proves is that the hardware is faster it does NOT prove that an engine is better on faster hardware. It only proves its results are obtained sooner.

Has anyone had experience with previous versions of Rybka?

I have, and it's simply terrific. I haven't seen any other engine (still today) that could give analysis of this high quality and this natural (non- "computerish") at the same time.