Rybka 2.3.2 is about 50 Elo stronger....

[fern]

Great news. At last I could face an opponent of my stature........

Tal Send his Regards
fernando

[Dann Corbit]

Hi Uri,
Thanks for the info....
But note Uri that most probably the Baron will participate with it's non public version which is MP as far as I know

Yes but the same is also for rybka
It is not going to participate with 2.2 and it is going to participate with more than 2 cpu.

I assumed that there was no big difference between the improvement in rybka and the baron.

The only problem of the baron is that the time control of the event is faster than one move per week.

Uri

The solution is simple then, find out how many CPUs Rybka is using and then repeatedly multiply that value by 2 until you calculate that you have a time advantage of one week per move.

Of course you will have to factor in SMP loss. Maybe ASCII RED is available.

[Harvey Williamson]

....at least this was announced by Vasik in his forum and in the playchess server at blitz time control

http://rybkaforum.net/cgi-bin/rybkaforu ... id=1003#fp

Wow,unbelievable indeed

On rybka forum it says v previous versions at blitz. We will wait to see the evidence. New Hiarcs here at Leiden is of course 300 elo stronger and working up an appetite for Fish. 300 must also be true just because we say it!

[M ANSARI]

I think people should interpret what Vas said more accurately. He said that 2.3.2 is more than 50 ELO stronger than 2.3.1 in head to head competition. You have to realize that 2.3.1 had a few bugs that probably made it play around 30 ELO weaker. So this 50 ELO point gain might actually only be 20 or 30 ELO against all engines. Ofcourse it might be more ... but let us not get ahead of the ball. Actual strength assessment will come after it has been tested by CCRL and CEGT.

But it is an exciting upgrade. 2.3.2 plays pawn endgames like a Super GM. It knows how to push for an advantage and how lethal pawns can be. 2.3.1 on many occasions would blindly go into an endgame with Bishop vs. pawns thinking it is winning ... while in actuality it was totally losing. This has been addressed and is showing a lot of promise. The difference in evaluation in some positions is dramatic ... so this is a big upgrade.

Ofcourse Rybka will be favourite in Lieden, but others are prepared as well. Anything can happen, and there is always the factor of a nasty book line that someone has cooked up while burning the midnight oil. No doubt about it that Rybka is by far the strongest engine ... but the strongest engine doesn't always win. Also Zap has great scaling and might close the gap if they use 16 cores or more. So it should be a very interesting week indeed.

[Uri Blass]

I think people should interpret what Vas said more accurately. He said that 2.3.2 is more than 50 ELO stronger than 2.3.1 in head to head competition. You have to realize that 2.3.1 had a few bugs that probably made it play around 30 ELO weaker. So this 50 ELO point gain might actually only be 20 or 30 ELO against all engines. Ofcourse it might be more ... but let us not get ahead of the ball. Actual strength assessment will come after it has been tested by CCRL and CEGT.

But it is an exciting upgrade. 2.3.2 plays pawn endgames like a Super GM. It knows how to push for an advantage and how lethal pawns can be. 2.3.1 on many occasions would blindly go into an endgame with Bishop vs. pawns thinking it is winning ... while in actuality it was totally losing. This has been addressed and is showing a lot of promise. The difference in evaluation in some positions is dramatic ... so this is a big upgrade.

Ofcourse Rybka will be favourite in Lieden, but others are prepared as well. Anything can happen, and there is always the factor of a nasty book line that someone has cooked up while burning the midnight oil. No doubt about it that Rybka is by far the strongest engine ... but the strongest engine doesn't always win. Also Zap has great scaling and might close the gap if they use 16 cores or more. So it should be a very interesting week indeed.

No
I think that the 30 points improvement against other engines does not make sense.

Rybka2.3.1 beat previous version of rybka in head to head cometitions but did not get significantly higher rating because the bugs caused her to have problems against weaker engines.

Fixing the bugs probably can help her mainly against weaker engines and not against previous rybka's so if rybka2.3.2 is at least 50 elo better than rybka2.3.1 based on head to head competitions then I expect her to earn even more rating points when it play against Zap,Hiarcs,Naum and other engines with high CCRL rating (It does not make sense to test her against engines that are more than 300 elo weaker)

Uri

[ArmyBridge]

Hi ANSARI
Well, It seems that your you know much about Rybka, if your you were a programmer that going to face Rybka, what kind of strategy will you adopt to fight Rybka?, will you select some kind of opening's?, I know ,Rybka play very well almost all type of opening , but Stefan Mayer once prepared to Shredder to play Urusov gambit to face Rybka, would it work once again, prepares a good repertoire of wild openings to try to fight the positional knowledge of Rybka?, I would like to know your opinion

[M ANSARI]

Yes Rybka could possibly be stronger than 50 ELO points ... but just in case testing does not come out that way ... people should understand that Vas mentioned the 50 ELO figure in a head to head match of 2.3.1 vs 2.3.2


As for beating Rybka, to be honest I don't know how I would prepare for that. If it was Rybka 2.3.1 I would say try to get it in an endgame where it would go for a Bishop against 3 pawns. This is extremely difficult proposal and is not possible now as this has been fixed. Rybka is a really formidable opponent ... I mean it is a real monster. While trying to get new lines for my book I decided to let Rybka play Zap Zanzibar on my Quad running 64bit OS and 3.2Ghz. Zap is no push over ... but from the 100 games I put it to play against Rybka ... Zap only managed 3 wins vs 41 losses and rest were draws. So as you can see it is not easy. Best hope is to wish that one of those 3 wins occurs in the match. Even Fruit, Shredder, Fritz and Hiarcs can get wins from Rybka if they play enough games ... it is just that those wins are very very difficult to come by. Add to this fact that Rybka will most likely be booked to the hilt, things look tough.

[PauloSoare]

But it is an exciting upgrade. 2.3.2 plays pawn endgames like a Super GM. It knows how to push for an advantage and how lethal pawns can be. 2.3.1 on many occasions would blindly go into an endgame with Bishop vs. pawns thinking it is winning ... while in actuality it was totally losing. This has been addressed and is showing a lot of promise. The difference in evaluation in some positions is dramatic ... so this is a big upgrade.

The Endgames! If 2.3.2 has more knowledge in endgames, I goes to be still more happy.

Paulo Soares

[ArmyBridge]

ouch!!! Zanzibar get only 3 wins??? that's what I call to give a good beating to somebody poor Zap, personally I love that program, wow .... something or somebody must appear and offer a good battle to Rybka.. I Think that we needed that chess scene going to be more balance.

[Milton]

I think people should interpret what Vas said more accurately. He said that 2.3.2 is more than 50 ELO stronger than 2.3.1 in head to head competition. You have to realize that 2.3.1 had a few bugs that probably made it play around 30 ELO weaker. So this 50 ELO point gain might actually only be 20 or 30 ELO against all engines. Ofcourse it might be more ... but let us not get ahead of the ball. Actual strength assessment will come after it has been tested by CCRL and CEGT.

But it is an exciting upgrade. 2.3.2 plays pawn endgames like a Super GM. It knows how to push for an advantage and how lethal pawns can be. 2.3.1 on many occasions would blindly go into an endgame with Bishop vs. pawns thinking it is winning ... while in actuality it was totally losing. This has been addressed and is showing a lot of promise. The difference in evaluation in some positions is dramatic ... so this is a big upgrade.

Ofcourse Rybka will be favourite in Lieden, but others are prepared as well. Anything can happen, and there is always the factor of a nasty book line that someone has cooked up while burning the midnight oil. No doubt about it that Rybka is by far the strongest engine ... but the strongest engine doesn't always win. Also Zap has great scaling and might close the gap if they use 16 cores or more. So it should be a very interesting week indeed.

Thanks for information about the improvement in Rybka's endgame. Out of curiousity, are you a beta tester; or did you get this information from Rybka's forum or from some other source?

Regards,
Milton