Beating Rybka3 engine in a 1CPU

[dr_zied_haddad]

Hi all,
Do you remember a few of my works improving the Shredder7.0 7.04 and 8.0 engines with new settings.
At that time, Shredder was the best engine ever, and it's still possible to improve it.
Nowadays, Rybka3 is a clear number1. All the other engines are just trying to lower the gap, but none could beat this outstanding engine.
Recently DeepShredder 12 was released. It'll be i think the 2nd best engine in 1CPU may be not the best in 4CPU or using 64 bit engines.
I'm now testing some new settings for Shredder 12, using a special testset database based essentially on the sicilian opening. I was surprised that Shredder 12 using the default setting was excellent when facing Rybka3 engines. I was even more surprised when one of my special settings beat Rybka 14-6 or 15,5 - 4,5(where the default setting holds the match but looses with a tight margin 9,5-10,5 or 8-12). It seems that this setting is the Rybka3 killer. It's less impressive vs Human and Dynamic settings.

I keep you informed and i'll publish a few of the games later

[dr_zied_haddad]

I forgot to say that the tournament was a 2' for the whole game, in a single CPU, played on my notebook (which in theory favours Rybka the strongest engine in the old hardware). No learning, positions from a selected testset from sicilian opening.
The same hash of 64Mo.

[Uri Blass]

I forgot to say that the tournament was a 2' for the whole game, in a single CPU, played on my notebook (which in theory favours Rybka the strongest engine in the old hardware). No learning, positions from a selected testset from sicilian opening.
The same hash of 64Mo.

I doubt if old hardware and 2 minutes for the all game favours rybka.

You can test if your fast time control help rybka by playing the same positions at slower time control like 20 minutes for all the game.

[Guenther]

I forgot to say that the tournament was a 2' for the whole game, in a single CPU, played on my notebook (which in theory favours Rybka the strongest engine in the old hardware). No learning, positions from a selected testset from sicilian opening.
The same hash of 64Mo.

How much memory is available on that notebook?
(I am asking because Rybka alone will use much more hash
than you had allocated, thus it could be that your results
are just a lottery with memory swapping all over)

Guenther

[Dayffd]

I remember and have those settings still - in the near future I I will be trying the 7.04 settings. One setting from another person, Columbus Egg, seems very good here. I would be very interested in seeing your new settings; over time I will be trying them at time controls of 30 minutes each engine per game. While I have many if not all of your setting for Shredder 7.04, I do not have any of them for Shredder 8. At least I don't think so although they may be on a floppy somewhere around here. I have to go through all of those soon.

Hi all,
Do you remember a few of my works improving the Shredder7.0 7.04 and 8.0 engines with new settings.
(snipped)

[dr_zied_haddad]

HI
I CAN ALLOCATE 512 Mo OF MEMORY FOR THE ENGINES.
INTERESTINGLY I'LL TRY TO MAKE THE SAME TESTS BY ALLOCATING THE MAXIMUM MEMORY FOR THE ENGINES

[dr_zied_haddad]

Happy to share my creations with real funs. You'll be among the first persons to know my settings . Still need a few tests for 1-3 weeks to be sure.