CSTAL on Steroids

[fern]

Due to that I use only red wine, not steroids.

Fernando

[George Tsavdaris]

I disagree with you and Uri as well,no one of you has seen the source code of CSTII so no one can make speculations regarding possible future improvements

Uri hasn't made any speculations about possible future of CSTal.....
How did you get that impression?

Read again carefuly and you'll see.....

Do you refer to what Uri said in this topic or another one?
Because in this topic i still don't see where Uri said anything about future development of CSTal.... Can you show me where he said anything about that?

Of course Uri is a respectable programmer and he can express his opinion,but I don't agree with him here

With what exactly statement he said you disagree?

[Uri Blass]

"If Cstal does not search deep enough it means that it does not know which lines to search...."

I am not so sure of the logic of your statement. At least in my human experience, you CAN know that a line is important and deserves to be searched, BUT not being capable of going really deep,you can reject it or consider it important BUT dangerous or bad.
And on the contrary, without knowing nothing about importance, not even having a hint about what "importance" means, you can pick p a good line just because going deep enough you finally see a material, tangible evidence of worthiness.
It is what many programs do, in fact.

My best
Fernando

You are right that for humans it is not correct but basically fpr computers not being able to search deep enough is because of one of two factors:
1)not searching fast enough
2)not being able to imagine clearly the position some plies after the board position.

2 is no problem for computers and 1 is not the reason for the fact that cstal is weaker.

cstal is a slow searcher but based on my knowledge the gap between cstal and other programs today is so big that even searching equal number of nodes is not going to help cstal to be a top program.


cstal seems to be at least 400 elo weaker than programs like Toga and it means that even making it 32 times faster is not going to help cstal to beat toga and I doubt if cstal is really more than 32 times slower than toga in nodes per second.

From the chess war ratings:

http://loirechecs.chez-alice.fr/chesswa ... 011CLs.htm

CHESS SYSTEM TAL II 2214 ENG

http://loirechecs.chez-alice.fr/chesswa ... 011ALs.htm

TOGA II 1.3x4 2760 GER

Uri

[emerson4301972]

CSTal plays sacrifices that are sometimes unsound sometimes sound, against computers unsound sacrifices dont work, but it will confuse humans. For one thing, when it sacrifices against humans, humans dont know whether its sound or not.

CSTal is good for generating ideas and is a great contribution to Computer chess. Hopefully there would be a multi processor version to compensate for its slow search

[fern]

That is the key of Cstal attraction: he play US, that even if being GM not always can calculate 18 ply to reject a sacrifice.
Cstal is great

Fernando

[fern]

Probably, Uri, you are right concerning comp vs comp. Concerning us, a vastly faster Cstal would give us even lot more fun.
My best
fernando

[mclane]

in CCC, they talk highly of CSTal, how other programs now have evaluations in its style and what a important thing it was for CC

10 years late. As always with the new paradigm. They all agree one day. Usually banning the original guy first or killing him or whatever.

First they ignore you, then they laugh at you, then they fight you, then you win

Gandhi said that

[ArmyBridge]

Hi friends:
I love CST it is a great program, I never seen before something like this, it produces great games , Thorsten dou you think that we will see CST MP??. Also I have interest in the difference between the way that evaluate programs like Junior that also very aggressive is compared with the evaluation that makes CST I like to know your opinion

[mclane]

Thorsten dou you think that we will see CST MP??.

thank you for your comments.

i am working on the continuation of the program.
one has to see that version3 was nearly finished when the company who made it was sold.

it never came to a release of Version 3 of CSTAL. this version 3 has a STYLE menue (as the DOS version) and it has a multi-engine concept.

the bitboard engine is faster than the old 2.03 version. but IMO it would be better to invest some more time to program the latest development in search progress made in the last 8 years.

You can believe me that i am working on this.
lets keep the fingers crossed.

[ArmyBridge]

It sound great Thorsten , i just can not wait to have the new CST , I´m sure that you are do a good job, we are hungry to see wild and amazing games
Regards