Winboard Forum

While Rybka chocking the whole computer community beating both Shredder and Fritz, and many others with just a relative small Notes/pr. second, I think Rybka will change all ideas how to make chess engine with more knowledge rather than raw calculation, and that we can expect more joy watching a engine/engine matches for the future.

I can imagine how big headage Shredder and Frits team member has for the moment...they went the wrong way...time to start over...dam Rybka..

Its possible he counts nodes differently though. He might not count quiescence nodes. Assuming an 80% quiescence/total nodes ratio, I'm guessing its nps is really 5 times faster than it shows. Nevertheless, you are correct, Rybka is shocking.

My private opinon:I've tested both Rybka and the new Hiarcs 10. Rybka won the match... but I think that Hiarcs is stronger. Just try to compare KNOWLEDGE of Rybka and Hiarcs. Hiarcs knows almost everything about chess - and Rybka? - seems to be 'calculation machine'. I'm going to play long game match between those two engines - I'm sure that Hiarcs will break 50%.

My private opinon:I've tested both Rybka and the new Hiarcs 10. Rybka won the match... but I think that Hiarcs is stronger. Just try to compare KNOWLEDGE of Rybka and Hiarcs. Hiarcs knows almost everything about chess - and Rybka? - seems to be 'calculation machine'. I'm going to play long game match between those two engines - I'm sure that Hiarcs will break 50%.

Well, I have seen horrible results for Hiarcs 10 in longer matches
vs. Rybka (also longer tc) from reliable testers.
How do you define 'being stronger' BTW?
Moreover, how do you compare the knowledge both have, have you seen
the sources? I can tell you at least that Rybka has a lot middlegame
and general chess knowledge.
'Hiarcs knows almost everything about chess' is IMO a marketing
hype and a quite big exaggeration, which also leaves you unable to
prove it.
Will you still claim Hiarcs 'looks stronger' or 'must be stronger',
even if it loses overall on your machine?

Regards,
Guenther

Karol Majewski wrote:My private opinon:I've tested both Rybka and the new Hiarcs 10. Rybka won the match... but I think that Hiarcs is stronger. Just try to compare KNOWLEDGE of Rybka and Hiarcs. Hiarcs knows almost everything about chess - and Rybka? - seems to be 'calculation machine'. I'm going to play long game match between those two engines - I'm sure that Hiarcs will break 50%.

I did not test hiarcs but I read that hiarcs is better than rybka in tactical test suites and it contradicts your opinion that rybka is a calculation machine.

I guess that rybka is simply better in knowledge that you simply do not have when hiarcs is better in knowledge that you have and this is the reason for your impression that hiarcs has more knowledge.

I also believe rybka's nodes per second and it is very sad that the programmer is going to change definition of nodes to show amount of work that rybka does only because some stupid people complained about the fact that rybka search less nodes than other programs.

Note that I do not believe that rybka does not count nodes in the qsearch and my guess is that rybka simply find tactical things by evaluation without the need for qsearch.

I guess that rybka simply see forks and pins that wins material and other tactical ideas by evaluation and it is the main reason that the evaluation of it is expensive.

Uri