Author: Ed Murak
Date: 08:40:04 12/13/05
Go up one level in this thread
On December 13, 2005 at 03:50:00, Thorsten Czub wrote: >On December 13, 2005 at 02:51:35, Ed Murak wrote: > >>Actually, I see the opposite. RYBKA (latest) has more problems versus Shredder >>9 than against another "obvious" leader, Fritz 9. >i can explain you the things you see: > >when rybka is intelligent, it will kill the stupid programs (f9), :-) >and it will not so heavily kill the less stupid programs (s9) >and will have the most problems with the more knowledged programs (hiarcs10). > >Of course sometimes this is not working. e.g. if you play against a program that >compensates knowledge with deep search (togaII 1.1 and gambit fruit 4bx are good >examples for THIS). Not by my tests. >>Rybka (beta4, 32 bit) v. Fritz 9 >> +8 =3 -1 > > >yes. rbyka kills fritz because fritz is not really clever. When I play against F9, myself being a patzer of course, my only problem but a big one is tactical lufts. If F9 gets a strategic "idea", it seems inevitably to be a wrong one. :) Yes, on the whole I agree with you, the Fritzy's have been stupid. F9 less so. More relevant knowledge in it. As GHz go up, this has to be the way forward. Not just searching more wrong moves super-deep. >>That is 79.1%, with a TP of +231 ELO, even without endgame tweaks and EGTB for >>RYBKA. Basically F9 isn't making it into the EG with both legs attached too >>often. :) > > :-)) Sometimes F9's whole body is missing by the 1st time-control, and it is spinning all unco-ordinated with its king performing a dance. :) >>Performance vs Fruit 2.2.1 is almost as high, >75% which I believe could have >>been thought impossible. > >test togaII 1.1 and this new gambit fruit and you will see that it is more >difficults. its IMO time for Fabien to come out with a new Fruit, because the >"old" 2.2.1 is not anymore the best in the FRUIT-family. > >>But this (Fritz 9 and Fruit 2.2.1 results against RYBKA b4) runs a little >>contrary to Chrilly's (possible) conclusion of special RYBKA v. Shredder tuning, >>maybe. Unless when Shredder comes back round the loop for a match, RYBKA scores >>90% or something! :) > >i don't think chrilly meant that rybka is a killer engine to s9. I judge by what Chrilly _wrote_, not what he _meant_. >it is true: rybka is in the moment one of the strongest and has lots of >potential and is an interesting program. It is great fun to play, and even fun to watch vs other engines. Best ever in that regard. IMO. Strategically, it seems to pick (one of the) correct ideas almost all the time. The losses I see for it (against engines; my own human results are private) are generally to king-assaults (NPS issues) and of course because there is _zero_ endgame knowledge in these early releases and no EGTB access. >i am really interested to see how HIARCS10 will play against it. Likewise. I hope H10 is delayed if necessary, to make sure that it too is not swallowed by the little fish and it, H10, addresses these issues. There is poison inside RYBKA that is _new_, if anything in this cc world is "new".... that can be sensed even by weak players by watching (not just accumulated results). For me the experience is qualitative, quantitative I leave to the crowds to do their ++-+=-+.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.