rybka
Moderator: Ras
rybka
can anyone speculate what rybka does right that everyone else seems to get wrong? im trying to write a world championship chess program and wonder if anyone here can shed some light in what it takes to rise to the top. rybka did it very fast and thats the way i want to do it . is it more about search or is the eval the most important ? why is the gap between rybka 3 and rybka 2 so big , is that normal to increase over 150 elo a year?? im trying to create a learning based program anyone here have any good ideas for learning in chess ? i did so test on crafty learning and i was unable to make it learn enough to beat rybka without a oppening book . in theory i thought it will work if i reversed colors after each game . so as of now i have not found a chess program with a learning maodual that makes enough diff.
-
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: rybka
My personal guess is some sort of clever search enhancement. Positional knowledge will not take you as far, as fast, as a really good search. Reductions come to mind in that when Shredder started using them it passed everyone and stayed there for a couple of years. Ditto for null move when it came out in the late 80's. What he has done I don't know. Someone will figure it out, they always do, whether by studying Rybka or by being creative and re-discovering the improvement independently. And then we wait for the next step, which will surely come as well.
Re: rybka
Rybka forum is here :
http://rybkaforum.net/cgi-bin/rybkaforum/forum_show.pl
There are enough useless topics about it here...
http://rybkaforum.net/cgi-bin/rybkaforum/forum_show.pl
There are enough useless topics about it here...