Author: Timothy J. Frohlick
Date: 09:08:35 12/09/05
Go up one level in this thread
Albert, I agree that Rybka is not playing typical computer chess. We are now in the Rybka Zone and I am feeling surreal. Salvador Dali PS Would you be willing to give Vas a little more $$ to fund his postgraduate education. I would. PS2 I am not affiliated with Dr. Rybka...oops....Mr. Rajlich. On December 09, 2005 at 09:55:46, Albert Silver wrote: >It hardly seemed possible that Rybka "slightly positional" could improve on the >score of the default "very positional" settings against Chess Tiger 15, having >scored 30.5-9.5. Sure enough, it couldn't, but nor did it do much worse, scoring >29-11, conceding 3 more draws. > > >1 Rybka "slightly positional" +21/-3/=16 72.50 29.0/40 >2 Chess Tiger 15.0 +3/-21/=16 27.50 11.0/40 > >I haven't checked the games other than the few I actually saw it play. I will >say this much, and it is something I must say I have never said about any engine >to date: I honestly get the feeling I could improve just by watching it play. >It's not perfect, and I won't even mention the well-documented endgame oddities, >but the piece maneuvers and placement, not to mention pawn play... Really >something else. > >Here is one move that oddly enough struck me not only as typical of Rybka, but >illustrative of its difference. > >[D]1k6/1p5R/2p5/P2b4/3P1K2/5P2/P7/8 w - - 0 55 > >White is completely won, so you'll wonder why I even bother showing it. The fact >is that it played 1.a6! bxa6 and then 2.a3. This doesn't make the position more >won by any means (if you know what I mean), but it shows the focus of the >engine. It chucked the pawn in order to guarantee the king remains cut off on >the 8th, and these 'little' things about piece placement and enemy restriction >are a big part of its success IMHO. I've seen a number of sequences where a key >behind a move it made was to prevent the opponent from doing something positive >such as improving piece activity or just placement. > >Here is another pearl that I found remarkable: > >[D]rn3rk1/2pb2q1/3p1b1p/pPpPppn1/2N4B/PQN2P2/4B1PP/1R3R1K w - - 0 22 > >Rybka played 1.b6! here, a move that may not seem so unusual in itself, and one >that no doubt other engines may find. What is striking is the sequence and >'idea' behind it. CT15 played 1...cxb6 sure enough, but contrary to expectations >(including my own), Rybka didn't take back on b6. Instead it played 2.Nb5!!, a >really striking move. It practically forces the exchange of Black's >light-squared bishop, after which White will stand a lot better. Obviously, >White will win back its pawn, but above all, it has removed a good defensive >piece (note the bishop on d7 was one of the only to successfully communicate >between both wings) and made it easier to penetrate. I'm not saying this wins >the game by any means, I'm just saying it is a strong move that is completely >atypical of computer programs. > >Match conditions were: > >Athlon64 Sempron 3400+ >Hash: 256mb; ponder off >Time Control: 10 mins + 2 secs >Nunn2 Openings set. > >The next opponent, chosen according to the CEGT results, will be Hiarcs 9 as it >scored far better than expected. > >The games are in a separate post. > >Albert
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.