Author: Vasik Rajlich
Date: 12:16:29 12/09/05
Go up one level in this thread
On December 08, 2005 at 08:29:26, Albert Silver wrote: >On December 08, 2005 at 05:48:39, Vasik Rajlich wrote: > >>On December 08, 2005 at 01:05:57, Albert Silver wrote: >> >>>I can't recall Deep Fritz 8 ever looking in such bad shape. It lost even winning >>>attacks. Here is one position where I was certain Rybka was done for: >>> >>>[D]5k1r/pb3q2/1p2p1Rp/3nPp1Q/2r2P2/P5P1/1P2N3/4KB1R w K - 0 32 >>> >>>yet 20 some moves later the position looked like this: >>> >>>[D]8/p7/1p2p1k1/1P2Pp2/P2r1PnP/5Q1K/R5B1/2q5 b - - 0 56 >>> >>>I think Rybka is a poor name. Houdini seems more appropriate. >>> >>>Bear in mind that this was played using Rybka's "slightly positional" profile. >>>My purpose is to test the other as well under identical conditions in order to >>>better see how they differ in terms of performance and style. That, and the fact >>>Rybka doesn't have its own book, is why I choose to play with the Nunn2 set. >>>CEGT reportes that Rybka suffered a marginal loss to Hiarcs 9 as well as a >>>serious one to Chess Tiger 15, so Tiger will be the next opponent in the series. >>>Then I'll see what the other styles show. >>> >>>Athlon64 Sempron 3400+ >>>Hash: 256 Mb; ponder off >>>Time Control: 10min + 2 sec >>>Nunn2 Openings set >>> >>>1 Rybka "SP" Beta 32-bit +24/-3/=13 76.25 30.5/40 >>>2 Deep Fritz 8 +3/-24/=13 23.75 9.5/40 >>> >>>The first match against the default "very positional" style yielded >>> >>>1 Rybka "VP" Beta 32-bit +18/-10/=12 60.00 24.0/40 >>>2 Deep Fritz 8 +10/-18/=12 40.00 16.0/40 >>> >>>Bear in mind, one match means nothing so no conclusions other than it is a >>>different and interesting style should be inferred. >>> >> >>I just want to make a comment about the positional vs tactical settings. These >>slightly affect the way search is handled. > >Could you elaborate a bit? > > Albert > Tactical settings will spend relatively more time on "forcing" moves. These include captures, checks, and certain threatening moves. Also, none of the settings are especially extreme - I could believe that any of the four might be the best. It would also be possible to make a special testsuite-crunching version, ie. an ultra-tactical setting. Going too far in the other extreme wouldn't really make much sense - it's hard to play good positional moves when you don't see any tactics :) Vas >>They will not change the general feel >>of the program for a human observer. Only different evaluation terms can do >>this. >> >>Vas
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.