Author: Robert Hyatt
Date: 07:18:22 01/13/04
Go up one level in this thread
On January 12, 2004 at 19:37:51, scott farrell wrote: >On January 12, 2004 at 12:04:52, Anthony Cozzie wrote: > >>On January 12, 2004 at 08:19:58, Tord Romstad wrote: >> >>>On January 12, 2004 at 07:59:23, Anthony Cozzie wrote: >>> >>>>On January 12, 2004 at 05:37:48, Tord Romstad wrote: >>>> >>>>>FEN, please? >>>>> >>>>>Tord >>>> >>>>[D]r6r/k4pp1/pq2p2p/1p1n4/3N3P/1R6/PPP1QPP1/2KR4 w - - 25 >>> >>>Thanks! >>> >>>>One of the reasons I like this position is that Crafty needs some minutes to >>>>come up with the solution. Gothmog should have no trouble: this position is >>>>very sensitive to checks in qsearch. If you do recapture extensions you will >>>>find it even earlier. . . . >>> >>>Actually, Gothmog doesn't find it that easily. It needs longer time (but >>>much fewer nodes) than Zappa to find Nxb5+. The right move is found after >>>10 plies, 13 seconds and 2.4 million nodes on a PIV 2.4 GHz. >>> >>>I don't do checks everywhere in the qsearch, but only at nodes where the >>>engine guesses that it is necessary. This is of course not an easy guess >>>to make, and it doesn't work perfectly. The main factors influencing the >>>decision are the material left on the board, the king safety evaluation, >>>the tactical complexity of the position, the static eval for the node, >>>and the upper and lower bounds for the root score. >>> >>>Recapture extensions are not used at all in Gothmog. I've tried them >>>many times, but they never worked for me. >>> >>>Tord >> >>Zappa is not a tactical monster. Back when I had 1-rep and recap, it was pretty >>decent, but I'm more or less following Fritz/Crafty in my approach: limit the >>extensions, get a lot of brute force depth, and outsearch in the middlegame. >> I'm not quite sure what the above means, since Crafty _has_ the recapture and one-reply extensions enabled all the time... > >I've recently come to a similar conclusion. > >I currently ripped a few extensions out, and reduced some to half plies, and >implementation of SE (my own version), lead to greater full width searches, but >with still plenty of more controlled SE extensions, to I ended up with the best >of everything, similar on test positions, much much better in games. > >>anthony
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.