Author: Robert Hyatt
Date: 05:08:38 02/02/99
Go up one level in this thread
On February 02, 1999 at 07:38:16, Bas Hamstra wrote: >Of course I know it's not secret :) > >Bob I saw you referrring to the new experimental code somewhere. I am very >curious what your cheapo version of SE looks like. Does it work? How much does >it cost? Do you score better in suites? Do you win more games? > >If it's to early to say - first impressions? > > >Bas Hamstra. No real impression yet. This is code I experimented with about 3-4 years ago, early in the crafty development. I saved it because I didn't like the overhead back then, when I was a whole lot slower. I fixed the code to match the new SMP code in crafty, and sent it to Mike Byrne (running fitter on ICC). He has continued to tweak around with it, and has reported some good things on some types of positions (ie the mate in 6 in WAC 141 has always been a hard position for crafty. It now finds the right move in about 15 seconds on my quad xeon, but mike has reported something 5-10 times faster on his dual PII/400. I'll certainly post results as we test this... the 'first impression' is that you will lose about a ply of basic search depth, but for that you get a lot more tactical acuity. It doesn't work so well in endgames, last time I tried, because it is _easy_ to over-extend since there are so many singular lines that advance passed pawns... part of what Mike is doing is finding out where/when/how-much to limit them..
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.