Author: Vincent Diepeveen
Date: 12:46:22 08/18/00
Go up one level in this thread
On August 18, 2000 at 15:27:57, Djordje Vidanovic wrote: >On August 18, 2000 at 13:59:03, Christophe Theron wrote: > >>On August 18, 2000 at 07:17:16, Vincent Diepeveen wrote: >> >>>On August 18, 2000 at 04:32:38, Joshua Lee wrote: >>> >>>>I am wondering what result's Diep get's on the Positional section of the LCT II >>>>test? >>> >>>I have not tried that test lately. Diep sure is a slow searcher. >>>So even if it solves them all, then it has to be said: other programs >>>are real quick in solving them, matter of tuning to known testsets. >> >> >> >>You really have nothing better to do than throwing mud to your fellows chess >>programmers? >> >> >> Christophe >> >> >> >> >>>I'll have a look for you this evening. > > >Christophe, > >I have a feeling that Vincent meant merely that programmers often use test >suites to fine-tune their programs. Nothing over and above that. That's how I >understood him... > Have a nice weekend, both of you :) The result is the same of course. Like that LCT position where Bb5 gets played. Some programs that don't know a thing about strong versus bad bishop solve it... :) Even worse is some aggressive tuned programs solve Nb1 in the bs2830 test set completely, despite that they don't have evaluation in their program to measure the trapnessed of the queen, as that's what is needed to solve this. Usually side effects in tuning have the same result as sovling it for the right reason. Diep has a function that measures whether a piece is trapped. How many programs have this also? How many programs solve Nb1 in the bs2830 test set quick? Note that by giving a bit extra points for attacking certain pawns also solves this position, that's how new versions of programs solve it nowadays. > >*** Djordje
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.