Author: Vincent Diepeveen
Date: 14:37:08 12/27/00
Go up one level in this thread
On December 27, 2000 at 13:55:09, Ricardo Gibert wrote: >On December 27, 2000 at 13:32:33, Gian-Carlo Pascutto wrote: > >>On December 27, 2000 at 13:07:05, David Rasmussen wrote: >> >>>But still. My program which is fairly strong, searches 13-15 plies (I don't >>>remember) in 5 or 10 seconds on my computer, which is far from enough to see >>>that line. How can one see that line with a normal chess program in one second? >> >>Extend heavily on checking moves and moves where there are few replies >>to check. Extend if one side has mating threats. > >Extending when there are few legal responses is sufficient. The apparent depth >is something of an illusion since the defenders replies are forced. This should >be easy for CNS. It is indeed, but the extension problem is that you also take a look deeply at from opening: 1.e4,h5 2.Qxh5 3.Rxh5 However, CNS weak problem is the fact that the search tree is completely dependant upon your evaluation. If your evaluation is happy for something you search that area real deep. If it doesn't care and all values are about the same then it is not taking a look at the entire trees at all! Despite the many problems, which for new algorithms there always will be, CNS is a very interesting algorithm and is getting researched a lot by Ulf Lorenz who made a quite good impression with PConners. Regrettably we hardly see output from his program on testsets, so i have no idea how it tactically is doing on the tactical positions like bs2830. For sure it solves the above published mate. >> >>Note that this won't necessarily make your program stronger. >> >>-- >>GCP
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.