Author: Pham Minh Tri
Date: 21:14:26 05/11/02
Go up one level in this thread
On May 11, 2002 at 14:43:44, Nolan Denson wrote: >Let me first say that the idea's below are a friend thats helping me understand >chess programming. (Any help is greatly appreciated) > >When generating captures is it best to assign a value to the move, >the difference between the captured move minus the >capturing move (i.e NxP => 100-300 = -200, or do >100-3=97). Also is it ok or best to first try all the moves that have a >= 0 >score and then I calculate the swap value more >accurately via the SEE function. I heard its possible to do this >straight away but it's too slow, but are there any advantages?? According to Bob, the move order helped by SEE may speed up 10%, but use of SEE also slows down 10%, so the gain = 0. However, SEE will help much in qsearch (cut 1/2 of q nodes). >Once that is done is it ok to do the following in order ... if not what are some >improvements that can be done?? > You should try hash move first, then the captures of last moved piece. >- try all the captures that SEE says >= 0 >- search killers, if legal >- generate non capturing moves and search them in >order by its history heuristic score >- search captures whose SEE score is < 0 > >In each step one move may generate a beta cutoff, thus >avoiding the work we would do in the others
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.