Author: Robert Hyatt
Date: 06:03:45 07/27/00
Go up one level in this thread
On July 26, 2000 at 23:26:44, Larry Griffiths wrote: >On July 26, 2000 at 22:29:07, Robert Hyatt wrote: > >>On July 25, 2000 at 22:04:28, Larry Griffiths wrote: >> >>>On July 24, 2000 at 21:28:49, Larry Griffiths wrote: >>> >>>My original Bruteforce for producing valid moves counts ran in 121 seconds. >>>I implemented extracting all the captures at the next ply and searching >>>thru them to find a king. I return if I find a king and adjust the move counts >>>at the previous ply so that they will be correct. My counts match just like my >>>original did... Except that it took 404 seconds! Is this the method that you >>>use in Craft's perft command Bob? The only thing I can think to do is use my >>>old method at the leaves of the tree... >>> >> >> >>I am getting confused here. Are we talking about an alpha/beta search, or >>about 'perft' type testing? Let's define that before I answer more. I don't >>try to optimize my perft results, I only use those for sanity checking the >>move generator and board update code after major changes... >> > > >My BruteForce is another name for your 'perft'. I am writing another chess >program from scratch and want to check the sanity of my move generation and >board update code after major changes... :) >> OK.. here is probably what you are doing wrong. You can use the "I captured your king" to mark previous ply moves as illegal. But what about at the _last_ ply? There is is more efficient to check for "in check" as opposed to driving the search one ply deeper but not counting the moves. You are probably doing about 40 times the work that you would do with the incheck function. What I was explaining really works for the search itself... because you _always_ generate moves after making one, if only to find out that there are no more captures to try. And that is enough to detect that the previous move was illegal...
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.