Author: Andrew Williams
Date: 05:33:36 06/26/04
Go up one level in this thread
On June 26, 2004 at 06:05:07, Laurens Winkelhagen wrote: >On June 25, 2004 at 19:27:37, Stuart Cracraft wrote: > >>On June 25, 2004 at 19:24:56, Andrew Williams wrote: >>>>move-ordering is poor. Do you measure first-move beta-cutoff percentage? This is >>>a nice measure, if only because most people generate it. It is the percentage of >>> nodes where a beta-cutoff is available and that beta-cutoff is discovered on >>>the first move tried in the node. Over 90% for this measure is what you're >>>looking for. >>> >>>Andrew >> >>Thanks -- never thought of that measurement. >> >>I've added to the todo. >>list > >Hi > >I implemented this (beta cut-off) measure also and I get values of 99% and >sometimes 98%. > >This seems to me to be a little to high: I have move ordering with pc-sq values, >lvvmva, history and hashmoves. I do not have that much confidence in my >moveordering, so I think the high value warns me of a possible bug in my >negascout algorithm. > >Is such a high value possible, or does anyone have an idea where the NS-bug >could be? (on the other hand, it could also be a bug in my implementation of the >measure) > My guess would be a bug in your measure. >Also I have another question: >I program in c++ and use an object Board B to represent my chessboard. This >object is a global variable. I know that other chess programs (e.g. beowulf) use >a structure *Board B which they pass to functions. > >Is either method faster than the other? (I'm a c-newbie;-) > Can't speak for speed as I'm no expert on C++, but one reason for using a pointer/reference is for doing a multi-processor implementation. Then each CPU needs its own board. Andrew
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.