Author: Robert Hyatt
Date: 18:17:40 01/11/02
Go up one level in this thread
On January 11, 2002 at 19:22:18, Russell Reagan wrote: >Good idea or bad idea? > >My idea is to, on my opponent's turn, have my program create several threads for >probable moves that my program thinks my opponent might make and have several >searches running in different threads on my opponent's time. Then when my >opponent makes his move, my program will kill off the threads searching moves my >opponent did not make, and will be left with one thread searching on the current >position. > >My thinking is that instead of having to start the search from scratch and try >to get back to where you were through transposition table entries, you save time >by continuing the search on the current position. > >Before I haul off and change the structure of my program to support this >feature, is it even worth it? Am I going to get that much of a search depth >increase or advantage by using this method instead of searching probable moves >by my opponent individually on my opponent's time? > >Thanks in advance for your comments. > >Russell Just think about it for a minute. Suppose you have 30 possible moves. If you start 30 threads, after 60 seconds each would have searched for 2 seconds. That isn't going to buy you much. If you search just one move, and are right only once in every 10 moves, (which is very low prediction rate) you will come out better than your idea, because once in every 10 moves you will use the 60 full seconds on the right move. In your algorithm, for every 10 moves, you only use 20 seconds on the right move(s)...
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.