Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: another MTD drawback - EXECUTIVE SUMMARY

Author: Pat King

Date: 13:27:38 09/11/02

Go up one level in this thread


Because MTD always uses min window searches, it
     a) requires an upper and lower bound to be stored in the hash table, or
     b) risks "bouncing over" the true score.
     c) takes too much time using "fail hard" search and creeping up on correct
        value (WARNING (C) IS A PERSONAL OPINION INTERPRETING CERTAIN PERSONS'
        OBSERVATIONS IN WAYS THAT FAVOR MY OPINIONS).

I have used (a) with my engine, but but I have my own problems, and my results
cannot be considered conclusive of anything. I have done comparisons of MTD and
PVS, with nearly equal results, but they used the same hash structure storing
upper and lower bounds. It seems to me, given all the preceding argument, the
right comparison would be MTD storing both bounds, and PVS storing one bound,
each using about the same physical memory, with PVS storing less information
about more nodes. Anybody done this?

Pat King



This page took 0.08 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.