Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Search Stability - ways to improve?

Author: Robert Hyatt

Date: 15:19:25 07/02/04

Go up one level in this thread


On July 02, 2004 at 04:14:50, Steve Maughan wrote:

>Bob,
>
>Thanks for the reply,
>
>>If you are talking about PVS, where you search on a null-window and fail high,
>>then re-search on a normal window and fail low, the fix is easy.  If you fail
>>low, _ignore_ the fail high, keep the old best move, and keep searching.
>
>Yes this is what I currently do.  The situation that I'm concerned about is when
>I have a tactical position where, for sake of argument, the key move is Bxg6+.
>Monarch fails high on Bxg6+ at ply 7 only to fail low on the research.  The same
>is repeated at plys 8 and 9 until it find the solution at ply 10.  Clearly
>Monarch is "seeing" something at ply 7 but is not able to resolve it until ply
>10 - this seems a bit odd - probably a bug of some sorts.
>
>Having make the original post I'm thinking that the bug will most likely reside
>in the hashing code.  That's the first place I'll look.
>
>Thanks again,
>
>Steve

This is _not_ unusual.  you store bounds and scores.  Bounds can cause fail
highs, but then will not help produce a true score so you can fail low.  It is
an artifact of inexact draft match hashing.  IE using an entry from a deeper
search to make a shallower search more accurate...





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.