Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: History Heuristic

Author: Pat King

Date: 14:26:08 03/17/04

Go up one level in this thread


On March 16, 2004 at 14:57:13, Robert Hyatt wrote:

>On March 16, 2004 at 11:12:34, Renze Steenhuisen wrote:
>
>>On March 16, 2004 at 09:51:00, Pat King wrote:
>>>"All moves searched" = fail low = move ordering doesn't matter here. I think you
>>>ought only store fail-highs in the history table. It will lose its effectiveness
>>>more slowly that way.
>>
>>Have you had any success with this way of applying the HH or is it just a
>>suggestion? It doesn't seem to work here, or I am not counting my move right!
>>
>>Cheers!
>
>
>The only time you should update the history counter is when you either fail
>high, or you back up a value that is > alpha.  IE at PV and CUT nodes, but
>_never_ at an ALL node...
>
>Any other way will fail.

Bob has it right, I was slightly too restrictive in the way I stated it. A
fail-high or an exact score yields useful information that you want to remember,
but a fail-low says "I've searched everything, and nothing works." So no move is
better than another. No point to updating the history table, nor to saving a
"best move" in the hash table at this point.

Pat



This page took 0.01 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.