Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hash probe success rate

Author: Dann Corbit

Date: 19:45:43 06/30/04

Go up one level in this thread


On June 30, 2004 at 15:14:54, Robert Hyatt wrote:

>On June 30, 2004 at 14:36:21, Dann Corbit wrote:
>
>>On June 30, 2004 at 14:19:02, Andrew Wagner wrote:
>>
>>>I have a single transposition table, replace-always. I count a success when I
>>>get back a score which allows me to return immediately, without making a move.
>>>About what percent should my success rate be?
>>
>>about 1/3 (33%) seems usual, especially for long searches.  Sparse boards seem
>>to do the best for some reason.
>>
>>>Currently, my success rate starts around 5%, then drops turn-by-turn, until it's
>>>0, then jumps back up and starts falling again. Any suggestions what might be
>>>causing this? I've tested all my keys with generating them from scratch, and
>>>have no problems there. My code is modelled after the TT code on bruce morelan's
>>>site. Any suggestions appreciated....
>>
>>Your hashing has a bug for sure.
>>
>>Here are some Beowulf ratios of probes/hits (from the IQ3 test suite):
>
>I would assume these are "signature match" hits, not entries that terminate the
>search instantly, as he was talking about.  There is a big difference.  Crafty
>runs around 30% hits in middlegame when you count signature matches.  Far less
>when  you only count "useful entries"...

The hash has been found, but not qualified for usefulness yet.



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.