Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The Scalable Search Test / Results of "Der Bringer"

Author: Gerrit Reubold

Date: 16:20:15 06/18/00

Go up one level in this thread


On June 18, 2000 at 15:12:43, Gian-Carlo Pascutto wrote:

>On June 18, 2000 at 14:55:43, pete wrote:
>
>>Looks pretty impressive to me ; unfortunately you will have to wait for Gerrit
>>Reubold to exlain the interesting part of your question :-)
>
>I'm interested for sure. The results you posted beat everything else
>hands down.
>
>The maximum effective branching factor (measured by time) is as low as
>5 AND it spots the mate quickly.
>
>--
>GCP


Hi all,

unfortunately I am unable to reproduce Pete's results with the current version
of Bringer. I don't know which version he used (I sent him many betas), his
hashable settings etc. These are my results (PII-300 14 MB Hash):

0:00:00.2  ( 4/12)       2691  -1.19  c1-b1  a6-a4  d3-e4  g8-f8
0:00:00.9  ( 6/16)      55650  -1.26  c1-b1  b7-b5  d3-e4  g8-f8  c4xb5  a6xb5
0:00:02.0  ( 7/21)     138751  -1.39  c1-b1  b7-b5  c4xb5  b8xb5  e3-e4  a8-b8
0:00:03.7  ( 7/22)     267661  -0.88  h3xh5
0:00:05.7  ( 8/26)     381065  -1.03  h3xh5  g6xh5  h1xh5  g8-f8  h5xh7  f8-e8
d3-f5  f7-f6
0:00:08.6  ( 9/26)     663786  -0.63  h3xh5
0:00:11.6  ( 9/29)     960843  -0.23  h3xh5
0:00:18.1  ( 9/29)    1572723   0.38  h3xh5  g6xh5  h1xh5  g8-f8  d3xh7  f8-e8
h7-f5  e8-d8  h5-h8  d8-c7  h8xb8  a8xb8  f5xf7
0:00:24.8  (10/36)    2335509   0.78  h3xh5
0:00:31.3  (10/36)    2942408   1.18  h3xh5
0:01:10.4  (10/36)    7566834   12.80  h3xh5  g8-f8
0:01:21.5  (11/36)    8816266   13.20  h3xh5
0:01:29.9  (11/36)    9654231   13.60  h3xh5
0:01:42.0  (11/36)   11323902   Matt in 11  h3xh5  a6xa2  h5xh7  e7-e5

Bringer is lucky with the extensions. Just the right extension and not too many
unnecessary extensions. I guess this is more luck than good tuning, the next
position can easily show a bad performance of Bringer.

Here are some statistics which were collected during the search, which looked at
11.414.517 nodes:

Fail high 1st move:   80 %  <<< move ordering not so good here, Bob Hyatt's
percentage of strongly ordered trees is 90%
Moves per fail high:  1.86

Extensions
 Check:      252678  <<< the well know check extension
 Recapture:  13542   <<< recapture extension, also standard
 PawnPush:   8019    <<< pushing passers to 6th or 7th rank
 SingleMove: 46035   <<< extend when player has only one move
 Threat:     5180    <<< extend when the opponent has a severe threat
 Razor:      8189    <<< a negative extension, some pruning, razoring
 Perpetual:  24230   <<< extension to detect perpetual checks, hurts the
performance in this position
 Endgame:    9963    <<< these extensions are probably very unnecessary in this
position

Null Search
 Null Searches: 764285
 Success rate:  79 %
 Node rate:     43 %


You can download Bringer at www.reubold.onlinehome.de
Bringer displays these search statistics, you can do experiments and get a
feeling for the positions in which the extensions do / do not work.

Greetings,
Gerrit




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.