Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rebel 10C - Bench Test

Author: Ed Schröder

Date: 10:55:27 02/28/99

Go up one level in this thread


>Subject: Rebel 10C - Bench Test

>Posted by m.d.hurd on February 28, 1999 at 09:25:22:

>It appears that when the bench test is run it now no longer solves number 1 and
>3 giving a score of 2600 and giving overall score of 2596 which is only based on
>the solving of number 2. Has anyone else seen this at all ?

Ignore the "bench test" in version 10c. It will produce incompatible results
with the original Rebel10 and 10b chess engine.

The bench test is based on Rebel10's function "combination=on" that loads
a special tactical search only valid for positions and its first computed move
(see the manual). This has all changed in Rebel10.0c as now "combination=
on" is also valid for the whole game. The new "combination=on" is much more
careful with extensions in order to get a reasonable depth while playing normal
games.

Note that the bench test isn't about solving positions. The 3 positions were
carefully chosen at the time to produce an accurate bench number which
is valid for the Rebel10 chess engine. Now that we have version 10c
(which is totally different) it would have been better to remove the bench
test or rewrite the thing.

Ed Schroder



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.