Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Chesspartner 5.3 (or ChessTiger 2004?): inconsistent header claims

Author: Lex Loep

Date: 11:53:18 12/02/04

Go up one level in this thread


On December 02, 2004 at 02:04:56, Robert Allgeuer wrote:

>On December 01, 2004 at 14:45:05, Lex Loep wrote:
>
>>On December 01, 2004 at 08:20:20, Robert Allgeuer wrote:
>>
>>>Lex,
>>>please note also the other problem I came across concerning mate claims of ERT,
>>>posted a few messages below.
>>>
>>
>>ERT relies on the information it receives from the engines, it only does some
>>sanity checks such as is the king in check.
>>
>>When using winboard/uci engines its important the winboard adapter is properly
>>configured.
>>
>>Lex
>>>Robert
>
>If I were you I would check under which conditions ERT outputs the peculiar
>string "Checkmate !!".

This string gets output when the absolute value of the engines returned score is
32767.
Problem is that in the winboard protocol the value for mate is not strictly
defined and not all engines are returning equal scores.
For native engines the score for mate is clearly defined.

>When you look at my examples you can immediately see that it has NOTHING to do
>with the engine, because it occurs with many different ones. It also is not an
>issue with the UCI/WB adapter because the problem appears with native, WB and
>UCI engines.

See my comment above.
By enabling the logfile you can see the scores that the engine returns.
ERT tries to do validate the score that the engine return, e.g. when the engine
reports checkmate, ERT does not simple checks validate this claim.
The same for score threshold, both engines must be in agreement before the game
is declared a win for one the two.

>
>I am a bit amazed how apparently reluctant you are to fix problems. There is
>also the old problem that Rebel claims mate one move to early which has not been
>fixed, yet.

I suggest you compare the scores returned by the Rebel engine to some other
winboard engines.
Look at the score when its about the get checkmate and also at the scores when
giving checkmate.

Of course its possible to fix it by makings some special cases for each engine,
but yes you are right I am a bit reluctant to fix this.

Lex

PS. If you find something strange, like the header score problem I be interested
to fix this but I need a log file. Especially for problems that are rare.

>
>Robert



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.