Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Knowledge

Author: Daniel Shawul

Date: 01:21:07 09/27/04

Go up one level in this thread


On September 27, 2004 at 03:10:49, Uri Blass wrote:

>On September 27, 2004 at 02:34:11, Daniel Shawul wrote:
>
>>On September 27, 2004 at 02:17:24, Uri Blass wrote:
>>
>>>On September 27, 2004 at 01:58:56, Daniel Shawul wrote:
>>>
>>>>On September 26, 2004 at 00:17:07, Sergei S. Markoff wrote:
>>>>
>>>>>There are a lot of positions with total 1 pawn on the board and equal piece sets
>>>>>of both sides. If weaker side have 2 or more minors or (majors&&minors) then the
>>>>>position is drawn in most cases. You must also use king position analysis to
>>>>>improve this knowledge.
>>>>>
>>>>>The same to 1 pawn vs 2 pawns with no passers.
>>>>>
>>>>>This knowledge not help a lot, but after 2 weeks I make a version of ST which is
>>>>>a little stronger than w/o this knowledge.
>>>>>
>>>>>I need your results to compare.
>>>>
>>>>same thing here.Mine also is better without the RPKRK evaluator.
>>>
>>>Sergei did not talk about KRPKR evaluator so I do not understand what you mean
>>>by same thing.
>>
>>      ok i see Rook < 2 minors. Anyway i think what he said can also be applied
>>for rook pawn ending too.
>>
>>   >
>>>>After seeing my engine loose agaisnt Movie (whit i think doesn't have this fn) ,
>>>>I disable the evalator and just scale down the normal score by 1/2.
>>>>Ditto for queen pawn ending, and opposite color bishop  ending.
>>>>
>>>>daniel
>>>
>>>I have no special evaluator for KRP vs KR
>>>
>>>I do similiar thing to dividing the score by 2 when there is a single pawn in
>>>the board and when both side have equal material except the pawn I simply reduce
>>>the score of the better side.
>>>
>>>Here is the relevant code from movei
>>>score is the score from white point of view.
>>>
>>>if (numpawns[LIGHT]+numpawns[DARK]==1)
>>>			{
>>>				if (valpieces[LIGHT]==valpieces[DARK])
>>>				{
>>>					if (numpawns[LIGHT]==1)
>>>						score-=60;
>>>					else
>>>						score+=60;
>>>				}
>>>
>>>Uri
>>     But you don't give a draw (0) score here. And what sergie is talking
>>about is giving a 0 score. I did and it lost against movie_178 on rook pawn
>>ending which doesn't.
>>daniel
>
>Of course
>
>I do not give draw scores unless it is clear that it is a draw or I evaluate the
>chances as equal.
>
>Uri

   ofcourse i don't give 0 score for endings where one side has two pieces.
Buf when only one piece exists,it is possible to recongize dead draw positions.
I do  for some rook pawn and queen pawn endings. For all other cases i reduce
whatever score I get from my normal eval.




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.