Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: King, rook pawn and wrong bishop endgames

Author: Dieter Buerssner

Date: 09:32:32 10/19/00

Go up one level in this thread


On October 19, 2000 at 11:10:53, Uri Blass wrote:

>On October 19, 2000 at 09:18:45, Dieter Buerssner wrote:
>
>>On October 18, 2000 at 14:14:20, Uri Blass wrote:
>>
>>>I can give another rule:
>>>Blacks draw if it is black to move and the distance between the black king and
>>>the corner in king moves is smaller than the ditance between the corner and the
>>>white pieces except the bishop(white king in king moves and white pawn in pawn
>>>moves)
>>>
>>>If it is white to move then use distance-1 instead of distance for white.

>>Rarely, your rule predicted a draw, when the position is lost.

And this was wrong, sorry. I did not implement the rule exactly as written
above. By an oversight, I rather used smaller or equal instead of smaller,
because this was more or less similar to what I allready had.

>No my rule does not say that it is a draw because the distance of the king to
>the corner is equal and not smaller then the distance of the pawn.
>
>I guess that in most of the cases when the distance is equal it is a draw but my
>rule was made to be always correct when it says draw.

Also this is correct, when I did not make another stupid mistake. I tested
all KBPK positions with an h pawn and the wrong bishop. There were 639768
different positions (a value pretty close to an estimate of 64*55*32*6).
Of those, there are 233366 draws due to TBs. Your rule did not predict
any position as draw, that is lost for black. I also supplemented your rule
by the rule given by Ricardo and I think Bruce, namely that when the black
king is on g/h 7/8, it is a draw. This gives a slightly better success rate.

With this, the rule returns "may be lost" in 114475 cases, that are draw.

Just out of curiosity, I also made the same test with smaller or equal.
Then exactly 5049 are predicted wrongly as draws. "May be lost" is returned
in 66599 cases, that are draws.

>It is possible to do a rule that is 100% correct when it says win when the other
>positions can be divided to probably draw and probably win for white but I did
>not do it.

I wouldn't mind to hear from your next rule ;)

Thanks again,
Dieter



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.