Author: Robert Hyatt
Date: 07:42:04 06/01/00
Go up one level in this thread
On June 01, 2000 at 06:58:55, Kai Skibbe wrote:
>Today an experimental fritz engine of Gromit3 reaches the following
>position against Junior6.
>
>[D]1r6/5B2/n2n4/PN4kp/4P3/8/1K6/R7 w - - 0 48
>
>Here Gromit plays the move Bxh5 with an evaluation of +0.48. Junior says
>-5,09 in favour of black. After Rxb5+ Kc3 Kxh5 white is totally lost.
>I remember that I have implemented the evaluation code for bad trades
>similar to crafty and so it isn´t surprising that Crafty also want to play
>Bxh5 with a slightly negative evaluation (-0,28). The code in Crafty is :
>
>if (WhiteMinors != BlackMinors) {
> if (WhiteMajors == BlackMajors) {
> if (WhiteMinors > BlackMinors) {
> if (WhiteMajors==1 && !TotalWhitePawns && TotalBlackPawns) {
> score=0;
> break;
>
>The last if-condition should be :
>
>if( WhiteMajors==1 && !TotalWhitePawns && TotalBlackPawns &&
> 1==(WhiteMinors-BlackMinors) )
>
>So the materialscore isn´t set to 0, if one side is more than one minor up.
>
>Now Gromit didn´t play Bxh5 and after Bxh5 Gromit says -4,45 in favour for
>black.
>
>Bob, am I missing something or is this a bug in the crafty code ?
>
>
>Best regards
>Kai
This is actually pretty amazing. I made the _same_ fix last night. Someone
was watching a game some clone (crafty clone) played on ICC and saw something
just like this. We spent an hour going over the game and I reached the same
conclusion... a piece down, opponent has no pawns, you do, you have a hard
time losing. More than a piece down, you probably lose. In the game we went
over, the score went way up when it managed to sac a piece for the opponent's
last pawn, but it left itself a bishop+night down, with both sides having a
rook. Trading rooks was a killer of course and it knew it... when it became
forced.
That was definitely a bug...
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.