Author: Shep
Date: 04:37:57 07/13/99
Go up one level in this thread
On July 13, 1999 at 04:57:53, blass uri wrote:
>
>On July 13, 1999 at 03:55:43, Shep wrote:
>
><snipped>
>
>>Besides, don't forget these are _manually played_ games! It gets really boring
>>for me to watch them play on when both I and the programs know that nothing new
>>is going to happen.
>
>I understand your decision but you guess that nothing is going to happen and do
>not know it and I am sure that in part of the cases this guess is wrong(I guess
>more than 10% and maybe more than 20%).
Of course their is an uncertainty. But I rather produce some "wrong" results
than have a match overthrown by some rare and obscure mistake from one of the
programs.
>
I saw a case in the ssdf games when chessmaster6000 lost against Fritz5.32 in a
>dead draw position(KRPPP vs KRPPP on the same side of the board).
>
>The reason was not a bug in chessmaster6000 but wrong evaluation
>[CM6000 liked the king in the centre and Fritz went with the pawn to h3 and
>after it won the h2 pawn(chessmaster got 2 pawns for it but these were
>unimportant pawns)]
My philosophy is a little different. We all want chess programs to play like
GMs. Then it is my philosophy to treat their games as if they were GM games.
That means, if both programs see no way to make progress, it will be declared a
draw.
(For this, I have two criteria:
a) the mentioned "draw rule" with 3 successive "+0,00" from both sides
b) the more fuzzy "draw adjudication rule":
"if both programs just shuffle their pieces around without moving a
pawn, giving check or capturing for at least 15 moves, the game is
a draw"
where the "15 moves" may sometimes be 20 or only 10, depending on my
humble assessment of the position)
We've all seen the "GM style draws" of 20 moves or less.
No-one would force a GM to play on just to see if he will make a wrong
assessment somewhere in the endgame.
Neither is the purpose of these tournaments to discover evaluation holes in the
participating programs.
Shredder 3 vs. CSTal is a good example. I was about to adjudicate this a draw at
move 40 because both showed around "+0,20" from move 30 on and the endgame
looked drawn to me. Then CSTal's eval suddenly rose to +0,80, so I played on,
and CST went on to lose the game by some endgame blunder.
I suppose no GMs would have continued after move 40.
Also take for example the "bug" related to tablebase use recently mentioned
here:
if you don't have all promotion TB's, programs like Hiarcs may not be able to
mate even though they can say "mate in 18".
How would you treat this? Is it the program's fault that it can't mate?
Is it your fault because you didn't set up the system correctly?
Is it a draw? Or a win? Hmmm...
Or the infamous "mate with KNB vs. K" bug in old Fritz etc.
---
Shep
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.