Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Nullmove crap

Author: Bruce Moreland

Date: 13:14:11 12/19/02

Go up one level in this thread


On December 19, 2002 at 07:51:33, Rolf Tueschen wrote:

>On December 18, 2002 at 23:08:13, Bruce Moreland wrote:
>
>>On December 18, 2002 at 22:40:20, Russell Reagan wrote:
>>
>>>On December 18, 2002 at 22:34:34, Robert Hyatt wrote:
>>>
>>>>Here is the point.  I will play you 100 games.  And in _each_ game, at a
>>>>place of my choice, I get to make two moves in a row.  I believe I can win that
>>>>match 100-0.
>>>
>>>Sounds like an interesting research project!
>>
>>It's not a research project, it's a chess variant.  White gets a pawn on e2, a
>>pawn on d2, and a king on e1.  Black gets the full army.
>>
>>White gets to make two moves in a row each time, and the white king may move
>>into check with the first move, but must leave check with the second move.
>>
>>I can usually beat someone with white a few times before they figure it out.
>>
>>It's pretty hard to mate white, it takes a lot of heavy material.
>>
>>But this doesn't have anything to do with null-move forward pruning.  All that
>>does is allow the program to say, "Am I still winning even if I pass my move?"
>>If the answer to that question is "yes", then there is probably no reason to
>>make a move, since the resulting position is almost always even *better*.
>>
>>bruce
>
>Bruce, you are wrong.
>
>Take a positional position, where a GM has exact knowledge that it is WON in the
>long run IF NOT the program makes such and such. And now the clue: it could well
>be that this is a position where your programs would ALL say I'm better or
>almost won. I see that you still haven't understood the weak sides of computer
>chess compared to human chess. Your long range chess sucks badly. Not that I
>would be happy with it. (But it is clear that the best GM are able to ALWAYS
>find a new trick to surprise the machine. It's ridiculous to assume that Kramnik
>suffered badly after ChessBase tweaked the allegedly "horrible" possibility to
>exchange Queens. Kramnik had hundreds of other weaknesses to exploit. But as I
>read there will be a re-match, and a re-rematch, so?)
>
>Rolf Tueschen

This is just a search technique that works.  It has nothing to do with how well
computers understand positional stuff.  Add this to a program and it *tends* to
prune out crap and keep good stuff, and general tactical performance increases.

If you take a normal program and add null-move forward pruning to it, it will
play better chess.

bruce



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.