Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: More doubts with gandalf

Author: Dan Newman

Date: 16:41:41 02/27/01

Go up one level in this thread


On February 26, 2001 at 12:21:00, Peter Berger wrote:

>On February 26, 2001 at 08:09:24, Frank Phillips wrote:
>
><snip>
>
>>Knowledge presumably takes cpu cycles to process, so faster machines help?
>>
>>If we had 32 man EGTBs, there would be absolute knowledge, no search and no
>>chess rule of thumb knowledge of the type discussed.
>>
>>Presumably chess knowledge just encapsulates guiding principles for those
>>position, which if we had enough searching power (or EGTB) we could prove were
>>won, lost or drawn.
>>
>
>It seems to be a very common belief that with 32men EGTBs chess is solved ; I
>haven't been able to understand this point of view so far .
>
>Availlable data and also intuition suggest that chess is a draw .
>
>It might be the case that every white first move draws ( at least that's my
>belief ) .
>
>I even believe that every black answer to every move of White leads to a draw .
>
>Now let's take this "perfect" program ( " ... absolute knowledge, no search and
>no chess rule of thumb knowledge of the type discussed ..." )and create a little
>challenge :
>
>This program , let's call it "Perfect" has to play against a group of strong
>Super GM players ( like in a typical Linares tournament ) .
>
>It randomly chooses one of the "perfect" moves availlable . How will it perform
>? It won't lose a single game : that's obvious . But how many games will it win
>? I suspect it will draw quite a few ; the move winning a piece but by some
>miracle allowing the opponent to escape will have the same probability than the
>move blundering a piece and letting "Perfect" escape in the last moment .
>
>As control group for "Perfect" I choose Kasparov who has to play against the
>same gang of masters ; I think it is very likely that he will perfom better and
>gain more points than "Perfect" that suddenly won't look that perfect anymore .
>
>In fact this effect can be clearly observed today already IMHO ; take a critical
>drawn tablebase position and play it on the weaker side against a typical
>chessprogram that uses tablebases ; you will often have a much easier time
>getting a draw ( as for example in KRP-KR it will simply sack the pawn ( a
>"perfect" move )) compaired to playing against it with tablebases disabled . I
>have seen _many_ games where the tablebase program let the opponent escape in a
>much easier draw .
>
>So without something like a "swindle" mode I am not sure that it is beneficial
>to even use the TBs if you are on the stronger side in a drawn tablebase
>position .
>
>pete

Of course 32-man tablebases will never be built.  The amount of matter in the
earth is many orders of magnitude too small to store them, and the amount
of energy required to calculate them is similarly much too large for this
ever to be practical :).  (This is assuming we don't have some amazing and
unlikely breakthrough...)

But (even though we might never know) I suspect that white wins with "perfect"
play.  I have a couple reasons for thinking this: 1) white has a slight edge
over black as evidenced by actual games played, and 2) white gets to choose
from one of 20 opening moves, and if any one of them is a win, white wins.

Of course it's entirely possible that all white's opening moves give rise to
draws and losses only or to all losses or all wins or all draws or even no
draws; no one knows.

Even if perfect play is a draw it's hard for me to imagine that any human or
machine could play against a perfect player and maintain the draw.  Just one
"blunder" and you're lost.  It seems to me that there will be at least a
few positions in any game where intuition or positional sense won't be a
perfect enough guide to avoid such a blunder--the position being too complex
or the result of the blunder too deep.

I suspect if the game is a draw overall, that the drawn games are extremely
long, giving the opponent of a perfect player an exremely difficult job to
do...

-Dan.



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.