[FOM] Checkers is a draw

joeshipman@aol.com joeshipman at aol.com
Sat Jul 28 13:14:20 EDT 2007


The context of my remark was establishing the power of takebacks. It is 
a form of handicap which allows me to get a very clear idea of exactly 
how strong a computer program is. My claim was that the same process 
would allow a strong grandmaster to, by interacting with a "perfect" 
player, form a justified conviction about the mathematical value of the 
game (a draw in the case of checkers, probably a draw in the case of 
chess if a perfect player is ever developed or encountered).

The reason I think that this is possible is that a strong grandmaster 
is probably within a few hundred rating points of "perfect", and I 
think the power of takebacks is at least this large based on my own 
experiences with very stong programs.

The metamathematical point I am ultimately making here is that the kind 
of certainty so attained ought to count as "knowledge" even though it 
is obtained neither by a formal mathematical proof, nor by an 
"interactive proof" of the kind recognized by computation theorists 
(Adi Shamir showed that PSPACE is the set of languages recognizable by 
this type of interactive proof).

This kind of certainty could only be obtained by someone of grandmaster 
strength, but exactly the same epistemological situation occurs for the 
game-theoretically easier question of whether Black wins the intial 
position of chess if you remove White's Queen Knight. Any reasonably 
good chessplayer KNOWS this to be true, but there is no hope of a 
mathematical proof.

If you are only a weak chessplayer, you still probably have enough 
experience to "know" that removing White's Queen from the initial 
position ensures a Black win (in this case, a mathematical proof is 
conceivable within a decade or two making reasonable assumptions about 
the development of computers, but the knowledge of chessplayers 
concerning this fact would not be made much more certain by such a 
proof).

-- JS


-----Original Message-----
From: Randall R Schulz <rschulz at sonic.net>


On Thursday 26 July 2007 08:30, joeshipman at aol.com wrote:
> I have a master's rating, and I own several Grandmaster-level chess
> engines. I rarely beat them when playing "straight", but I find it
> quite easy to beat any of them when I am allowed to use the "move
> takeback" option repeatedly.

Well, that's not chess as its rules define it, is it? It's chess with
time-travel into one's own past (while carrying back information from
the future, of course). It really tells us nothing about you or
these "engines," does it?


________________________________________________________________________
AOL now offers free email to everyone.  Find out more about what's free 
from AOL at AOL.com.


More information about the FOM mailing list