home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!paladin.american.edu!europa.asd.contel.com!darwin.sura.net!spool.mu.edu!uwm.edu!rutgers!micro-heart-of-gold.mit.edu!xn.ll.mit.edu!ll.mit.edu!shoham
- From: shoham@ll.mit.edu (Daniel Shoham)
- Newsgroups: sci.math
- Subject: Re: Chess Problem
- Message-ID: <1992Sep15.033010.19777@ll.mit.edu>
- Date: 15 Sep 92 03:30:10 GMT
- References: <BuFpLp.9nI@ecf.toronto.edu> <1992Sep12.174545.23214@ima.isc.com>
- Sender: news@ll.mit.edu
- Organization: MIT Lincoln Laboratory
- Lines: 59
-
- In article <1992Sep12.174545.23214@ima.isc.com> karl@ima.isc.com (Karl Heuer) writes:
- >Here's a more advanced chessboard question.
- >
- >A white king, a black king, and a white pawn are randomly placed on an NxN
- >chessboard. (Independent, uniform distributions.) What is the limit, as
- >N -> infinity, of the probability that the position is a win for white?
- >
- >(Note that it doesn't matter what you do about collisions or illegal
- >positions, since they occur with probability zero in the limit.)
- >
- >Karl Heuer karl@ima.isc.com uunet!ima!karl
-
- We will assume optimal players on both sides (otherwise the question is
- meaningless).
-
- As stated, the probability is zero. A chess game is declared a draw
- after 50 moves without an irreversible one. Even if white manages to get a
- queen, the probability of finding the black king within 50 squares of an edge
- is -> 0.
-
- Suppose the 50-moves rule has been removed, and all that is needed is to queen
- the pawn.
-
- In a very large board, the likelihood of any of the three pieces starting
- anywhere near any of the edges or each other is -> 0 so we will ignore it.
-
- As I see it, white has a choice of two stratgies:
- 1. Run the pawn down and queen it without a king defense.
- 2. Move the king toward the pawn and defend it. Then advance the two in tandem.
-
- (strategies where the king and pawn move toward each other may save time, but
- they work just as well as strategy 2).
-
- We will use an x-y system to describe the position of each square. The last
- (queening) row is defined as y=0, whereas the first row is y=1. x, is similarly
- ranged between 0 and 1.
- (px,py) = pawn's starting position
- (bx,by) = black (king) starting position
- (wx,wy) = white (king) starting position
-
- px, py, bx, by, wx, wy are iidrv with uniform (0,1) distribution.
-
- Strategy 1 works if:
- py < by or py < |px-bx|
-
- Strategy 2 works if:
- d(p,w) < d(p,b) where d(i,j)=max(|ix-jx| , |iy-jy|) (i.e. chess-distance)
-
- We could now use brute force to integrate over all cases, or (if your religion
- permits it) do a Monte Carlo simulation.
-
- As my religion does not forbid me from doing a Monte Carlo after I have found
- an analytic algorithm (that I am too lazy to integrate), I did it:
-
- P(Draw) =~ 12.5% (40,000 runs, i.e. +/- .5%)
-
-
- Dan Shoham
- shoham@ll.mit.edu
-