Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How To Beat These Pesky Computers: Increase The Size Of Chess

Author: Uri Blass

Date: 06:34:28 02/25/06

Go up one level in this thread


On February 25, 2006 at 06:20:43, Tord Romstad wrote:

>On February 25, 2006 at 06:15:01, Graham Laight wrote:
>
>>Instead of playing on an 8x8 board, play on an 80x80 board, with each side
>>having 80 pawns, 20 rooks, 20 knights, 20 bishops, 10 queens and 10 kings - each
>>of which must be taken to win the game (I'll call this "super chess").
>>
>>For illustration, late me make some sweeping assumptions about chess: suppose
>>that each position has an average of 37 moves, and that a chess computer looks
>>at 2 billion (2*10^9) positions per move. In super chess, there would be an
>>average of well over 10*37 = 370 moves per position, because rooks, bishops and
>>queens would have more moves, and knights and other pieces would also have more
>>moves available on average, so lets say that the average number of moves would
>>be 1000 per position.
>>
>>In chess, the number of ply the computer can search comprehensively is:
>>
>>37^n = 2*10^9
>>log(37)*n = log(2*10^9)
>>n = log(2*10^9)/log(37)
>>n = 5.93
>>
>>We all know, of course, that extensions can reach a much deeper level than this.
>>
>>In super chess, the depth of the comprehensive search is:
>>
>>log(2*10^9)/log(1000) = 3.1 - which is not nearly enough to play well. The
>>extensions will be even more seriously impacted.
>>
>>So - each time the programmers get a bit uppity, all we have to do is challenge
>>them to a game of super chess!
>
>I fear you would be very disappointed by the results.  Computer programs
>would play this game very poorly, but so would humans.  You would leave
>your pieces hanging all the time.  It isn't easy to see that the knight you
>just moved can be captured by a bishop 50 squares away.
>
>It's better to change to a chess variant with a roughly normal board
>size, but which computer programs can't evaluate well.  Shogi (Japanese
>chess) would be a good choice.
>
>Tord

Note that increasing the size of the board to more than 8*8 help humans in go(I
suspect that go 8*8 can be even solved by computers).

I agree that 80*80 is too much but increasing the board to more than 8*8 can
help humans relative to computers.

If you want to increase the branching factor then I have other ideas.

For example give the players additional option to change the square of one of
their pieces in the board(with no capture) when they make a move instead of
playing a normal move when the piece that they changed it's square cannot move
in the next move and needs to wait one move before moving again.

For example
1.Nb1-g5 can be a legal move but it is probably not a good move because after h6
the knight cannot move in move 2.


Uri



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.