Author: Vincent Diepeveen
Date: 17:31:00 12/24/00
Go up one level in this thread
On December 23, 2000 at 11:34:42, walter irvin wrote: >On December 22, 2000 at 13:39:27, Uri Blass wrote: > >>On December 22, 2000 at 11:49:22, Tania Devora wrote: >> >>> >>>How many positions or moves exist in chess? >>> >>> >>>I mean all the possible moves!, >>>good moves, stupid moves like. 1.h4 1.a5. 2.Th3 2.a4 Ta3 3.Ta6 3.Tg3 4.Ta6 Th3 >>>etc etc........ >>> >>>I think that the number will be verrryyyy HUUGEE!! I think that any machine or >>>computer in the future could see all the possibilities in chess. >>> >>>What do you think? Anybody have an idea? of how many possibilities exist in the >>>chess game? >> >> >>The number of possible games including stupid moves is very big and no computer >>in the future will calculate it but you do not need to calculate all the >>possible games in order to solve chess. >> >>For example you do not need to calculate both 1.e4 c5 2.c4 and 1.c4 c5 2.e4 >>inspite of the fact that the games are different. >> >>The more interesting question for solving chess is the number of legal >>positions. >>The number of legal positions is clearly smaller and there is less than 10^50 >>legal positions(I guess there are about 10^40 positions(1 with 40 0's)). >> >>I do not expect a computer in the next 100 years to have memory of 10^50 or even >>10^40 positions but it may be possible to solve chess with less positions >>because it is possible to prune illogical moves and get less positions. >> >>An example to illogical moves that it is possible to prune is 1.e4 Nf6 2.Qh5 >> >>I am sure that 2.Qh5 is losing and chess programs have no problem to win with >>black so programs do not need to have knowledge about the right move after 1.e4 >>Nf6 2.Qh5 and search is enough. >> >>The number of possible positions when you prune moves that are illogical may be >>smaller and computers may solve chess in the future but the number is still too >>big for the computers of today and I do not expect computers to solve chess in >>the next 50 years. > >> >>Uri > >you may be correct that no computer will solve chess in next 50 years , but you >could be wrong .i agree with everything you said in your post , i just want to >take it 1 step further .there could be more prunning .a strong program will win >a very large % of games where it is up + 2.5 .so that could be used as a bench >mark .also there is a large number of legel positions that while they are close >in score they could not be safely reached vs strong opposition .so if you >eliminated duplicated moves,moves with scores over 2.5 ,positions that can not >be safely reached .you might actually have a number that even todays computers >have a shot at . >walter irvin First of all the number of moves there are is quite limited it's like something near 1900. Secondly the number of total chesspositions might seem huge, aroudn 10^40 but who knows, perhaps only 10^20 of those pos are useful positions! The real problem is that if we get a machine with 10^20 memory, how are we going to tell the computer which positions are crap and which aren't? So basically we need to rely on branching factor, but a search must not be confused with the total number of positions possible! A search simply is not busy seeing each position once! It is searching for lines you can play, and you can create an infinite number of games. Nevertheless we must take into account that chess is a finite game, one day it will be solvable! That means white probably wins somehow, especially seeing how some top GMs nowadays basically get a real good position with white usually says empirically a lot, whereas with all kind of draughts variations we see that a draw is to be expected (so chess IMHO more interesting as draughts as all those draughts games are always boring draws!. The last word on how many positions we need to search versus how big our database needs to be is not going to be said real soon, but my own calculations i expect that considering lineair progress in software speedups because of increased hardware, that in 2066 we'll face some real interesting truths about what playing perfect chess brings for kind of result. Win for white or a draw? I bet on a win for white still but majority is betting on a draw!
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.