Author: Tom Kerrigan
Date: 08:28:59 03/17/00
Go up one level in this thread
On March 17, 2000 at 02:43:24, pete wrote: >On March 17, 2000 at 02:18:14, Pete Galati wrote: > >>On March 17, 2000 at 01:05:34, Vincent Vega wrote: >> >>>Sorry, math error. >> >>I never understood the interest in solving the game. In theory, I guess you >>could have a computer that can solve the game. But when that's done, and you >>set the 2 greatest Chess players down to play a match against eachother, they'll >>still be playing at human capabilities, and it'll still be a case of whoever >>makes the second to last mistake wins. >> >>So I'm sure if the interest in computer Chess continues that the game will >>eventually get solved, but I don't think that that's very interesting, and >>probably it doesn't need to be done all that much. >> >>Refinements in the programs is what's needed, not the ability to solve the game. >> >>Pete > >I suspect it is even worse ; when the game is solved you maybe will have won >much less than expected . > >Let's assume the game is drawn after all white's first moves and by most black >answers to them ( which seems to be very likely ). > >This will continue throughout the whole game : the move which wins a pawn but by >some miracle allows the opponent to escape to a draw is just as good as the move >which blunders a knight and by some miracle let's _you_ escape to draw . > >So what you get is "only" a search tree which has been limitted very much ( nice >enough ) . Choosing between the moves you still have to build a strong engine ; >I think I remember to have read similar things about draughts and the program >chinook ( name might be slightly different ) . > >But now this perfect program will probably still allow a human which for example >achieves to swap pieces escape to a draw from time to time . > >Now imagine a tournament where several top humans and this perfect program play >each other round-robin ; it might well be that the machine even then wouldn't >win . > >It seems we already have this problem now ; when two programs which use 5-men >tablebases play against each other the draw ratio seems to go up ( anyone ever >checked this statistically ? ) . This is a very good point. If you download a program that plays Connect 4 perfectly, you will see exactly the situation you describe. It is not hard for a human to beat a program that plays perfectly (assuming, of course, that the human gets to move first). And most of the moves played by the program are pretty random. -Tom
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.