Author: Howard Exner
Date: 01:08:32 01/23/00
Go up one level in this thread
On January 22, 2000 at 09:15:57, David Blackman wrote: >On January 22, 2000 at 03:07:59, Howard Exner wrote: > >>How will computers be taught to adjust their playing style when facing weaker >>opponents? Two examples come to mind, one recent and one older. Recently >>the game of Fritz (rating: not yet established vs humans at 40/2) vs >>Hasidovski, A. (rating: 2177) saw Fritz cheerfully swap down into a drawn >>endgame. The older example that comes to mind is the Deep Thought vs Wchess >>draw. >> >>This I think is not just a Fritz or Deep Thought behavior but common to most or >>all programs. Chess games often follow quiet paths that make it difficult for a >>tactically stronger player to utilize that advantage. What can programmers do >>to compensate for their tactical wizards not knowing the opponent? Could having >>the program receive input as to the players rating be of use? They might then >>play a "second or third best move" but keep the position more alive with >>tactical possibilities. Or will computers always play what they evaluate as the >>"best move" regardless of opponent's strength? > >It's reasonably easy to persuade a program that it should go for tactical >complications and swindles, if that's what you want. A few programs have had >code for this. I've noticed this with Rebel Century. Some of the parameters called King Safety, Attractiveness and Attacking plus antiGM = strong make Rebel more alert to tactical possibilities. There does seem to be a trade off when setting these too high. I guess nothing comes for free, as other aspects of the game may suffer. > >It's somewhat harder to tell a program to head for a position that is quiet and >positional, but NOT drawish. Or to tell it to head for an ending that is roughly >even but difficult for both sides to play instead of another ending that is dead >drawn. > >To really make the most of an opponents weaknesses, you need those options as >well. I think they are possible, but tricky, and i don't think anyone has tried >very hard to do them yet. > >All of these probably rely on the operator adjusting the programs settings >before the game to reflect what the operator thinks the opponents weaknesses >are. Ideally you would like to tell the program "You're playing Gelfand >tomorrow" then you go to sleep while the program scans through it's database for >all Gelfand's games, analyses them for systematic weaknesses and decides what >openings to use and what settings for the engine. That's probably very >difficult, but i would guess that it could be done. Perhaps in the distant future programs may do this automatically, given a large database of players. That would be a bit unnerving knowing you are playing a machine that has a collection of your games and favorite openings. > >It's often said that you can play good chess just by playing good moves, and you >don't need to play to your opponents weakness. Almost all chess programs try to >play that way. Hardly any of the top human players do. Right, Lasker seemed to epitomize the risk taker by purposely playing moves that he knew would provoke his opponent. On the other end of the spectrum is Fischer, who many chess authors describe as one who "plays the board".
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.