Author: Frank E. Oldham
Date: 13:30:17 04/14/98
Go up one level in this thread
On April 14, 1998 at 13:03:17, Mats Winther wrote: (with some elisions made) >Look at a Petrosian game where he makes a positional sacrifice (a rook >against a knight). Petrosian wins this game against a strong >grandmaster. >Would Fritz5 make the same move? Probably not. Fritz5 will cut off that >whole variation tree. He doesn't consider it much since Fritz5 has a >materialist view of chess (and must have since he is a computer). >But actually, that positional sacrifice may be the best move in that >position. > >There is a materialist side of chess, but there is also a positional >side of chess. These sides complement each other. There is a side of >chess that has nothing to do with computational force. > >Probably it would be very hard to create a chess engine that does not >cut off that variation tree but instead takes the positional sacrifice >into consideration. It's hard for a computer to realize it's good since >Petrosian doesn't regain the material later in the variation tree. >His play has something else is mind. > >Some sophisticated positional qualities cannot be reached by >computational >force. > >The ideas above is my credo in computer chess. But I'm afraid I'm >repeating myself. Since my view does awake so much negative emotion I >have decided that this is my final message in this forum. > >Mats Winther Mats, please keep posting -- these topics are worthy of discussion. If we use crafty as an example (since we have the source), it would actually be fairly easy to get it to make Petrosian's positional sacrifice. It just takes some (possibly adaptive or learned) modifications to the evaluation function. The real problem would be to do this without weakening play in other positions! A major task still in computer chess is "discovering Petrosian's *evaluation function* -- and Kasparov's, and Fischer's (and maybe even CSTal's !). Frank
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.