Author: Gian-Carlo Pascutto
Date: 10:17:18 04/01/00
Go up one level in this thread
On April 01, 2000 at 11:18:10, Georg v. Zimmermann wrote: >It would be helpfull if the program I'm working on could detect whether a move >is hard to see for a human. Then the program would not sack a rook because it >would else get mated in 5 in a complicated way. It could also set up traps for >humans. This would make it stronger and more fun to play against (wouldn't you >rather want to pull of a m5 than win with an extra rook?) But how do you do >that. Here are the ideas I had : > >- Detect big score-changes on deeper plys = deep tactic a human might miss. Very risky...you have no way of knowing how 'natural' the combination is. And losing a rook is still better than getting mated... >- Count how many decent moves there are in one position. If there are only few >which are non recaptures, this could be hard to see for a human. (Maybe >impossible cause of alpha-beta pruning?) You're effectively trying to guide your opponent down an as narrow path as possible. This is such a natural thing to do that the fact nobody's doing it must mean its very difficult or expensive. But don't take my word for that ;) -- GCP
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.