Author: Robert Hyatt
Date: 07:56:33 07/12/00
Go up one level in this thread
On July 12, 2000 at 10:27:13, Daniel Chancey wrote: >Computers should be more like Tigran Petrosian. He mastered the art of >prophylaxis and stopped any chance of attack his opponent might have seen. > >Castle2000 This is not easy to do. It is possible that you appear to have more pieces around your opponent's king than he has around yours. And you initiate an attack. And it turns out your pieces are not nearly as coordinated as your opponents, and you get smashed. Detecting the critical 'difference' in the above is non-trivial, and is just _one_ of the things that gives programs (and programmers) fits. Some of it you tune out via the book. Some you tune out via eval (ie don't castle into a strong attack, which many programs gleefully do). But computers are going to be subjected to attacks for _many_ years to come, because that is the hard part of the game to evaluate. It used to be that I would much rather see Crafty play a complex middlegame where it got attacked by a GM, rather than simplifying into an ending where the GM would rip it easily. This is no longer true. I see it play endgame after endgame very nicely, only to fall to a kingside attack I could see coming, but which it had no clue about. One day, of course, this won't be true. And many times the attack will fail when the opponent overlooks some subtle defensive move. But the machine ought not get into the complications, rather than getting into them and then hoping there is some escape hole... When that happens, the GM's days will be numbered...
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.