Author: Jay Scott
Date: 14:36:47 02/20/04
Go up one level in this thread
I wrote: >Full-width searches slow down exponentially with depth. I interpret the time >control finding as meaning that the human "algorithm" may have asymptotic >behavior better than exponential. I want to amend that. It's misleading. Taken literally, the asymptotic cost of any reasonable chess algorithm is constant, because in the limit you solve the game. :-) But it still makes sense to try to plot "chess strength" (on some absolute scale) versus time per move. In practice, interesting positions cannot be solved, and we can pretend that any curve we get goes on to infinity. It's clear to me that the shape of this curve is different for humans and existing programs. If judged in the computer science way, "How well can you do in the limit?", then the judgment is: Given a long enough time control (if the relation observed so far continues to hold), then humans play better. That is a powerful motivation for chess programmers to study human chess play. Jay
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.