Author: Larry Kaufman
Date: 20:25:51 11/04/98
Go up one level in this thread
On November 04, 1998 at 22:52:47, Jim Phillips wrote:
>Question: How much does positional knowledge add to a chess player's elo
> rating, compared to having no positional knowledge at all?
>For example: Chess player/program A does a full-width search 8 plies deep
> on every move, but looks only at material balance. Player/program B does
> the same, but also applies theoretically perfect positional knowledge.
> How much higher is B's rating?
>I have done a few "thought experiments" on this, and I've made the guess
> that positional knowledge is worth very roughly 500-600 elo points. Would
> anyone else care to give their estimates or opinions? Has anyone done a
> rigorous test to determine a value?
>-Jim Phillips
Actually, there have been experiments done on these lines. Hans Berliner
ran tests on "HiTech" vs. "LowTech" (HiTech with only minimal chess knowledge)
and I believe found a difference at 8 ply in the 200-300 range. However there
is a vast difference between minimal knowledge and none at all. With only
material values the program will play so ridiculously that the loss would be
greater than 500-600 points (assuming a GM level for the smart program); I would
guess 800 points or even more. However with only a simple centralization table,
the loss would be cut at least in half.
I.M. Larry Kaufman
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.