Author: Marcos Christensen
Date: 06:54:15 07/22/00
Go up one level in this thread
On July 22, 2000 at 07:49:57, blass uri wrote: >On July 22, 2000 at 07:13:40, Ralf Elvsén wrote: > >>On July 22, 2000 at 06:05:00, blass uri wrote: >> >>> >>>It is not clear that programs are better than me in static evaluation in games >>>but the opposite is also not clear and I believe that the evaluation of >programs is more comlicated than the evaluation of humans even if it is not >better. >> >>The evaluation of programs maybe consider more factors on average than >>humans. But humans have an ability to concentrate better on the >>important things in a position. If there is a kingside attack you >>don't care about overall pawn structure. You concentrate on tactics >>and king safety and try to refine that part of the evaluation. >>And in other positions it's the other way around. To code that >>ability to concentrate on the important things would be extremely >>hard, I think, and this is a part of the evaluation function that >>in programs is essentially blank. Suggestion: look at e.g. Crafty's >>evaluation. Then think about what you do yourself. I would be >>surprised if you still would think Crafty's evaluation is more >>complicated, or better for that matter. (I'm talking about >>static eval of course). > >There are cases that I am better in evaluating king attacks but not always. >I remember a case when I avoided a move because I was afraid of king safety >problems. > >I analyzed the position with programs after the game and found that they were >not afraid of the problem and they were right and I simply overestimated the >opponent's chances against my king. > >It is not clear to me that my static evaluation is better. > >Uri Perhaps your evaluation were correct. It is well known now (vide leko and kramnik games) that computers are not SOOO good in king safety terms. Why you not belive in you? :)
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.