Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Evaluation at start versus eval at node. Why not mix them?

Author: James Robertson

Date: 10:39:21 06/12/99

Go up one level in this thread


On June 12, 1999 at 07:05:35, Torstein Hall wrote:

>On June 11, 1999 at 12:01:38, KarinsDad wrote:
>
>>On June 11, 1999 at 10:15:19, Torstein Hall wrote:
>>
>>>Its known that Fritz and some other fast programs reach their high nps because
>>>they do their evaluation mainly at the start of the search.
>>>
>>>Is it not possible to mix this up a bit. So that you do a "big eval" at the
>>>start, then at certain depths in the search. Or even better, if some condition
>>>sets in, like swap of queeen, endgame etc, then do a new thorough eval, else
>>>just the fast eval?
>>>
>>>But perhaps some programs do it that way already? Or perhaps its just not
>>>possible to implement?
>>>
>>>
>>>Torstein
>>
>>I am trying something similar to this in my program. I will let everyone know
>>how it works out (once the program actually works, hopefully later this summer).
>>
>
>Very intersting,
>
>Perhaps you can get the best from both worlds, a fast nps and a correct eval!
>
>I think this also can be the most human like approach to chess programming. I
>belive strong players do a more thorough eval at the end of forced variations,
>when entering the endgame etc., else its most tactics!
>
>Torstein
>
>>KarinsDad :)

The cool thing about being a human is that a good player will look at a position
(even if it is very far into a search) and evaluate it instantly (I hear this is
called pattern recognition?), so they can afford to do a full eval at the leaf
nodes!

Anyway, if a human has a NPS of 3 (or in my case .2), doing a full eval doesn't
really reduce the number of positions one could look at. :)

James



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.