Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hubner - Fritz: 3-3 is really disappointing !

Author: Otello Gnaramori

Date: 02:59:11 07/24/01

Go up one level in this thread


On July 24, 2001 at 04:40:21, Uri Blass wrote:

>On July 24, 2001 at 03:36:04, Otello Gnaramori wrote:
>
>>On July 23, 2001 at 17:54:26, Kevin Stafford wrote:
>>
>>>>The point made above was just to stress the importance of the tactics , let me
>>>>add the following inspired words by I.Marin :
>>>>
>>>>"Chess is ONLY moves in a chessboard, chess is only tactics. If you calculate
>>>>thoroughly enough you will beat Kasparov and everybody else because "positional
>>>>mistakes" simply don't exist: they are simply tactical errors with long term
>>>>consequences."
>>>>
>>>>Best Regards
>>>
>>>While Ignacio is correct that chess is only tactics in the most abstract sense,
>>>this quote is also slightly misleading. This is because both humans and
>>>computers have a finite tactical window beyond which they cannot accurately
>>>calculate (at least in a reasonable amount of time). It has been shown that
>>>increasing the depth of this window is difficult, as the search tree grows
>>>exponentially with each ply increase. It is for this reason that positional
>>>considerations are in fact relevant.
>>
>>I think that the above consideration is applicable to the humans, since the
>>comps can easily manage huge search trees and have tactical skills far beyond
>>any human player.
>>
>>>
>>>I define a positional move as one that leads to no obvious, immediate tactic,
>>>but is speculative in that it might lead to tactics which are beyond the current
>>>tactical window. No one is disputing that tactics are not extremely important
>>>(and in my opinion, are the very soul of chess), but I do believe that playing
>>>for tactics alone will put you at a disadvantage to a person or machine with
>>>positional understanding, unless your tactical window is sufficiently deeper
>>>than theirs.
>>
>>Exactly , If I can see one move deeper than you, you can have a fantastic
>>strategy but at the end I will be the winner.
>>
>> It is for this very reason that there has been a shift in chess
>>>programming away from exclusively fast searchers towards engines with more
>>>"chess knowledge" in their evaluation functions.
>>>
>>>-Kevin
>>
>>There are different "schools of thought" about the above point: you have to get
>>an eval that doesn't slow down too much the search speed (i.e. not too complex),
>>otherwise the chess knowledge is of no avail , since you have not enough
>>deepness of search.
>
>I do not believe in it.
>a complex evaluation function can help you to be better in tactics because you
>can get better rules which lines to prune and which lines to extend.
>
>Uri

I agree with you , Uri.
But if the function is "too complex" don't you think that there is a clear risk
to impact the overall search speed ?





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.