Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Engines for long analysis

Author: Uri Blass

Date: 13:08:25 02/20/05

Go up one level in this thread


On February 20, 2005 at 15:34:37, Jonas Cohonas wrote:

>>Most improvements are improvements in every time control.
>>
>>Uri
>
>So some are not?
>
>Just out of curiosity, how many times have you left out code in your engine,
>that would, in theory be good, but the added knowledge would slow your engine
>down to degree that made you leave it out? (if any)

I never left code that I know that in theory will be good.
I left code out of the program because of fear from bugs or because I did not
like the results but I think that adding knowledge to the evaluation does not do
the engine significantly slower.


>
>How come some engines do poorly at blitz controls, but get better at longer time
>controls? i don't mean to bring up the good programmer vs bad programmer issue
>here, but focus more on the potential possibility that, those engines might have
>some code that makes them better at longer time controls.

I think that an important factor here is better search algorithm.

>
>How about selectivity, i speak in laymans terms here, is it not so that with
>this implimented the engine will "filter out bad moves" upfront and only
>concentrate on moves that makes sense early in the search, would the value of
>this implimentattion (selectivity) not mean that the engine might find stronger
>moves, if it was given much longer to think? (i am not sure that selectivity is
>this feature of filtering out moves, but it is the feature i am talking about
>not the name :)


pruning is one of the things that may help.
I think that it is important at all time control but I guess it is more
important for long time control.

order of moves is something that is more important for long time control.
I cannot say something general about the value of evaluation.

I guess that there are parts that are more important for blitz and parts that
are more important for long time control.

suppose your strong engine does not have the evaluation knowledge to win KR vs
K(I read that fruit has problems with it in
http://wbforum.volker-pittlik.name/viewtopic.php?t=1619).
At blitz it may be a problem because it does not search deep enough when at long
time control it may find the win by search.

Uri



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.