Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Searching for an old programming discussion ...

Author: Eelco de Groot

Date: 14:59:33 07/21/00

Go up one level in this thread


On July 21, 2000 at 17:44:42, Georg v. Zimmermann wrote:

>On July 21, 2000 at 17:39:52, Eelco de Groot wrote:
>
>>I'm sure more than 10 people would be interested in idea number two if it had a
>>search function. What programming discussion were you looking for, exactly?
>>
>>Eelco
>
>Some explanation about Futility Pruning and lower bound by E.Heinz. I don't even
>know when it took place, someone pointed me to it.
>
>Georg :)

I found this in my own files:
--------------------------------------------------------------------------------
Subject: Re: Several chess programming questionsSubject: Re: Several chess
programming questions



[ Post Followup (without quoting) ] [ Post Followup (with quoting) ] [ Computer
Chess Resource Center ] [ Message Listing ]



Posted by Ernst A. Heinz on April 07, 2000 at 12:26:11:
In Reply to: Re: Several chess programming questions posted by Tom Kerrigan on
April 07, 2000 at 04:14:39:
>> 3) does "razoring" mean pruning in general (or even alfa-beta method)?
>
>No, razoring is when you do futility pruning at the last ply of your full-width
>search.

No, Tom -- what you describe is not razoring but standard futility
pruning at frontier nodes with a remaining search depth of 1 ply.

> I think most people don't like it.

Wrong again.

Most people do not like razoring. Futility pruning at frontier
nodes, however, is _theoretically sound_ for searchers with a
capture-only quiescence and a futility margin that is at least
as large as the maximal positional score produced by the static
evaluation function. Standard futility pruning simply "lifts"
stand-pat cutoffs from horizon nodes to frontier nodes.

For more details, please see my article about "Extended Futility
Pruning" in the ICCA Journal 21(2) and my book on "Scalable Search
in Computer Chess" (http://supertech.lcs.mit.edu/~heinz/node1.html)
for  more details.

>I tried it and all my test suite
>scores went way down, even when the futility threshold was pretty conservative.
>It just causes to the search to miss too much...

According to my own experience and that of numerous others, this
seems very strange. Maybe, you did not implement futility pruning
but real razoring at frontier nodes. Your above quoted results
sound like you did so.

=Ernst=






  Re: Several chess programming questions - Tom Kerrigan 13:03:13 4/07/2000 (0)



[ Post Followup (without quoting) ] [ Post Followup (with quoting) ] [ Computer
Chess Resource Center ] [ Message Listing ]


--------------------------------------------------------------------------------

I don't think I have the follow-up from Tom , though.
Eelco



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.