Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Static Exchange Evaluator

Author: Robert Hyatt

Date: 12:30:07 08/30/04

Go up one level in this thread


On August 30, 2004 at 13:56:57, David Dahlem wrote:

>On August 30, 2004 at 13:50:02, Robert Hyatt wrote:
>
>>On August 30, 2004 at 11:10:24, David Dahlem wrote:
>>
>>>On August 30, 2004 at 11:00:12, Robert Hyatt wrote:
>>>
>>>>On August 30, 2004 at 10:42:34, Rick Bischoff wrote:
>>>>
>>>>>Hello all,
>>>>>
>>>>>I currently use the MVV/LVA technique in my queiscent search and am not really
>>>>>happy with the bloat (it's running on my G4 now and getting some good NPS, but
>>>>>the depth isn't all that great due to aforementioned bloat) so I was wondering
>>>>>if anyone would be kind of enough to explain the principles behind SEE to me?
>>>>>
>>>>>i.e., how do you determine the correct capture order?  What do you do with bad
>>>>>captures?  Do you just give them a really low score or do you just exlude them
>>>>>completely from the search move list?
>>>>
>>>>
>>>>See is a series of captures on a _single_ square.  Just like you would do, your
>>>>program should always use the smallest possible piece next, when making a
>>>>capture.
>>>
>>>"always"? Are there not some situations where capturing with a larger (checking)
>>>piece would force a recapture, while a smaller piece would not, which would be
>>>advantageous?
>>>
>>>Dave
>>
>>Not in SEE.  There is no concept of "checks" in any SEE implementation I have
>>done or seen.
>>
>>SEE (static exchange evaluator) really is "dumb" in that regard. :)
>
>My point, the way i read your statement, "Just like you would do, your program
>should always use the smallest possible piece next, when making a capture.", is
>that you seem to be saying this is an "always" unchangeable chess rule, whether
>for humans or engines. :-)
>
>Dave

I was saying it in the context of SEE.  There I do _always_ use the smallest
piece next...



>
>>
>>>
>>>You minumax the result to see if the capture on that square wins,
>>>>breaks even, or loses material.



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.