Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Move ordering metric

Author: Uri Blass

Date: 08:44:06 11/13/03

Go up one level in this thread


On November 13, 2003 at 11:29:18, Daniel Clausen wrote:

>On November 13, 2003 at 11:10:53, Uri Blass wrote:
>
>>On November 13, 2003 at 10:53:22, Daniel Clausen wrote:
>>
>>>On November 13, 2003 at 10:13:34, Uri Blass wrote:
>>>
>>>>On November 13, 2003 at 09:44:22, Daniel Clausen wrote:
>>>>
>>>>>On November 12, 2003 at 18:40:32, Peter Kappler wrote:
>>>>>
>>>>>[snip]
>>>>>
>>>>>>>I get an average of better than 90%. It takes tuning. Work on killers and
>>>>>>>history pays off. Make sure you search bad captures before other moves also.
>>>>>>>Taking capture out of killers gave me a boost also.
>>>>>>>
>>>>>>
>>>>>>Hmmm, I think I do bad captures at the end, and as I recall it was a clear win.
>>>>>
>>>>>IIRC, for most people trying bad captures at the end was better with SEE, but
>>>>>worse with MVV/LVA.
>>>>>
>>>>>Sargon
>>>>
>>>>How do you know that a capture is bad with MVV/LVA?
>>>
>>>When someone uses SEE, a capture is bad when the SEE-value for it is negative.
>>>When someone uses MVV/LVA, a capture is bad when the MVV/LVA-value for it is
>>>negative.
>>>
>>>It's that simple. :)
>>>
>>>Sargon
>>
>>It is not so simple.
>>
>>By this definition every pawn capture by knight,bishop,rook,queen,king is a bad
>>capture.
>>
>>It does not make sense so it does not make sense even to try to put it after non
>>captures.
>>
>>You can have a function to detect part of the captures as bad captures and also
>>use MVV/LVA for the captures that you do not detect as bad.
>>
>>The function that detect bad captures is not SEE because SEE gives you the
>>expected value of the capture and not only the information if the capture is
>>good or bad.
>>
>>SEE can order QxN before QxQ when both are good captures when the special
>>function does not do it.
>>
>>In that case the question if to put bad captures before history moves or not to
>>do it is dependent on the definition of bad capture.
>>
>>Uri
>
>
>I retract everything I said in this thread. :)
>
>I still had the (false) idea in my head that to calculate the score of a capture
>in MVV/LVA, you subtract the value of your piece from the value of the captured
>piece, which leads to the problems you pointed above.
>
>But I guess MVV/LVA just means that the order of captures to try is
>
>PxQ, B/NxQ, RxQ, QxQ, KxQ, PxR, B/NxR, RxR, QxR, KxR etc
>
>With this definition there's simply no way to distinguish between good and bad
>captures.
>
>Sargon

This is exactly the way that people use MVV/LVA but in order to do it they use
scores.

I have

g->index = (piece(to) * 8) - piece(from);

I have
pawn=0,knight=1,bishop=2,rook=3,queen=4,king=5

These scores mean that I have exactly negative scores for NXP,BxP,RxP,QxP,KxP

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.