Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: MVA/LVA sorting order

Author: Tom King

Date: 05:48:01 03/15/99

Go up one level in this thread


On March 14, 1999 at 01:23:38, Peter McKenzie wrote:

>On March 13, 1999 at 17:23:24, Tom King wrote:
>
>>On March 13, 1999 at 07:48:31, Tom King wrote:
>>
>>>On March 12, 1999 at 02:45:52, Peter McKenzie wrote:
>>>
>>>>On March 12, 1999 at 02:26:49, JW de Kort wrote:
>>>>
>>[SNIP]
>>>>>e.g suposse we have the following possible captures: PxN, PxR, QxR
>>>>>
>>>>>using methode 1: PxR, PxN, QxR
>>>>>using methode 2: PxR, QxR, PxN
>>>>>
>>>>>Wich one is MVA/LVA?
>>>>
>>>>It is method 2.
>>>>As a finesse, you can try sorting King captures first (so KxR comes before PxR).
>>>
>>>Is this good (King captures first)? Doesn't this go against the definition of
>>>MVVLVA?
>
>The idea is that captures by the King always win the piece, because the King
>can't move into check.  Captures by other pieces may result in the capturing
>piece being recaptured.
>
>This might be complicated by the way you detect King moves into check.
>

Right. I allow moves which put the king into check(!). They are filtered out at
the next ply.

>>>
>>I just ran some quick tests. For my program following MVVLVA rather than MVVLVA
>>(but sorting these king captures first) seems to be about 5% faster. Anyone else
>>tried this?
>
>I tested it ages ago, can't remember how, but putting the King captures first
>seemed to help.  What tests did you do?

I just ran two version my program on a bunch of positions. V1 used MVVLVA and V2
used MVVLVA-with-king-captures first. I found "pure" MVVLVA to be faster about
80% of the time. Sometimes it was only slightly faster, but sometimes it was
significant (say 30%).



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.