Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: PV and Killer moves

Author: James Swafford

Date: 17:27:23 04/29/99

Go up one level in this thread


On April 29, 1999 at 15:31:54, Dann Corbit wrote:

>On April 29, 1999 at 15:15:37, Pham Hong Nguyen wrote:
>>Could anyone explain how is different between PV and Killer moves?
>pv is preferred variation.  That means that it is the sequence of moves the
>computer imagines is going to happen if both parties do their best.
>
>Killer moves are moves that look like really good ideas at first glance.  In
>evaluating what move to make, you want to look at pawn x queen and stuff like
>that before some quiet sort of move.  That way, if you run out of time, you will
>more likely have finished looking at the good moves.
>
>A pv is more or less the computer's current thinking about what the future
>sequence of moves is going to be.
>
>A killer move is something that goes into move ordering as far as which
>possibilities we should check out first and foremost.


To expand a bit on what Dann said, most programs keep a list
of 1-3 killer moves per ply of search.  I _believe_ most programs
do not allow killer moves to be captures.

A typical move ordering scheme would try the principal continuation
(or pv) move first, if one exist, followed by captures sorted
based on the expectation of material gain, followed by killer
moves and finally noncaptures and captures that are expected to
lose material.

A move becomes a "killer move" by causing a fail high.
Whenever a move qualifies as killer, the least popular move
in your killer list gets replaced.

There are countless other tricks, too.

--
James




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.