Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Interesting mate test for hashing

Author: Bas Hamstra

Date: 20:03:25 09/11/99

Go up one level in this thread


On September 11, 1999 at 22:23:18, Robert Hyatt wrote:

>On September 11, 1999 at 15:59:19, Ed Schröder wrote:
>
>>On September 11, 1999 at 15:36:18, Robert Hyatt wrote:
>>
>>>OK... then we are 'in sync' here it seems..
>>>
>>>As far as PVS, the main advantage is that since almost everything is searched
>>>with a null-window, it saves nodes, _IF_ you do well at move ordering (I have
>>>no doubt that you do well so PVS might be a win for you too)...
>>>
>>>It reduced my trees by 10% and loses nothing at all...  unless you screw
>>>up move ordering, then it can make the tree bigger as you first search with
>>>a null window, then you have to re-search with the normal window...
>>
>>Thus, PVS is aspiration search with a null-window? Is that all there is?
>>
>>Ed
>
>
>basically yes.  You search the first move at any ply with the window that is
>passed in.  You search the remainder of the moves with alpha,alpha+1.  And all
>you care about is the case where another move comes in >= alpha+1, because
>then you have to re-search that move with the original window.  Of course,
>99.99999999999% of the time, the window passed in is x,x+1, so this doesn't
>happen...
>
>It is very efficient, since it is a more controlled version of mtd(f) when you
>think about it.  IE mtd(f) inside the tree, but not along the PV...

I'm working on it too, based on your advice... I understand the principle: it's
cheaper to prove a move is worse than the first move, than to determine it's
exact score.

You say PVS gives you as least 10% over alphabeta%. Others have reported that
MTD(f) gives you at least 10% over PVS (someone mentioned 15%).

ab      100.000 nodes
PVS     90.000 nodes
MDT(f)  81.000 nodes ???

What is your opinion on this? Is there any reason to not do MDT(f) right away?

My intuition says this: if everyone *still* sets an aspiration window for PVS,
it must somehow be *non* optimal. Don't you agree? For you have seen it still
searches more than necessary (on average, in games), and try to restrict that by
setting an initial window. An optimal method wouldn't need that. Do you have any
interesting philosophies on that issue?

Regards,

Bas Hamstra.





































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.