Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Question bout auto pv playing

Author: Wylie Garvin

Date: 19:22:17 01/31/02

Go up one level in this thread


On January 28, 2002 at 18:35:53, Andrew Williams wrote:

>On January 28, 2002 at 17:04:01, William H Rogers wrote:
>
>>What you stated might be what has really happened. I will give you an example of
>>what I mean and hope it makes sense.
>>Suppose that this is my pv:
>>
>>1. e2e4, e7e5, g1f3 etc.
>>
>>I really don't mean at the beginnig of the game, it well could be move 20 or
>>something. The question is if a program predicts e7e5 would it be logical to
>>just make the move g1f3 with recalculating all other moves once again?
>>
>
>I can't see what the advantage would be. Surely that would be just thinking
>once every two moves? I don't think I'd want my program doing that.
>
>>Please do not confuse this example with the real thing. Some of the older
>>programs include "My Chess" and a few others. I don't think that they used
>>pondering at that time.
>>Thanks
>>Bill
>
>Andrew

I think this auto-pv playing would NOT be advisable, and you should work on
pondering instead.  After all, if you let your program search N plies deep to
make the initial PV, and then the opponent makes the predicted response and you
play your previously calculated reply, that is essentially the result of a
search to only N-2 plies.  If you made a third move off the same PV it would be
N-4 plies.  This would probably cause a noticable drop in strength..

The only case where I can think of where it wouldn't matter is if every line
searched terminated in a *known* game-theoretic value (i.e. every line ended in
mate or stalemate or reached into the EGTB's).  In that case, repeating the
search to a deeper depth would just cause the same thing to happen.  AFAIK, if
you never evaluate or quiesce during a search, then this will happen and you can
stop searching.  But by the time this happens, you may have already found a mate
and even if you haven't, I don't imagine that it costs much to re-search the
remaining tree two plies shallower for each move.

Anyway--there's a lot of interesting things you can do with pondering.  E.g. If
your program decides that you've thought enough about the reply you would make
to the opponent's apparent best move, you start thinking about his 2nd-best
move.  Or, you could start "pondering" on his reply to your reply to his reply.
Or you could start drawing Mandelbrot images... =)

cheers,
wylie



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.