Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: When to do a null move search - an experiment

Author: Vincent Diepeveen

Date: 04:49:10 04/30/04

Go up one level in this thread


On April 29, 2004 at 09:28:53, Ed Schröder wrote:

>On April 29, 2004 at 07:37:23, Vincent Diepeveen wrote:
>
>[ snips ]
>
>>>This is all very poor Vince, I assume you don't play much with nowadays top
>>>programs. From 1982 to 2001 Rebel won its games by positional understanding and
>>>not by search and Rebel lost its games because it was outsearched. Today Rebel
>>>isn't outsearched at all, it now loses its games because the current top
>>>programs have a better positional understanding than Rebel.
>>>
>>>You should have a good look at the current tops, the positional progress has
>>>been great the last years. To me it all seems to indicate (provided your search
>>>is okay) the only way to make progress is to improve on chess knowledge. But
>>>what's new, I already came to that conclusion in 1986 after some intensive talks
>>>with Hans Berliner.
>
>>What i mean is Ed, is that you would not have accomplished the great results
>>with Rebel which you managed, had you just searched with a fullwidth search +
>>bunch of checks in qsearch.
>
>No of course not, brute force is silly, Rebel since day 1 has been a selective
>program. But I am getting your point, in the days before the nullmove was
>discovered Genius and Rebel had the best (static) selective search, a dominant
>factor in their successes, is that what you meant to say? If so, it is true.

That's what i mean, using a simple form of search in those days meant fullwidth.

>If only Frans had kept his mouth shut to Chrilly (Chrilly leaking nullmove in
>the ICCA journal) it is very likely Fritz would been the next Richard Lang still
>dominating all the rating lists and WCC's for the last decade. But Frans didn't
>and then all bets were off.

Frans' world title in 1995 had a huge positive impact.

>
>>I am under the impression that you just like diep try checks at several depths
>>in the qsearch. In diep i can try at the entire 32 ply of the qsearch checks.
>
>Checks in QS are relative cheap nevertheless I have limited them more and more
>the last years. There is little sense doing long range QS checks if you already
>hit 12-14 plies.

I see that different, but i must add the note that Rebel is basically tactical
selective in main search already and DIEP isn't. I do very little extensions in
diep other than singular. Even check hardly gets extended, i'll extend it when
it's singular.

>
>>Doing things like attacks in eval and mobility and scans for all kind of things
>>which are trivial for chessplayers and i do not even dare to write down the name
>>for here, they slow down once engine.
>
>>I would search 3 ply at 1991 hardware with it, simply because the code size is
>>so huge, the nps at 1991 machine (i had a 10Mhz XT at the time) would be around
>>a 100 nodes a second or so.
>
>You are overreacting of course.

Well of course when programming for a 10Mhz machine i would add huge selectivity
too, but i'm just talking about a theoretical math here now.

My code size doesn't even fit within the ROMs you had in 1991 :)

>>My point is would you have become world champion in 1991 searching 3 ply?
>
>8-10 plies was sufficient.

8-10 plies at 10Mhz with for its time a great eval is really a grandmaster piece
of work. I actually was not so long ago at a friend of mine (IM) having a
chessmachine with inside an Ed Schroeder program. I really was amazed that you
got such a depth out of such a tiny hardware!

In fact it completely outsearches diep 1997 which got 8 ply at the world champs
1997 at a PII300Mhz.

>
>>I very deliberately ask it this way, because fritz3 (1995) searching at todays
>>hardware handsdown would search 20 ply in any middlegame, when it would be
>>converted. Apart from that it single cpu would search 3.5 million nodes a second
>>hands down.
>
>20 plies?
>Come on.

No kidding.

Huber clearly has proven that with MTD and a simple eval (basically material)
you can search depths of 30+ ply.

The passive way in which fritz3 developed its pieces is very helpful. You get
*everywhere* nullmove cutoffs.

With an agressive tuned engine such a thing is of course a lot more difficult to
do, if not nearly impossible at todays hardware.

>
>
>>Todays fritz searched in 2003 world champs at a quad xeon 2.8Ghz about 13-15
>>ply.
>>
>>The 2003 fritz at a 386 , 10Mhz would have a problem getting beyond 4 ply.
>
>I don't believe that.

In 1999 fritz searched 17 ply everywhere at a quad xeon 500Mhz in middlegame.

In 2003 fritz at a quad xeon 2.8Ghz searched 14 ply in middlegame.

Those 14 ply from fritz2003 however completely destroy that 17 ply from
fritz1999.

I know why this happens. I see it at many engines including DIEP.

>
>>Would you beat it with Rebel-Madrid?
>
>Fritz3 was a weak program, in tactics and positional play. Fritz5 was a
>revelation, excellent search for those days (the nullmove era starts here) still

Fritz3 won the world champs 1995, so for its day it can't have been too bad.

>Fritz5 was positional weak. Rebel at that time was typically out-searched by 2-3
>average, it had to win its games by positional play. Then from Fritz6,7 and 8
>Fritz slowly became an excellent positional player.

>And regarding Rebel at 1991, don't get too exited, you can't imagine the number
>of holes in its selective search at that time.

I'm sure there were huge holes, but search depth in those days were more
important than it is now.

>My best,
>
>Ed



This page took 0.01 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.