Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Participants 2003 world champs

Author: Vincent Diepeveen

Date: 08:49:20 10/01/03

Go up one level in this thread


On September 30, 2003 at 19:44:02, Peter Skinner wrote:

>On September 30, 2003 at 17:15:00, Jorge Pichard wrote:
>
>>One of these four will win:
>>
>>Junior
>>Shredder
>>Fritz
>>Brutus
>>
>>The others will gain experience :-)
>
>I think Diep might have a chance. Vincent is pretty confident with his 500
>processors, and has been working hard. (People might think hell just froze over
>with me saying that..)

Thanks for the compliment Peter. Especially from you that's nice to hear.

>Out of your four possible winners I would replace Brutus with Deep Sjeng.

>Fritz may have a shot, although I think it will lose to both Shredder and
>Junior.

Forget Fritz, even despite that it will get 0.5 points for free from Brutus,
perhaps even more, it has no chance.

Even at equal hardware fritz has a problem and so far it showed up world champs
at faster hardware than the other strong programs.

World champs 1999: quad xeon 500Mhz. No one had that except junior.
Of course zugzwang, cilkchess and p.conners are programs from the 80s in some
respect.

Shredder is always the favourite of course and junior because of their mercilous
mentality at the board.

Yet i feel you're not very well understanding DIEP yet.

It's 5th of world with a poor search depth in 2002. All kind of horrors happened
and some only a few insiders know that i was not to blame for. I slept 2 hours a
day, not taking into account disasters like that in july the machine would be in
maintenance, to start with.

>I would love to see Tiger, Rebel, and especially Gandalf enter. Crafty on a good
>Alpha machine would be pleasant.

None of these programs join of course. Gandalf would enter, i'm sure of that,
when it would be possible for them physically to enter. They always were there
previous world champs when diseases didn't cause problems.

>If all those participants were to enter my top 4 choices would be:
>
>1. Shredder (Stefan always seems to win these events.)

Shredder is always the favourite.

>2. Junior (Past two years it has played very tough)

too.

>3. Chess Tiger (I personally think Christophe has something up his sleeve still)

Tiger scores near 0% against other engines when other engines are dual, quad,
not to mention 500 processors.

It is perhaps good at SSDF at equal hardware and a few coins for SSDF dudes.

>4. Gandalf/Diep (Kind of a crap shoot. Has Steen fixed all his passed pawn
>problems?)

So that places me 3d, that's cool, because gandalf ain't there and tiger isn't
either. Also the general public is never going to understand what is important.

I also guess many will bet at brutus.

However consider this.

He could get a 192 processor supercomputer from university Paderborn with 100
fpga cards (Wuellenweber was telling that proudly world champs 2002 against me).

With all respect but i can get in software effectively a lot faster at that 192
processor than Brutus in hardware.

I read at chessbase.de it will join at 4 processors or so now. 4 hardware cards
in i bet a dual or quad machine.

That's quite different from 100 cards.

Chrilly has nothing to do except program for brutus and get it parallel to work
at this supercomputer. I gotta earn a living too, but in the meantime also
managed to get it to work at 500 processors.

What's harder. 192 processors or 500?

So no doubt brutus will be a strong program, but there is no hardware advantage
compared to say a quad opteron for example.



Yet a few bad games like in paderborn2003 doesn't say a thing about how it will
play in world champs 2003. We all know how good Chrilly is at focussing at a
program and getting results without feeling strong.

Shredder feels strong and plays strong. DIEP feels strong and plays strong.
Nimzo/Brutus software always feels weak but scores better than they positionally
seem to be.

So no doubt it will be dangerous from that viewpoint.

Most dangerous though will be Kure's preparation.

Book really matters more these days because openings phase without exception is
the weak phase of all engines (when compared to humans) and in chess the weakest
chain counts, not the strongest chain.

In diep however all chains are a lot stronger thanks to 500 processors.
We're not talking about a factor 2 faster here but hundreds!

Just imagine how many plies those hashtables give in endgames!!!!!!!!!!

I'll blindfolded take any endgame i can get!

Another advantage of 500 processors is that programs which are based upon
outsearching the opponent have a problem against you. Best examples are of
course Junior and Shredder and Fritz.

I remember diep-junior world champs 2002 where i was searching 9 ply with diep
which is a very poor depth. And diep - shredder where i was searching 10 ply in
a won position for diep and diep missing the win.

Of course better debugged evaluation in such case can find it, but another 5 ply
or so is easier.

>I think those would be good. Alot of people will be betting on Brutus. I don't
>think it will fair well, and finish in the middle of the pack. I think Crafty
>might do well on a fast machine. The lastest version of Crafty on ICC has been
>playing extremely well.
>
>Peter.



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.