Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Play-the-game test Introduction/FAQ

Author: Thorsten Czub

Date: 14:29:10 12/14/97

Go up one level in this thread


>I think a few years ago, the only thing that mattered was tactics in
>computer chess.   The programs had to see past various horizon affect
>problems and survive simple tactics just to be in the game at all.
>Even when it looked like they were getting good tactically, a simply
>threat or two could push everything beyond their horizon.   So good
>positional play helped, but was a minor factor in their success.

Right !

>But the right formula for computer chess, in my humble opinion, will
>be very fast searchers that are as smart as you can make them!

And everybody tries this. Junior is one good example. Amir always
repeats this point of view. Your above sentence could be a quotation of
what Amir told me in Paris.

>This sounds so obvious I do not know why it's even debated.  And it
>seems to me that the cream of the crop best programs are following
>this formula very closely!

I think this is true for Rebel, Junior, maybe also Fritz (sorry - Frans)
tries it. But I don't think Hiarcs or Mchess really DO this.
And Johann de Koning also has not "increased" the speed until now (but
on the other side he has not increased the strength of The King in the
last versions).
Hiarcs and Mchess HAVE definetely increased their strength although
their speed has not increased much.

>
>But it's getting harder and harder to avoid the evaluation part.  I
>once believed that as programs became faster, knowledge would matter
>less.  But I now believe just the opposite.    My eyes have been
>opened!

Brilliant.
But i still don't want to think ONLY in this category you mentioned:
fast and smart !
We should also have the strength to try different approaches. With the
risk that this approaches lose.
Therefore I still try to work with chris. As long as his program
increases in strength although the NPS goes down, it makes sense to me
to continue working on different approaches. If we would all only follow
the same ideas, life would be boring.
But it is very very very hard to stand watching that the outsearching
aspects (more plies, more NPS) makes the fast-searchers WIN against
CSTAL.
Not in Paris. In Paris CSTAL luckily survived , only against stronger
programs it died, but against the bunch of fast-programs it defended
although it was only doing 6000 NPS. Still I would like to know how
CSTal would play on Cilk-chess hardware. This could be an interesting
approach. When CSTal is able to stand against fast-searchers with 6000
NPS, how would it play with 60.000 NPS?


>
>It's getting harder to beat anyone tactically, most programs easily
>steer their way around the crap and what is left, is which program
>understands the deeper issues best.

Right.


>But this is a really exciting time for us!   We will soon be able
>to write programs that are better than the best human players.
>I think we almost have enough hardware now, but we need smarter
>programs to really do it right.

Confirm completely.

>But I also believe there is a long way to go past this.  I believe
>Kasparov is not that close to being the ultimate player, although
>he may be the best human (maybe even best human ever.)   At some
>point, computer programs will exist that can CONSISTANTLY beat all
>the best human players.   This means computer chess  will be alive
>for a long time and the emphasis will be on which program rules.

Best famous sunday words to end this day. You are right again IMO.



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.