Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: GM Kasparov stupider then most.......An overview of top programs!

Author: Timmay

Date: 08:48:30 08/30/03

Go up one level in this thread


I agree with Garry's assessment (this is coming from a National Master by the
way email me for my information if you want to). First of all I think Junior 7.0
is stronger than Junior 8.0, its more realistic, more useful in analysis. When
thinking on its own to come up with one move it can get quite stupid, no doubt.
It's useful to analyze with those engines though and see a few different top
lines and make an educated choice as a result, but not just run it forever and
think that it will be amazing. It might in wild complications (like any program)
but not in quieter positions!

Fritz (5.32 better for general tactical analysis, Fritz 8.0 better for "thinking
on its own for one move" and in quiet positions) is by far a better program than
Junior, Chess Tiger 14.0 (I think 15.0 weakened quite a bit, the solid tiger
14.0 is a better all around program. I'm basing all this off of my experiences.)
also is a truy very strong program with a great understanding of dynamics of a
position and the big p-word programmers love to hear: PLANS!(great job
Christophe!). This counters Fritz nicely. I think the best code anyone can put
in their program once the initial program is set is "mobility"! The best
programs believe in that a lot. ALL plans stem around getting all active pieces
in one way or another (pawn structure is the other part that should be included,
shredder and hiarcs to a good job of understanding pawn structure). The trickier
part is "coordinating" active pieces with active pieces for a common goal. Why
are weak squares nice for an opponent? Because it gives them an active knight.
It's easy for a computer to see an active knight for one, and its easy for a
computer to see how many black pieces it influences (being posted close to the
opponent's position). Two active bishops, also an advantage. Rooks on open
files, an advantage. And from my understanding Christophe used mobility
extensively in the Tiger engine. That is the FIRST program I've ever seen who
can make two-three-occasionally-four move plans. It's amazing how well it
manuevers pieces, attacks weaknesses, etc. The only clear weakness in Chess
Tiger 14.0 is not knowing what to do with the rooks. Fritz doesn't mind playing
moves like Re1 or Rc1 preparing pawn breaks, but Tiger is not so good with his
rooks sometimes. Not sure what the technical reason for that is I'm not a
programmer. Maybe because Re1 moves don't "immediately" give a rook mobility,
but after a standard pawn break like e4 it could get quite a lot. Could be
something to work on. Also Christophe my advice would be to make some sort of
"scale evalulation" like Fritz 8.0 has to change the emphasis from material to
position. Do that on the Chess Tiger 14.0. 15.0 is weaker because there's more
knowledge about king attacks etc. which i've noticed has made most programs
"weaker" not stronger. It's important to know about king safety, but Junior
cranks out h4's whenever it can!

Umm, another great program is Nimzo 8.0. What do you know also quite a bit of
emphasis on mobility! This is also a nice addition to Fritz and Tiger. The
problem is this is much more material-oriented than Tiger. But it's better with
the rooks, and good at pushing pawns to the control the center. Junior 7.0 is
occasionally useful to look at in analysis, but occasional being the key word.
In most positions it's not totally useful. The "knowledgey" programs such as
Hiarcs 8.0 or Shredder 7.0 are good at decision making like when to exchange, or
not weakening pawn structure, but when you're not making decisions like that
they're much weaker than the other programs mentioned. Shredder's moves when he
thinks on his own don't impress me much. Fritz's moves after thinking a long
time are always good, (they may be slightly worse in the strategic situation
because a computer can't plan as well as humans), Chess Tiger 14.0 is also a
great program and it seems to me it only needs a minute or a minute and a half
to shoot out some great moves!

So a quick overview, in my opinion, the absolute strongest programs today are
Fritz 8.0 (or 5.32 in analysis with multiple moves), Chess Tiger 14.0 (my
favorite program), Nimzo 8.0, and if you want a knowledge program (pawn
structures, understanding when to exchange) I'd pick Shredder 7.0. But I can
decide move about pawn structure or when to exchange pieces on my own, so
Shredder isn't one of my favorites. So if Christophe could get a "scale
evaluation" like what fritz 8.0 has on the Chess Tiger 14.0 (not 15.0)
evaluation and search, and could find the root of the problems Tiger has with
rooks, I think that program would be the closest thing to Garry Kasparov ever
invented. Best wishes.



This page took 0.01 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.