Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: comps. GM strength maybe but can we agree that they are IM strength

Author: Josh Strayhorn

Date: 21:04:01 01/13/01

Go up one level in this thread


On January 13, 2001 at 23:00:54, Dann Corbit wrote:

>On January 13, 2001 at 17:13:56, Mike S. wrote:
>
>>The problem with this comparison is, that comps have a much different and much
>>more imbalanced profile than the average human master IMO.
>>
>>The games within this posting illustrate that (see at the end):
>>
>>http://f23.parsimony.net/forum50826/messages/7051.htm
>>
>>Some top programs on P600, still falling victim to the trojan sacrifice. As long
>>as this happens, I think we cannot talk of real IM strength.
>
>I don't think that single set of examples disprove it.  IM's can make blunders
>too, especially at fast time control games.  If you can demonstrate that
>statistically some program definitely has a weakness that can be explioted then
>we could say that the computer programs with that defect probably are not IM
>strength either.  That is because the weakness might be expoitable over and over
>again.  Even that would need to be demonstrated.  But a few examples won't prove
>that point.  I suspect every GM has made more than one blunder in his/her
>career.
>
>I also think that only games at tournament time controls should count in the
>discussions.  Typically, the trojan horse attacks are made at blitz time
>controls.  If you give the computers 40/2 time control, they are far less likely
>to get sucked into such a thing.


I think it's been proven beyond doubt that computers are quite capable of
crushing a grandmaster under certain conditions.  At the sime time, though, it's
also pretty clear that ordinary masters, and even some much weaker players, can
learn some simple techniques that allow them to turn the tables and solidly
defeat the machines.  That's what makes this difficult.

I recall about a year or so ago I stumbled on a page describing how to play
anti-computer chess.  This particular one was describing the stonewall setup,
which I'm sure most here are familiar with.  I read the document, played a
couple of practice games to get the idea, and then enjoyed an afternoon of
destroying the few free programs I had at the time until I got tired of it and
went back to 1. e4.

Now, I'm a fairly strong player, but even so, I don't think there's any book or
web page that will teach me, in less than 15 minutes, how to beat a human
grandmaster with regularity.

My opinion is that we can't really call computers grandmasters, or 'GM
strength,' until they are considerably less susceptible to tricks like this.
'GM strength' isn't a Platonic form, it's a measure of performance.  The
machines still have too many holes.







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.