Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Optimal Designs for Anti-Human and Anti-Computer Play Different?

Author: Bob Durrett

Date: 07:05:03 01/09/04

Go up one level in this thread


On January 09, 2004 at 08:53:24, Mike S. wrote:

>On January 08, 2004 at 19:07:23, Bob Durrett wrote:
>
>>Your observations seem to suggest, perhaps astutely, that the best design for an
>>anti-human engine may be significantly different from the optimal design for
>>play against other computers.  Agree?
>
>Yes and no :-) I think there are good arguments for both extremes:
>
>1. Even more tactical, because that's the biggest strength of programs compared
>to humans (and very deep calculation might replace some sophisticated positinal
>evaluations), or
>
>2. Less tactical (because they can easily be much better than humans anyway),
>and with every possible improvement of the traditional program weaknesses
>positional play (and/or even strategical decisions).
>
>As very simplified (and therefore not quite correct, but illustrative) examples
>we could say, that Fritz has been more the type (1) approach, and Shredder more
>type (2).
>
>Actually Fritz was moving towards type (2) in recent version history though.
>After the X3D match, Frans Morsch said Fritz will now be entirely developed
>further considering human opposition (only).
>
>Hiarcs is somewhat difficult to categorize between (1) and (2), because it
>always had a good "positional reputation", but was among the fastest combinators
>at the same time (at least since version 6).
>
>Maybe the "most type (1)" engine currently is Junior.
>
>In practise, I think most engines are a 1/2 mixture. Both a quick search and a
>good evaluation will obviously also be good against other engines too, not only
>against humans. Maybe the optimal anti-human and anti-computer designes would be
>much less different than we assume.
>
>Regards,
>M.Scheidl

Actually, I was keying on your "deep search and effective extensions."  It
occurred to me that maybe the searching, pruning, and all the other wonderful
tricks [futility pruning, lazy evaluation, etc.] discussed here at CCC may need
to be revisited.  Maybe some of these are better for engine vs engine and others
better for play against human opponents.  Personally, I do not know the answers
to all this since it's "above my pay grade."  : )

Bob D.



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.