Author: Christophe Theron
Date: 09:39:02 11/09/00
Go up one level in this thread
On November 08, 2000 at 05:17:03, Hermano Ecuadoriano wrote:
>On November 07, 2000 at 17:36:00, Hermano Ecuadoriano wrote:
>
>>Thanks.
>>I think the smarter program, Tiger, should try to keep more pawns on the board.
>>But I can see how its style would be especially difficult for humans: the
>>structure keeps moving. The argument keeps changing...
>>
>>Tiger understands this principle, or method, of the continuous conversion of one
>>advantage into another (as opposed to the older idea of the "consistent game").
>>So, Tiger plays according to what used to be called the "Soviet School",
>>according to my understanding.
>>I think Tiger plays like Bronstein played against Botvinnik.
>
>Bronstein, in his excellent book on the 1950 match, said (I paraphrase):
>"Above all I sought to avoid the grip of standard technical devices. And
>in this respect, I succeeded. In no game did Botvinnik display his usual
>orderly conversion of some standard technical advantage into an engame win
>(because his technical demonstrations were distracted or diverted by my
>irruptions). But in order to achieve this, I wore myself out unnecessarily".
>
>While Botvinnik wanted to have a "discussion" about a weak pawn, Bronstein
>wanted continuously to change the subject, ultimately getting compensation
>elsewhere. This requires faith, because one's (non-material) dynamic
>potential must eventually be transformed into something concrete, like a
>checkmate or a new queen, which cannot yet be seen. ("Faith is the
>substance of things hoped for", etc.)
>Kasparov characterized his victory over Karpov in these terms.
>
>Now I will explain how this applies to Tiger.
>In the old days of computer chess, positions of static equilibrium often
>arose in games between fairly equal computers. Each sought its own values,
>and then conservatively held on to whatever it had achieved. And unless they
>(sometimes haphazardly) discovered something else to do, their play would
>often become embarassing. I think everyone knows what I mean. Much of
>"anti-computer" play is about exposing exactly this.
>To escape from this scenario, a random thrashing about is not good enough
>(because the opponent is a presumably equally strong calculator). What is
>required is some very sophisticated knowlege, some "dynamic savvy".
>This is clear: some kind of valuation, be it a number or a "feeling", must be
>assigned to the "dynamism" in the position. And the valuation must not be
>nonsense: it must consistently "lead" to "something". And here is my claim:
>The programmer of Tiger has made a lot of progress here.
>
>(And now this discussion is getting difficult for me because I was always
>extremely "conservative" in this context. Since I don't understand it, I must
>admire someone who has encoded it successfully, because one must understand
>before writing the code.)
>
>If you play through the PGN generously provided, above, I think you will see
>this frequent mobile transformation of values. I think it is very
>noticeable and very interesting. And I credit Tiger for this, because I already
>know how Fritz plays. :)
It is a nice compliment, thank you. I'm not even sure I could tell you how I did
this in Tiger. I try to improve little by little, when I see a weakness I try to
fix it, and I do it over and over. Eventually the program sees things I do not
even understand myself.
Christophe
>I drew the first game I played against Rebel 10 because I understood the pawn
>structure and was able to keep the game under control. (I don't remember the
>time control. PII-300.) I might never draw a similar game against Tiger for
>the reasons given above.
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.