Author: Albert Silver
Date: 08:02:17 06/01/00
Go up one level in this thread
On May 31, 2000 at 11:15:01, Simon Waters wrote:
>On May 26, 2000 at 16:50:55, Albert Silver wrote:
>
>>On May 25, 2000 at 19:32:13, Simon Waters wrote:
>>
>>>On May 22, 2000 at 19:55:54, Peter McKenzie wrote:
>
>>>Hmm - well I tried a similar thing with simpler games and genetic type
>>>approaches.
>>>
>>>Here we needed some weighting to optimise play at a game - so I created a scheme
>>>to modify the weightings of winning programs to create new programs, and made
>>>them self play.
>>>
>>>The new generations would learn to beat the old ones, their off spring would
>>>beat them, and so on - however after a few generations it was clear the
>>>strategies were going in circles, like a never ending games of paper, scissors,
>>>stone.
>>
>>Do you mean that older versions could beat the latest ones? i.e. Program A was
>>beaten by Program B that was beaten by Program C that lost to Program A? What
>>were some of the changes done and how and when was the circle closed?
>
>Exactly.
>
>As I said not chess alas - but I would be surprised if similar things were not
>true in any complex enough problem.
In computer vs. computer play it is more difficult to apply, because a lot
revolves around the opposing tactical and positional play. Programs as a rule
are very lacking in many aspects of positional play so that even so-called
positional programs that top the lists reveal themselves to be tremendous
tacticians as well. To be fair, according to Irazoquoi's private tests, Junior 6
would be an exception to this rule as it is not tactically superior but _is_
beating the competition. In any case, as a consequence of the lopsided nature of
their strength, the SSDF tests have demonstrated which programs excel in tactics
more than any other. Human play is quite different and encompasses a far greater
range of strengths. So that when a program plays a human player, it may be
fighting with the same weapons (tactics), or it may not if the player is a
die-hard positional player. Some argue that in the latter case, this wrongly
prejudices the computer's possibilities and therefore limits its potential
rating, but that is silly as chess is not just tactics, otherwise you might as
well be like a car salesman:
- Yes, with its supercharged V8 and 350 hp, it will go from 0 to 100 km/h in 4.5
seconds flat.
- Wow, that's amazing. (Doesn't know that the car's speed limit is 110 km/h)
I don't think that piling everything on one strength will cut it, but I don't
think we'll be going in circles either. The reason is that we aren't working
with a fixed set in which nothing progresses. New knowledge is added, and new
techniques to improve tactics or accelerate the search are added all the time.
Unless one is preparing the program to play against a specific opponent, I would
assume that finding the best balance is the wisest approach.
Albert Silver
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.