Author: Tina Long
Date: 05:29:56 11/11/99
Go up one level in this thread
On November 11, 1999 at 07:51:13, Eelco de Groot wrote:
>
>On November 11, 1999 at 06:10:34, blass uri wrote:
>
>>On November 11, 1999 at 05:15:30, Eelco de Groot wrote:
>>
>>>For Blitz you are probably right that 'depth' is probably better than
>>>'knowledge', Tina... But maybe only up to a point? If you know that you will be
>>>playing against a tactical monster might a more positional style and knowledge
>>>not be what could save you then?
>>
>>If you play against a tactical monster it is very dangerous.
>>one tactical mistake is enough to lose the game.
>>
>>I think that tactics is important also for correspondence games.
>>Deeper blue did not see the draw against kasparov in the final position of game
>>2 because of tactical problems.
>>
>>Ed discovered that tactics was important in his 1 hour/move games
>>
>>I do not believe that knowledge=500 is productive in practical games even at
>>slow time control.
>>
>>Uri
>
>Maybe you are right about that last bit, Uri. Only if you happen to be in a
>position where that last bit of knowledge applies you would of course be helped
>by it. The idea was more that if you had some extra defensive extensions that
>mainly look at tactical possibilities by opponent, you wouldn't have to
>sacrifice quite as much knowledge in Blitz (or rapid or even under tournament
>conditions against a fast opponent), or alternatively be a little less selective
>closer to the root. You could also use a lttle more time for looking at pins
>maybe, like Roy Brunjes described in his review or use one of other approaches
>like lessening value of pawns. But tactics remain more than 90%, I'm sure, of
>the game. Unfortunately not my strong point, tactics...
>
>Eelco
Maybe I'm off on a tangent, but I'd expect a "pareto principal" in the knowledge
parameter. The loss in search depth and nps not necesarily being offset by gain
in knowledge, particularly in the first minute or two of analysing a position.
I'm assuming that Rebel looks at all the MAJOR or IMPORTANT "knowledge" aspects,
or maybe applies these to the important parts of the search tree when knowledge
is set at 100. Going beyond 100 introduces less important knowledge aspects, or
applies them to a greater proportion of the search tree.
Moving knowledge from 400 to 500 is reducing the depth of search & the nps, but
providing what benefit in better evaluation.
I've looked at the analysis lines for positional positions from Informator with
various levels of knowledge, and while the speed & depth is clearly measurable,
the variations in analysis lines (the strength) is too subtle for me to judge.
I'm agreeing with us all in that I'll only be using knowledge=500 for ONite
analysis, along with ONite search method.
I'll get the depth and nips that Fritz would in 1/10 time, but each np (?) will
be more knowledgably evaluated. I am happy with this as Fritz may do nothing,
or perhaps gain 1 ply in the extra 9/10 of time.
If chess was a mathematical game, and the solution to any position could be
definitely stated {correct move 27.h6 +210} then we could tell what's best.
I'm eagerly awaiting the results of the competition. Will there be 120's & 85's
or 0's and 500's in the winning setups?
Anyway, I just play it like a toy, & I cheat like crazy, I've worn the print off
the back button.
Hi guys,
PS: in the style of that ahead-of-its-time British comedy Monty Pythons Flying
Circus, I wander these pages bellowing "Bella....Bella....Bella"
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.