Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: ProDeo: what is Chess Knowledge?

Author: Ed Schröder

Date: 09:56:58 08/12/04

Go up one level in this thread


On August 12, 2004 at 12:42:56, Albert Silver wrote:

>
>>>Thanks again, and interesting. You mentioned in the descriptions the various
>>>advantages of higher settings for some of the parameters, notably the Chess
>>>Knowledge. Aside from the King Safety, I take it that you now believe the
>>>default values are still best.
>>
>>Yes.
>>
>>Although Thorsten will strongly protest :)
>>
>>
>>
>>>A number of the settings aren't clear to me all the same. At least the
>>>differences between them. There are a number of parameters linked to attacking,
>>>such as Attack, King Safety, and Attractiveness. The difference between them is
>>>very unclear to me.
>>
>>[Attacking = xxx] is meant to be the successor of [Anti-GM = xxx] but since I
>>don't compete any longer against GM's I have no idea how it would perform. The
>>function searches for complexity on the board.
>>
>>[Attractiveness = 125] simply is a multiplier of all positional components of a
>>position. A value of 125 will multiply the evaluation with 1.25
>
>This merely multiplies the values of the components (presumably only those that
>apply)? Sounds like an odd idea. Basically it would only serve to exacerbate the
>relative importance of the elements, no? I mean for example:
>
>Attractiveness = 100:
>
>Passed pawn = 0.30
>Doubled pawns = 0.20
>knight on outpost on f5 = 0.16
>
>and if the attractiveness is increased to 150:
>
>Passed pawn = 0.45
>doubled pawns = 0.30
>Knight on outpost = 0.24
>
>In other words it will favor pushing the passed pawn as it increases in value
>more than others. Much more than it would normally. Does this really help at
>all? If the values are off then they should be readjusted, but here it seems it
>will distort the overall picture, serving to solve some positions quicker at the
>cost of messing up more of the others.
>
>                                       Albert


The function is (was) useful for me to test (pawn) sacrifices. You will also
notice that increasing the value (101-130) will result in more attractive play
without losing much strength.

Ed



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.