Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: On the Harry Schnapp book - intended primarily to Dieter Eberle

Author: GŁnther Simon

Date: 10:16:05 02/21/06

Go up one level in this thread


On February 21, 2006 at 13:05:25, Uri Blass wrote:

>On February 21, 2006 at 12:44:39, GŁnther Simon wrote:
>
>>On February 21, 2006 at 12:10:27, Uri Blass wrote:
>>
>>>On February 21, 2006 at 12:04:05, GŁnther Simon wrote:
>>>
>>>>On February 21, 2006 at 11:19:27, Uri Blass wrote:
>>>>
>>>>>On February 21, 2006 at 01:30:34, Gabor Szots wrote:
>>>>>
>>>>>>On February 20, 2006 at 16:19:48, Dieter Eberle wrote:
>>>>>>
>>>>>>>Hi Gabor,
>>>>>>>none of the lines you posted is bad. They may not be played every day by
>>>>>>>Kasparov but if you check them with a strong engine like Rybka beta13b you will
>>>>>>>find out that the maximal deviation from zero is far below +/- 1, a good result
>>>>>>>for a general book in my opinion.
>>>>>>>Astonished, Dieter
>>>>>>
>>>>>>Hi Dieter,
>>>>>>
>>>>>>I submitted all games with the HS book. I did not claim all the lines were bad.
>>>>>>I said the were bizarre. Most of them is too passive to be good. If an engine
>>>>>>comes out of the opening with -0.65 (as was the case in Spike vs. Pro Deo) it
>>>>>>will probably lose against an opponent of its own caliber.
>>>>>>
>>>>>>GŠbor
>>>>>
>>>>>I do not agree with it.
>>>>>
>>>>>-0.65 is not a good predictor of result in the opening.
>>>>>
>>>>>take the following position from the noomen match
>>>>>
>>>>>r2q1rk1/3nppbp/b2p1np1/2pP4/8/2N2NP1/PP2PPBP/R1BQ1RK1 w - - 0 11
>>>>>
>>>>>Glaurung say 0.65 pawns for white after a minute of search
>>>>>
>>>>>
>>>>
>>>>...snip... AFAIK we speak about -0.65 for _White_! after book end
>>>>and that is quite different and IMHO already downhill.
>>>>
>>>>Guenther
>>>
>>>I do not see the difference.
>>>
>>>If +0.65 for white can happen when white is not better then +0.65 for black also
>>>can happen when black is not better.
>>>
>>>Uri
>>
>>
>>Uri are you joking? Of course evals are not evenly distributed for
>>the colours, as White has an advantage for the first move, which
>>is even exaggerated by the evals of chess programs.
>
>Based on the match between 2 versions of Movei black won both games inspite of
>0.65 pawns advantage by glaurung for white when Movei was even more optimistic.
>
>My point is that 0.65 pawns advantage can be clearly misleading regardless of
>the question which side get it.
>
>I see no reason to assume that -0.65 for white means that
>white will probably lose against an opponent of its own caliber otherwise +0.65
>for white means that white probably wins against an opponent of its own caliber.
>
>Uri

I still disagree, but you don't have to believe it.
IMHO -0.65 W POV == around -1.20 B POV , because of an unbalanced scaling
around a fictional zero point somewhere at +0.50 - +0.60.
BTW we could calculate correlations between results and book out evals
from a bigger sample than posted before I guess.

Guenther



This page took 0.1 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.