Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Crafty evals

Author: Miguel A. Ballicora

Date: 22:39:15 09/21/01

Go up one level in this thread


On September 22, 2001 at 00:52:06, Robert Hyatt wrote:

>On September 22, 2001 at 00:43:03, Robert Hyatt wrote:
>
>>On September 21, 2001 at 23:45:54, Miguel A. Ballicora wrote:
>>
>>>On September 21, 2001 at 23:01:29, Robert Hyatt wrote:
>>>
>>>>On September 21, 2001 at 17:01:35, Miguel A. Ballicora wrote:
>>>>
>>>>>On September 21, 2001 at 16:45:44, Robert Hyatt wrote:
>>>>>
>>>>>>On September 21, 2001 at 16:27:01, Gian-Carlo Pascutto wrote:
>>>>>>
>>>>>>>On September 21, 2001 at 14:57:20, Robert Hyatt wrote:
>>>>>>>
>>>>>>>>I don't mind that kind of mis-evaluation so much.  If anybody wins, white
>>>>>>>>wins.  What I would not like is to see that kind of evaluation and then lose
>>>>>>>>the game...
>>>>>>>
>>>>>>>The problem is when crafty prefers the 1.8 from this ending over the 1.7 from
>>>>>>>another ending that _is_ won for white.
>>>>>>>
>>>>>>>That'll cost you half a point.
>>>>>>>
>>>>>>>--
>>>>>>>GCP
>>>>>>
>>>>>>
>>>>>>I don't disagree.  But it is just as important to not lose when you can win
>>>>>>or draw.  The not drawing when you can win is another level of tuning.  I need
>>>>>>to study the ending because a pawn on the a and c file becomes very hard for
>>>>>>a lone king to stop...  it might have simply made a mistake and drawn a won
>>>>>>position in this case..  I didn't look at it very carefully (yet).
>>>>>
>>>>>It looks drawn. As you already know, the advantage of the outside passer is that
>>>>>you can sacrifice it and the king can go and eat up the opponent pawns on the
>>>>>other side. In this case it is impossible (that's the problem here). For
>>>>>instance, place the pawns in g6,h5,g4. How can the white king capture them? it
>>>>>can't because as soon as you capture g6 black plays h4 and promotes. On the
>>>>>other hand, two passers cannot beat a king on the Q side, they need the help of
>>>>>the wKing. however, the wK cannot go to the Q side because black promotes.
>>>>>
>>>>>Regards,
>>>>>Miguel
>>>>
>>>>
>>>>I don't believe it is quite that simple.  If the pawns are zugged on the
>>>>kingside, the two passers on the queenside can easily win.  This is the
>>>>attraction of "wild 7" on ICC, which is what the "mini game" is all about.
>>>
>>>I am not sure what you mean by zugged pawns. Anyway, white cannot capture the
>>>pawns in any case, so how can white win? two passers, if they are not separated
>>>enough, cannot win without king support.
>>>The only chance is a race, when white king rush to the Q side to support them
>>>letting a black pawn to promote.
>>>
>>>
>>>>In this zug-less position (at present) things aren't so clear, of course,
>>>>but
>>>>put the black pawns at g5 and h6 and the white pawn at g4, and this turns
>>>>into a simple white win...
>>>
>>>Black plays h5 and it is a draw. I think I did not understand.
>>>
>>>Regards,
>>>Miguel
>>
>>
>>You are assuming that a KQ vs KQPP is a draw?  IE after h5, I play gxh5 and I
>>promote first.

But I play g6xh5 first and you don't promote ;-)
I think we are talking about different things. You are setting a position
with only h6 and g5? then I have pawn less and has little to do with the game
itself.

>>
>>To understand the rp and bp vs a lone king, requires a bit of thought.  But
>>once you understand that this is all about zugzwang, you realize that if the
>>tempo is mine, and you have _no_ move to make but king moves, you can't stop
>>the two pawns.  Just visualize the two pawns side by side at a4/c4, with your
>>king at b6.  It is your move and you have to stop the two pawns.  If you play
>>Kc5, I play a5 and you can't take the c pawn.  If you play Ka5, I play c5 and
>>you can't take the a pawn.  So you retreat, I advance, and we reach this
>>position one rank further up.  It's a cute position that once caused a lot
>>of interest on ICC until most everyone figured out how to win (white starts
>>off with pawns at a2/b2/c2, king at d1.  Black has pawns at h7/g7/f7, king
>>at e8.  White moves and wins every time if he plays right.  One wrong move
>>
>
>Part of the above is right.  Part is wrong.  Too late at night to correct it
>now...  Depends on how far the pawns are advanced...

Of course, but it is not the case of this game!
Ok it is late for me too, I better go to sleep...

Miguel

>
>Bob
>
>
>
>and black draws or wins.
>>
>>So calling this a draw or a win or a loss is not so easy for a static eval.
>>If black has one "passing" move (wild 7 has white playing a4/c5/a5/c5 until
>>the black king stops the pawns, then using the b pawn to gain the necessary
>>tempo to win the game) then this doesn't work.  In the position I started with
>>above, if white has to move, either pawn move loses, as black can then capture
>>that pawn and catch the other before it can promote...
>>
>>That's why I said this was a hard one to evaluate without a lot of searching
>>or specific pattern analysis.  I don't want to include the specific knowledge
>>for the above analysis in Crafty as it would be slower, and it would hardly
>>ever be used...
>>
>>But I do plan to look at the position and see if the eval can be tweaked some
>>as it _does_ know that black has a majority.  I am probably incorrectly giving
>>credit to white for two "distant passers" when there is really only one.  And
>>black has a distant passer himself that it does know about...



This page took 0.01 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.