Author: Christophe Theron
Date: 18:39:15 02/19/01
Go up one level in this thread
On February 19, 2001 at 18:38:35, Robert Hyatt wrote:
>On February 19, 2001 at 18:17:58, Christophe Theron wrote:
>
>>On February 19, 2001 at 11:34:22, Robert Hyatt wrote:
>>
>>>On February 19, 2001 at 11:17:52, Chuck wrote:
>>>
>>>>On February 19, 2001 at 11:12:37, Christophe Theron wrote:
>>>>
>>>>>On February 19, 2001 at 07:44:30, John Wentworth wrote:
>>>>>
>>>>>>This is just an observance and may be completely wrong, but it seems that when a
>>>>>>program castles Queen side his chances of losing go up by a lot. Every time I
>>>>>>see a program do this, I say to myself he's going to lose and I bet more than
>>>>>>60% of the time he does. This may be a problem with humans vs humans as well, I
>>>>>>don't know. Anyone else notice this?
>>>>>
>>>>>
>>>>>I think you are right.
>>>>>
>>>>>
>>>>> Christophe
>>>>
>>>>I wonder if this a problem with queen-side castling or castling opposite (one
>>>>side castled short, one castled long)?
>>>>
>>>>Chuck
>>>
>>>
>>>It is a problem of the following:
>>>
>>>1. not recognizing king safety and the danger of a pawn storm until it is
>>>too late;
>>
>>
>>That's right. Human players are able to see very quickly that the position of
>>the king is unsafe, even if the pawn shelter is perfectly fine and there are no
>>open lines directed to the king. They know (by exerience) that the position of
>>the king is going to get really bad later.
>>
>>
>>
>>>2. not knowing how to attack the opponent, because when you castle to opposite
>>>sides, it becomes a race to see who draws the first blood. If you don't know
>>>how to break the position open (and I have not seen any programs do this very
>>>well) then while the program fiddles, Rome burns.
>>
>>
>>That's right, but on the other hand many programs already know that in case of
>>opposite castle it is important to start a pawn storm immediately.
>>
>>On the other hand many program are really good at launching a pawn storm that
>>end up in a blocked pawn position in front of the opponent's castle, which
>>actually protects the opponent's king from ANY attack. Really funny, and I would
>>not bet my own program is able to avoid this problem.
>>
>>
>>
>>>3. castling opposite is a direct challenge. Quite often the human will have
>>>his pieces positioned to support his attack, while the program's pieces are
>>>positions improperly to attack or defend. The time lost repositioning them
>>>leads to trouble.
>>
>>
>>Yes, absolutely.
>>
>>
>>
>> Christophe
>
>
>A good test is a position where white plays h6, black refuses to play g6 to
>create the hole at g6 for a white queen (mate), and then white plays hxg7 and
>is quite happy. While he just put a solid wall of granite up on the g-file
>since white can't capture his own pawn and black won't do it until the time is
>right.
Yes. I remember a long time ago I have seen that Genius (3 or 5) was able to
understand this. I have been very impressed (I have actually been impressed by
Genius at least a thousand of times).
However I still don't have anything for this in my own program. It does not
happen often enough to give me a good sample of test positions for this case.
> The problem I see is that some programs have huge scores for open files
>on the king. Even if the opponent can plant a bishop on that file and close it
>for as long as he wants. There are open files, and there are useful open files.
That's right. That's another challenge for programs trying to understand king
attacks.
Christophe
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.