Author: Christophe Theron
Date: 20:23:17 01/17/01
Go up one level in this thread
On January 17, 2001 at 13:01:05, Robert Hyatt wrote:
>On January 17, 2001 at 07:15:27, Christophe Theron wrote:
>
>>On January 16, 2001 at 21:54:50, Robert Hyatt wrote:
>>
>>>On January 16, 2001 at 20:56:11, Christophe Theron wrote:
>>>
>>>>On January 16, 2001 at 13:37:11, Roy Eassa wrote:
>>>>
>>>>>On January 16, 2001 at 12:07:49, Robert Hyatt wrote:
>>>>>
>>>>>>On January 16, 2001 at 05:29:55, James wrote:
>>>>>>
>>>>>>>H4, h5 was so repeated that i wrote a counter < not difficult> to it, just in
>>>>>>>case the deep f boys try it on a slower but learning puter :)
>>>>>>
>>>>>>
>>>>>>If you use a different program, this likely won't work. It is based on
>>>>>>the idea that white will play h4-h5 and g4, while black plays h6 to prevent
>>>>>>white's h6. Black then blissfully castles kingside and after white plays
>>>>>>g5, the kingside is exploded with rooks and queens occupying the files.
>>>>>>
>>>>>>All black has to do is (a) not play h6 unless it is absolutely forced (the
>>>>>>old idea of "making luft" is a lemon) and(b) don't castle into a position
>>>>>>where you can _see_ that there are going to be open files around the king.
>>>>>>
>>>>>>Most programs will not do one or the other of those. Others won't do
>>>>>>either...
>>>>>
>>>>>
>>>>>Christophe,
>>>>>
>>>>>Are you reading this?
>>>>
>>>>
>>>>No I'm not. ;)
>>>>
>>>>
>>>>
>>>>> Will this be among the fixes coming out soon??
>>>>
>>>>
>>>>Honestly I don't know. This problem could be avoided in several different ways,
>>>>but at this time I don't know for sure what I am going to do, if I do something
>>>>in order to fix it.
>>>>
>>>>Maybe a book fix would do the trick.
>>>>
>>>>
>>>>
>>>> Christophe
>>>
>>>
>>>Take it from an "experienced hand" here. If you don't fix the _code_ so that
>>>it knows when to castle and when to not castle, then on the chess servers you
>>>are going to get beat on.
>>
>>
>>I think this is going to be improved in the next versions anyway, because with
>>Gambit Tiger I'm working harder on king safety and king attack.
>>
>>
>>
>>> I have said this before... "the chess servers are
>>>the absolutely most hostile environment you can find to develop and test a
>>>chess engine." If you have holes, they will get picked on. By _everybody_
>>>from GMs to IMs to other computer operators.
>>>
>>>That is why it is so helpful to debug there..
>>>
>>>if you fix the book but leave the hole, they will find another set of moves
>>>to create the same problem...
>>
>>
>>That's right.
>>
>>
>>
>> Christophe
>
>
>Another weapon you need for a chess server is the "position learning" I do
>in Crafty. You are getting zapped because you have no book line to help with
>the h4-h5 opening. If crafty would fall for that trap, it would only fall for
>it a couple of times, because it will remember when the score drops and put that
>into the permanent hash table. The next time the same opening is tried, it
>will vary well before the point where it saw the score crash...
>
>that is the reason I did this particular type of learning (it was first
>described by Slate and Scherzer in the JICCA a few years back)... it offers
>a great deal of protection against someone finding a very short or non-existent
>book line which makes book learning ineffective.
You are right, it is a nice feature. I'm thinking about it since a while, but
unfortunately I had higher priorities.
But I like the idea and I wish I can spend some time on it.
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.