Author: Christophe Theron
Date: 10:27:00 08/10/98
Go up one level in this thread
On August 10, 1998 at 10:33:01, fca wrote:
>On August 10, 1998 at 08:36:36, Christophe Theron wrote:
>
>>On August 10, 1998 at 00:02:15, fca wrote:
>>
>>>On August 09, 1998 at 20:13:56, Christophe Theron wrote:
>>>
>>>>On August 09, 1998 at 19:18:52, Robert Hyatt wrote:
>>>>
>>>>>On August 09, 1998 at 10:22:31, Christophe Theron wrote:
>>>>>
>>>>>>Thorsten talked about this, but maybe it applied to older versions of Lang's
>>>>>>programs. Nobody has given evidence that Genius selects differently its moves
>>>>>>than the opponent's, except in the very end of the lines.
>
>>>>>actually, several posted a few positions where Genius appeared to show an
>>>>>asymmetric search. It couldn't find the key move with a very deep search,
>>>>>yet if you play the key move and let it play the other side, it would find that
>>>>>it was lost very quickly. Which lends credibility to the idea that it looks
>>>>>at everything for the opponent, but prunes (forward prunes) its own moves
>>>>>quite a bit in the right circumstances...
>>>>
>>>>
>>>>I remember Thorsten said he would post some of these positions, but he didn't do
>>>>it, or I just missed his posts.
>>>>
>>>>I would be interested to see such positions. If somebody has found such
>>>>examples, would he be kind enough to repost them, please?
>>>
>>>I think there is a misunderstanding here. By Bob, maybe by Thorsten and others
>>>too.
>>>Genius is highly "modularised." Depending on perceived game phase, a different
>>>evaluation module (code segment - not just a couple of variables) is used.
>>>But...
>>>Which module to use is only set at root level. Within, the same module stays
>>>used.
>>>So say the root move being considered would, if made, change perceived game
>>>phase (not necessarily a capture, but usually so, I believe). While it is
>>>examined - even to 32 ply depth - the lens used will still be for module x.
>>>However deep the search.
>>>
>>>Now the move is chosen. So module y is now used. Alas, the evaluation is
>>>significantly changed - i.e. Genius now "sees it" truly.
>>>
>>>This is the heart of the problem that is misdiagnosed as asymmetry (without
>>>denying or confirming that such asymmetry exists :-) ).
>>>
>>>Kind regards
>>>
>>>fca
>>
>>
>>What you are describing is sometimes called the "blemish effect", and is well
>>known by most of us, including Bob, Thorsten and me of course.
>>
>>Genius has indeed some trouble evaluating a queen exchange for example. It's not
>>the only one in this case! Fritz5 seems to be a little bit improved in this
>>regard.
>>
>>When Thorsten coined the idea that Genius was assymetrical, I think he wasn't
>>confused by the blemish effect, and really meant that some agressive pruning was
>>done for one side only.
>>
>>The problem is that it is not easy to show. In a normal game, Genius analyzes
>>only its own moves. To discover the assymetry stuff you have to replay the whole
>>game, ask Genius to analyze both sides, and watch carefully any score swing.
>
>Yes.
>
>>Did somebody try that?
>
>Yes. I remind you I said in the preceding post:
>"without denying or confirming that such asymmetry exists"
So you tried? What is your conclusion?
>>As you seem to have a close connection with Richard, maybe you could ask him?
>
>:-)
>
>>But I guess I already know the answer:
>> answer = "" ;
>
>Answer actually is:
>
>CbEFhPyMS/2x8P8Sqo/jCzbSKcoyxWIM8PgqoD45RiNtKAzTzYD+/gei41ytSw/V
>8RNbStX2xCrKJTlEGpe3UGbN4HKXVGDcTiNNKN/sKFpnvuwghhPI+HKAj8xB+joQ
>em0jM6GnCahRMGGJuJHWPN5JXLlPBLizzG2szlS5hx6SwGlhEJKSXaahK/+fANR3
>K9i2cUJ0J6uXWVKPi3vHYBNx9oXaNtTLKK4CusGWcWbTZHYAri8tHudzV0ssyNc7
>ANghB2VlrNKgZL8bAfOoD9pbD/AsJEwr99gR84ZiVM+lpBBzPNQbD30S/Ihmq5AV
>yc/fuotEfz8lWDoQY+lDpp88aPkbZxwt/hQIQ+uTf14smekjMa+P+0B1uXls
The longest answer I have seen from Richard since a long time. :)
>But PGP key might be hard to find. :-)
Does PGP mean "Proud Genius Programmer" ?
>:-))
>
>> Christophe :)
>
>Kind regards
>
>fca
>
>PS: Disassembling Genius would be instructive.... :-)))
I don't want to waste 1 month or more disassembling it when I am about to write
a better program by myself. :)
Not that I don't think Richard's code is full of marvels, but maybe I can do a
god job too? :)
But if you really want to help me, please ask Richard to send its sources to
ctheron@outremer.com. I promise I will have a close look to them. And I'll keep
them away from other eyes. Thanks a lot. :)
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.