Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Looking for "this type position" that make programs 2100 ELO

Author: Robert Hyatt

Date: 05:38:35 06/26/01

Go up one level in this thread


On June 25, 2001 at 17:58:25, Chris Carson wrote:

>On June 25, 2001 at 17:45:32, Sune Larsson wrote:
>
>>On June 25, 2001 at 17:30:13, Chris Carson wrote:
>>
>>>On June 25, 2001 at 17:21:56, Sune Larsson wrote:
>>>
>>>>On June 25, 2001 at 15:42:39, Chris Carson wrote:
>>>>
>>>>>It should be noted that FIDE sets everyones rating to 2000 for the first event.
>>>>>
>>>>>I am looking fot the "this type of position" that at 40/2 on 1Ghz machine or
>>>>>faster makes Fritz, Junior, Shredder, Rebel, or Tiger play like a 2100 or lower
>>>>>rated player.
>>>>>
>>>>>The "this type of position" must be a positon that leads to multiple wins for
>>>>>the player.  A single win will not work since program learning will avoid it the
>>>>>next time.  If the "this type of position" exists, the most of us (even me)
>>>>>should be able to beat the programs 50% or more over a stretch of games, say 20.
>>>>> We might even be able to get a string of wins going (like 5 or 6 in a row) once
>>>>>we have mastered the "this type of position".
>>>>>
>>>>>I am really serious about this.  I do not know of any "this type of position"
>>>>>and I have looked for a long time now.  Also most programmers and players would
>>>>>benefit from this knowledge.
>>>>>
>>>>>Blitz strategy does not count, only 40/2.
>>>>>
>>>>>Thanks in advance.
>>>>>
>>>>>Best Regards,
>>>>>Chris Carson
>>>>
>>>> Got a couple of wins vs different CM8-GM personalities with this
>>>> wellknown setup as black. Don't play it anymore. Repetition is boring.
>>>>
>>>> [D]r2q1rk1/pppnn1b1/3p2pp/3Pp3/2P1Pp2/2N3PP/PP1BQPB1/R4RK1 w - - 0 15
>>>>
>>>> I'll mail you a few games, though they were played in 40/40.
>>>>
>>>> Sune
>>>
>>>Thanks, I am only looking for 40/2 and for the programs I mentioned.  CM does
>>>not have learning does it?
>>>
>>>I will save the position for analysis and any others you send me, not sure how
>>>they will turn out.  :)
>>>
>>>Best Regards,
>>>Chris Carson
>>
>> The KI attack with black pawns on d6,e5,f4,g5 and rolling, causes big
>> problems for most programs. The theme is displayed in the three different
>> games I just sent to you. The plan for black is here more important than
>> the exact move order. White has to attack very actively on the queenside
>> and at the same time defend the kingside. Programs usually can't manage that.
>>
>> Sune
>
>Yes, the KIA is something I have success with, draws mostly.  Rebel and Fritz
>adapt quickly to it and my their performance goes back to normal after say 5
>games, unless I reset to default and just play what I have learned (they get no
>learning that way).  ;)
>
>A 2100 player can not keep up with the programs using the KIA in my experience.
>Same with the Stonewall and other Dutch variations.  I have also tried French
>and French in reverse, also English and different Indians.  Programs adapt to
>quickly and after a couple of games leave me in the dust.  If it takes a 2400
>player to maintain the advantage, then that is beyond my capabilities and other
>2100 players.  :)
>
>Best Regards,
>Chris Carson


I don't see how a program can "learn its way out" after 3-4 games.  With a book
that has millions of possible moves, there are many ways to play differently but
end up in the _same_ pit of snakes...

Book learning is not a cure-all.  It simply prevents repeating lost bad lines.
But it doesn't solve the kind of problem being discussed here.



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.