Author: Don Dailey
Date: 14:45:56 06/23/98
Go up one level in this thread
On June 23, 1998 at 14:36:49, Ed Schröder wrote:
>>Posted by blass uri on June 23, 1998 at 10:27:58:
>
>>your descreption in your homepage:
>
>>"In Other words Rebel(using anti-GM) does not need to know the sacrifice
>>is 100% correct for this position, Rebel smells the chance of an attack
>>and goes for it"
>>the example shows Rebel10(not using anti-GM) can miss Rxe6
>>assuming I do not have a long time
>>Uri
>
>Yes and no, it depends. The anti-GM style is not about tactics. The given
>position from my home page is just an example to show what anti-GM is able
>to. To pick a "good" example a tactical position is usually much more
>self-explaining than a quiet position.
>
>An example of a more quiet position...
>
>r2qkbnr/pp1n1ppp/2p5/4p3/3pP3/5BNP/PPPP1PP1/R1BQ1RK1 b
>
>BR .. .. BQ BK BB BN BR
>BP BP .. BN .. BP BP BP
>.. .. BP .. .. .. .. ..
>.. .. .. .. BP .. .. ..
>.. .. .. BP WP .. .. ..
>.. .. .. .. .. WB WN WP
>WP WP WP WP .. WP WP ..
>WR .. WB WQ .. WR WK ..
>
>01:26 09.36 0.00 g8-f6 d2-d3 d7-c5 f3-g4 d8-b6 (R10)
>00:48 09.36 0.07 d4-d3 b2-b3 d3-c2 d1-c2 d7-c5 (R10 + anti-GM)
>
>The above is a game fragment of one of Rebel's games at Aegon 1989.
>Rebel then played the (IMO) ugly 1..d3 Of course the move has advantages
>but personally I clearly prefer the quiet 1..Nf6 (do not attack while
>you are behind in development)
>
>Rebel since 1989 has always been in love with this 1..d3 move. Only after
>a deeper search Rebel switches to 1..Nf6 However Rebel10 using anti-GM
>will deepen its love for 1..d3 even more and never will switch to 1..Nf6
>as it fits in the principals of the anti-GM style.
>
>I guess I have to live with it. I am curious what other programs play.
>
>- Ed -
Hi Ed,
Can you tell me what the anti-GM principles are? Are they general
to anti-human or they aimed specifically at Grandmasters?
Here are a couple anti-human algorithms I have used, I don't know
how useful they are but they seem to help slightly:
1. Computer gets bonus for 1st and 2nd pawn exchanges (to get
active play)
2. Computer gets anti-trade-down bonus to encourage pieces to
stay on the board.
3. Computer gets penalty for pawn rams (pawns butting heads)
Both must be pretty small so as to not encourage stupid moves
but only give a general direction to "break ties."
I have not thought deeply about others. I am interested in yours
if you can share them, or others people may have.
I view all of these algorithms as very slightly weakening the play
of the computer, if I considered them optimum then the program
would always use these algorithms. The idea is to change the style
slightly to favor open positions again people.
- Don
P.S. Cilkchess seems to like Ngf6 in your example from 3 ply on.
It never considers d6. I don't consider this as having any
real significant and I would trust Rebel's positional feel
more than Cilkchess. It is unclear to me which is better,
I kind of like d6 myself! What actually happened in the
Aegon game? Did d6 prove to be a good move?
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.