Author: Hermano Ecuadoriano
Date: 06:47:18 12/11/00
Go up one level in this thread
On December 10, 2000 at 18:12:41, Jim Monaghan wrote: >The usual convention concerning chess style is to talk about a continuim with >strategy on one side and tactics on the other. Hartston in a K-K match book 1978 >deepens this concept by adding the concept of risk. So with S to T on the x axis >one can add low risk to high risk on the y axis one can develop a grid like so: > > >high risk Botvinnik Tal > Korchnoi Bronstein > > > > > > Capablanca Alekhine > Karpov Fischer >low risk Rebel Fritz > > > Strategy Tactics > >Essentially, risk means willing to plunge into the unknown either in a tactical >melee with compensating positional factors or choosing complex strategical lines >where both sides stand badly and where the only thing for sure is that the >better player will win .... the final outcome is not known. As oppossed to the >safety first group ... even the LT group above play fairly complex chess >pushing the limits beyond their opponents abilities but not necessarily their >own. > >Where am I going with this and how does it relate to computer chess? Well, I >think chess engines generally play it close to the x axis line. You have Rebel >that kills you softly placing it's pieces on slightly better squares, swapping >down. and eventually winning in a pawn ending by zugzwang on the 65th move. On >the other end you have Fritz annilihating you by winning a piece on move 16 with >the game shortly over. > >So I made up a little test suite of half dozen positions from the games of Tal >and Bronstein and ran it at 10 minutes per position with some engines: Amy, >Phalanx, and Yace. Well, the programs "failed" miserably. What have I proved ... >well that these engines don't play like Tal or Bronstein :-) However, Phalanx >showed a flash of inspiritation on the Kupper-Tal position and did choose 1 ... >Nxb2!? As Dusan Dobes says on his web page, his program plays like Tal .... well >sometimes!! > >Cheers, > >Jim > >PS: Test suite >2r3k1/pp4bp/3p2p1/3Ppb1n/1qr5/2N1B1PP/PP2QPB1/R1R3K1 b - - bm Nf4; id 1; >r2q1rk1/pb1pbpp1/np3n1p/1N6/2p5/P5N1/1PB2PPP/R1BQ1RK1 w - - bm Bxh6; id 2; >r3k2r/1b3ppp/pq1ppb2/8/np1NPPP1/2P2Q2/PPB1N2P/2KRR3 b kq - bm Nxb2; id 3; >rn2kb1r/1bq2p1p/p3pp2/P2P4/1pp5/2N2N2/1P2BPPP/R2Q1RK1 w kq - bm dxe6; id 4; >2r1r3/p1qbppkp/3b2p1/1pnB4/7B/4N3/PP1Q1PPP/3RR1K1 w - - bm Bxf7; id 5; >rn3rk1/ppq2p1p/1n4p1/3PP1B1/1bBQ4/2N2P2/PP5P/R3K1R1 w Q - bm Bb3; id 6; I like this subject. I think that as the programs get better, there will be more, or should be more, "philosophical" reviews. For example, we study the games Karpov-Kasparov and try to make inferences about the personality and psychology and, further, chess-philosophy of the two players. At the time, I was trying to answer a question like "How SHOULD chess REALLY be played?", which cannot be answered at this time, but might be interesting to humans indefinitely. Similarly, I like some programs whose rating might not be the highest. Thus, the enduring interest in such things. For the programmer, some more sophisticated parameters should be desirable. Just like "material" or "mobility", these parameters could be inferred from close study of GM games. My point is that these inferences gradually approach Chess-philosophy. The recent and ongoing "drive into the fog" debate is a foretaste of this phenomenon. It is not nonsensical. Though, of course, the individual philosopher could make it so.
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.