ditzel@ssc-vax.UUCP (Charles L Ditzel) (07/20/83)
Off on another tangent: The notion of a computer beating the world champion is still in my mind a long ways off. It is not the ability to calculate tactics that I wonder about. This function I feel is straight foreward. It is the more delicate process of evaluating a position that poses the roadblock. In chess the various styles/schools of play pose fairly complex problems. Life is easiest when two classical players meet (though even these type of problems will be difficult) but when a classical and a modern player meet the goals have shifted. Take for example the Grunfeld defense in queen pawn opening. The classical player has resolute and unshakable faith in grabbing the center as white and wondering despite his/her preponderance of central pawns whether the modern player's less direct central presence will win out (see Botvinnik-Fischer from MY 60 MEMORABLE GAMES & Gligoric-Portisch, Wijk aan Zee 1976(?)). The subtly of chess logic is not easy to pin down (if it were, there would be a lot more chess masters) "procedures" must be carefully chosen by the climate of the position and often the choice may come down to the school of play a player subscribes to. At one of the ACM conventions (it was in Seattle) I watched Yasser Seirewan (he was a young master then - since he has become a International Grandmaster and has been one of the few people to defeat Anatoly Karpov, the current World Champion) play a number of **speed games** against Chess 4.6(or was it 4.7 -in any cases it was a chess program tied to a developmental CYBER at CDC in Minnesota). Seirawan's first game was deeply tactical (all sorts of pieces were en pris) and HE LOST! There were more than a few suprised people. From the second game on Yasser played extremely positional games and won each one of them easily. In fact it was no contest. Tactical chess can easily be mastered by the software, positional chess is much more difficult and has yet (as far as I am concerned) to be demonstrated at even the A or B player level. It is going to be very difficult to create software that plays with Capablanca-like or even Petrosian-like judgement. If anyone out there is aware of master-level chess software I would be fascinated in knowing about it. cld