Author: Mike S.
Date: 18:25:29 12/11/00
Go up one level in this thread
Junior could be a particular interesting example in this respect. Maybe we should distinguish between positional style and positional knowledge. In my tests, Junior 5 solved 6 out of 10 positional tests, but i.e. CST II 10 of 10. One of the difficulties here is, that in some situations, programs will find a "positional" solution just by calculation. In other words, they see the resulting material gain already, while a human may choose the same move purely based on positional knowledge. But I think we can expect that a program, which has this knowledge too, should at least find the move faster: [D]3r1bk1/p4ppp/Qp2p3/8/1P1B4/Pq2P1P1/2r2P1P/R3R1K1 b - - 0 26 26...e5! A positional pawn sacrifice (from Nimzovitch-Capablanca 1927), to double the rooks at the 2nd rank. I'm sure there are more difficult examples for this. But I would suggest to call this a positional performance anyway, no matter how the program achieves it. It's the performance that counts. While in the 1st of your 3 positions, white would have to reject Nxb6 and allow black to win an exchange for a pawn by QxQ hg e4 if I see it correctly (haven't analysed much), in the other 2 positions nothing is sacrificed and no offer declined. The solution moves have no "testing characteristics", which is a general problem with testing of this component of a program's strength. Regards, M.Scheidl
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.