Author: Bruce Moreland
Date: 11:02:14 01/22/98
Go up one level in this thread
On January 22, 1998 at 13:06:12, Ernst A. Heinz wrote: >Good point -- I will let "DarkThought" compute thsi problem over >night on one of our 600MHz Alpha-21164a LX164 machines. > >BTW, what does "Ferret" say about the positions raised so far as >inconsistent or wrong? I've posted a couple of disagreements already, but I can't really make a more organized effort now, since my machines are busy. Normally I run test suites overnight, and they either sit more or less idle during the day, or I put the Alpha on ICC. But as of this moment I have 25 versions I am trying to choose between, so I'm running test suites 24 hours per day on all of them. I did run this particular batch at a few minutes per move (I picked one of those versions at random), and agreed with what you guys have been saying, although this ... e4 one is still interesting. A try might be 1. .. e4 2. Qc4 d5 3. Qb5 Rf3, and when I played out to this position earlier, white was having some problems. There's potential for an interesting sacrifice involving Rxg3+, because the e2 rook is undefended. Without the original book this is all a little harder. For instance, position 31, the point of the problem was obviously to illustrate that the queen attacks d5 indirectly (cross-checks) if white's back rank is cleared. That is why Nxd5 works, and that is why Nxd5 is the "solution" to the problem. The last sane version of Ferret found that move in 0.35 seconds and showed it as leading to an advantage of +1 within 0.45 seconds. So obviously that problem is easy. It would be bad to over-search this problem in order to find other ways to win the pawn at 19 plies or something, or try to argue about shades of positional advantage when the pawn can be won two ways. Maybe the same thing is true of this ... e4 problem, the point might be to win a pawn or something, and this is refuted somehow, so we're going to go beyond it and try to find some other sort of forced win, which is a little strange. It is becoming aparent that the book was written for humans, in order to show them combinatorial techniques, rather than as a repository for difficult problems with unique solutions, to be used to test computer programs twenty years later. bruce
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.