Author: Don Dailey
Date: 13:47:40 06/24/98
Go up one level in this thread
On June 24, 1998 at 16:01:13, Bruce Moreland wrote: > >On June 24, 1998 at 13:34:14, Ernst A. Heinz wrote: > >>I violently disagree with your "materialistic assessment". The programs that >>like any of the two sacrifices (Nxe6, Rxe6) seem to be happy to trade material >>for the attacking chances they see and score within their search horizon. In >>this particular case, the speculation turns out to be wrong -- so what? >> >>As for the "program P finds move X in time T and iteration Y" posts in >>general, you are right -- but sometimes its just nice to tell the world >>that your program solves interesting positions ... :-) > >I admit that I threw gasoline on this, and along with Bob also throwing gasoline >on it, we are probably bothering people. > >My criticisms aren't of Rebel, certainly, nor of speculative play. Bob seems to >have a problem with speculation against strong players, but I don't. I think >that computer chess is all about risks, even in a program that plays oatmeal >chess. We are constantly exposing ourselves to pruning and evaluation errors. >I don't think it takes that much more bravery to take risks that are visible on >the board. > >At the same time, I wonder what position these programs are trying to get to. >White is down a pawn, and immediately gets two pawns for a rook, which is a huge >sacrifice, but (in variations at least) a few moves later can pick up the b7 >bishop. This leaves white down the exchange for a pawn, which is not too much >different from the current pawn-down situation. Perhaps the scores are going up >somewhat because the programs are seeing that black gets ripped open as a result >of this trade, but that is hardly a cause for excitement. > >If this is really what they are seeing, then there isn't much speculation >happening here at all, is there? Maybe there is even a lot more that could be >going on, but if the programs aren't seeing it, why give them credit? > >When mine played this as white against Crafty (I forced Rxe6 fxe6 and went from >there), the material situation ended up the same as it started. Rxe6 is an >exciting move, but it doesn't appear that it is even a sacrifice materially, >it's not like these programs are seeing three pawns of positional compensation >or something. > >bruce My feelings on speculation: It should be a good thing if done well, it's all about percentages. Me and Larry have always been conservative about recognizing compensation for a pawn for instance, the idea being let's not take a chance on being wrong. But now I argue that we are STILL taking a chance on being wrong, if we don't recognize the value of an opponents sacrafice we will error in the opposite direction. So if you are speculative, it should not be wild speculation, but more in the category of "educated guess" and based on pragmatic considerations. If done really well you will occasionally be wrong in either direction. - Don
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.