Author: Bob Durrett
Date: 14:30:09 02/04/04
Go up one level in this thread
On February 04, 2004 at 12:44:00, Randall Shane wrote: >On February 04, 2004 at 10:09:18, Bob Durrett wrote: > >>On February 04, 2004 at 09:57:37, Robert Hyatt wrote: >> >>>_any_ gambit starts off with you material down. Do you still play 'em? I >>>played the King's gambit, the Latvian, the Evans, the Goring, the Danish, you >>>name it. And in every game I started off a pawn (or more) down. :) >> >>Doesn't this mean that Crafty is improperly evaluating the positional factors in >>gambits? >> > >Can any program properly evaluate the positional factors in gambits? >If the evaluation is absolutely perfect, after all, you wouldn't need search. Bob H., your logic is impeccable. : ) Not being a chess programmer, I cannot be aware of the special considerations and difficulties which may be involved. Nevertheless, I do read chess books quite a bit. It seems to me, intuitively, that any improvements at all in ANY part of the software would be good things. Motherhood, maybe, but surely reasonable. It seems logical that sacrificing a pawn for ***something*** would be a good idea if the value of that something were roughly equal [in practice] to the value of a pawn. Presumably, that something would account at least in part for the success of the gambit. [Note that the something would not be psychological advantage unless one is psyching out the programmer somehow.] I do not know whether or not any program can assess the value of the compensation for the material sacrificed, and maybe programming the evaluation code to do this is impractical or not feasible within reason. If so, then it would not make sense to try. On the other hand . . . Suppose, for the sake of discussion, that it is possible and suppose some brilliant ["Einstein-like"] programmer actually did that. Surely, the resulting software would be no worse unless the code ate up a big chunk of the processor's time. Is there any FUNDAMENTAL reason why trying to do this would be a bad idea??? Generally, I feel that chess engines smart enough to recognize positional [or other non-material] advantages and disadvantages would do a better job of determining which continuations to pursue during the searching. [I speak not about the specific algorithms since I know not. : )] Recall an argument I presented here awhile back. Paul Morphy has been credited with recognizing the need to create the preconditions for an attack before committing to that attack. [For example, having a preponderance of material in the vicinity of the enemy king.] Surely the position evaluation code can check for presence/absence of such preconditions and use that information to decide whether or not to look for attacking lines. [Attacking moves may be different from "ordinary" moves.] Bob D.
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.