Author: blass uri
Date: 10:01:34 08/04/99
Go up one level in this thread
On August 04, 1999 at 12:23:21, Robert Hyatt wrote: >On August 04, 1999 at 11:55:03, KarinsDad wrote: > >>On August 04, 1999 at 11:02:46, Fernando Villegas wrote: >> >>>I am not a programmer, but a seasoned player and maybe I can hint to other way >>>to see this problem of the ?easy moves? on the ground of what a human player >>>thinks and do when faced with an obvious move to do. And what an experienced >>>player do is to discriminate between two different classes of ?obvious moves?: >>>normal ones and those to answer maybe-winning-sacrifices moves by the rival. >>>A normal one is just to recapture a piece that has made a normal capture before. >>>If my opponent take my bishop with a knight, so not giving nothing for free, I >>>just analyze a few ply in order to discover if that capture is or not part of a >>>mate combination, specially if that happens near my king. Mates combinations >>>beginning with a normal capture -by example, to kill a defensive piece before >>>launching the mate attack- normally are inside an horizon of no more -normally- >>>than 6 ply. So in these kind of cases -recapture after a ?normal? capture- that >>>should be enough in a couple of seconds of search for the program. >>>But then we have the second class of obvious moves, those that follow an >>>strange, no normal move. If my opponent play Bxh3+ , so losing at once material, >>>I tend to thinks that is part of a masive attack and so i will not recapture at >>>once, but take a deep look, in fact deeper than against any other normal move. A >>>program should do the same each time an aparent sacrifice is being performed. An >>>ad-hon culd be to consier the quealuty of he previous moves. I suppose that an >>>opponent that have played several best moves according he program is not a >>>beginner and so every unbalanced mve that he do shuld be consdered with the >>>utmost care. And vice versa. >>>So, easy moves should be treated depending of what class of capture is being >>>perfomed by the opponent. Is a balanced one in terms of material? Happens near >>>the king or the queen? Is not balaced but on the contrary lose material? >>>I Hope not to bother nobody with maybe an obvious analysis...:-) >>>Fernando >> >>Good point. However, I often see in GM games an in between move after a capture. >>The opponent HAS to reply to the in between move and then a move or several >>moves later, the first GM picks up the "normal obvious retake". I think it would >>be difficult to differentiate between many different types of forced moves >>without searching further (in the current paradigms). >> >>I think this area of chess programming is a non-trivial one which can only be >>resolved when programs start thinking more like humans and less like deep >>searchers. >> >>KarinsDad :) > > >I think intuition has a lot to do with a human's "that was easy" move. IE >you can intuit 'suspicion' (say) in a game like the one where Shirov played >(I believe) Bh3 giving up the bishop for a winning endgame. A computer might >think taking the bishop was obvious... but a human would think "wait a minute, >this is a GM, GM's don't hang pieces very often, so lets look a bit further to >see what is going on.." > >That "intuition" is a difficult thing to match, programming-wise... It is easy for computers to see that the evaluation of gxh3 is going down and this is a good reason that it should not think of gxh3 as an obvious move. If you have a real positive evaluation you expect it to go up and not to go down. Uri
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.