Author: scott farrell
Date: 06:40:47 12/04/02
Go up one level in this thread
On December 03, 2002 at 13:44:21, Colin Frayn wrote: >Here's an interedting position that arose in an ICC game between Beowulf and a >human player. it's one of those positions where Beowulf knows it's behind, but >can't possibly realise how bad his position is because he doesn't realise how >his h7 rook is utterly trapped. > >[D]r4k2/2R2p1r/4pBp1/4P2p/pR3P1P/Pb2K1P1/8/8 b - - > >I'm not sure how to get round this without adding in some hideously slow code. >I already have mobility code in for the rook but then it's difficult to >distinguish between a rook that's in a nice defensive position that it could get >out of and which will transform itself into an open file fairly quickly after a >pawn exchange, and something like this where there is no sequence of moves to >retrieve the rook from its corner. > >Cheers, >Col Here is some unconventional wisdom (?) I have thought about this sort of idea quite a lot. It appears to me that this rook problem, is one of a whole class of similar problems. One I saw was in Kramnik vs Fritz games of recent is where the fritz rook was not trapped or pinned, but couldnt move away, as it had to defend a pawn, but effectively trapped based on human observation. Kramnik has such a decisive win from here, it was amusing to watch. It occured to me, that a search SHOULD be able to see this, my logic goes like this: 1) do a normal search 2) find how many rook moves result in some sort of failure (null-move, futility, etc etc) 3) score that piece down if there are a too large a penalty for moving that piece There are HUGE practical problems obviously: 4) everyone scores quiet positions 5) this requires scoring pieces through a dynamic sequence, and attributing the score back to a piece on a board towards the root somewhere 6) the need to recognize the rook (or other piece) as being 'trapped' and somehow tagging this is NOT quiet, and doing a 'trapped' search as opposed to a 'qsearch' now if I could only code as fast as I can think after a few beers with the lads. Scott
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.