Author: Robert Hyatt
Date: 08:42:57 11/07/02
Go up one level in this thread
On November 06, 2002 at 15:42:14, martin fierz wrote: >On November 06, 2002 at 14:51:50, Robert Hyatt wrote: > >>On November 06, 2002 at 13:58:24, martin fierz wrote: >> >>>On November 06, 2002 at 11:15:43, Robert Hyatt wrote: >>> >>>>On November 06, 2002 at 03:43:58, martin fierz wrote: >>>> >>>>>On November 06, 2002 at 00:02:57, Robert Hyatt wrote: >>>>> >>>>>>On November 05, 2002 at 14:17:55, martin fierz wrote: >>>>>> >>>>>>>On November 05, 2002 at 01:18:07, Jouni Uski wrote: >>>>>>> >>>>>>>>There is interesting and quite long analysis about game 2. which Kramnik won. >>>>>>>>They claim, that after whites 35. Rc5, there is no need for black to lose: >>>>>>>> >>>>>>>>[D]r7/3k1ppp/8/p1R1p3/Pp2P3/5PP1/1P5P/4K3 b - - >>>>>>>> >>>>>>>>According to analysis 35.-Rc8 draws! Really? >>>>>>> >>>>>>>Rc8 is obviously the best move in this position. shortly after the game i posted >>>>>>>this position asking if any program can find this move - only bob answered >>>>>>>saying that if a position is lost it doesnt matter what move you make (hmm, >>>>>>>sounds rather stupid to me...). >>>>>> >>>>>> >>>>>>Sorry to sound "stupid" but my point was that trying to find a "good" move >>>>>>after the game is lost is not very interesting. Much better to try to find >>>>>>a good move _prior_ to things going south. >>>>>> >>>>>>Rc8 might be good enough to draw. I'm not so sure being a pawn down with an >>>>>>active rook is that great if your opponent has a rook that is causing problems >>>>>>as well. I would rather try to find something _earlier_ in the game... >>>>> >>>>>that still sounds a bit stupid to me :-) every chess player gets into positions >>>>>where he wishes he had done something different earlier in the game. but then >>>>>the right reaction is not to kick yourself for the earlier mistake, but instead >>>>>to fight as well as you still can and make things hard for your opponent. >>>>> >>>>>you can also just see it as a test position which has no reference to what >>>>>happened earlier in the game. black to move and fight for the draw! the >>>>>discussion then was whether kramnik was playing specifically anti-fritz or not, >>>>>i.e. whether it was really such a big help that he got the *exact* copy before >>>>>the match. i still dont't think so, and this position is one of the reasons. >>>>>which computer will play Rc8 here? my guess: not one. which means that kramnik's >>>>>strategy (as seen in games 1-4, but no more later) would have worked against >>>>>most other computers too... >>>>> >>>>>aloha >>>>> martin >>>> >>>> >>>>I suspect it _would_ have worked. I'm not particularly worried by this >>>>position, because >>>>it requires dumping a pawn for a little activity, which is not going to be easy >>>>to implement. >>> >>>this is really a standard idea in rook endings. it's not like this is a "one in >>>a million" position... >>> >>>>The work spent on getting a program to do that in the right way will be _far_ >>>>greater than >>>>the work required to avoid the entire sub-tree earlier in the game, which was my >>>>point. This >>>>is about "economy of effort" more than anything else. >>> >>>which may be right for this particular game. but rook endings are the most >>>common endings, and this is a very common theme in rook endings. >>> >>>aloha >>> martin >> >> >>It's a common theme, but it is also very "special-case". For every position >>where giving >>up a pawn for activity is right, there are cases where it is also wrong. >that may be true, but then the challenge is obviously to find an evaluation >function which knows which case is which... > >>That's the problem I was talking about. And once you choose to give up the >>pawn, you are >>committed, right or wrong. Until then you have options. > >not true. in this position, after ...Rc8 Rxsome pawn Rc2, the white king is >trapped on the back rank. which is surely a big help for the defending side. you >cannot wait until white has activated his king - you don't have the option of >saccing that pawn any time you like. > >aloha > martin Of course not, but at that point I have a choice. After I sac I have no choice whatsoever... If the sac loses more times than it wins, then it turns out to be bad. This is one of many such cases where a human's intuition and experience will help, but trying to transform that into an evaluation function is nowhere near easy.
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.