Author: martin fierz
Date: 16:17:08 09/16/02
Go up one level in this thread
On September 16, 2002 at 18:39:11, Dieter Buerssner wrote: >On September 16, 2002 at 17:32:09, martin fierz wrote: > >>On September 16, 2002 at 13:59:02, Dieter Buerssner wrote: >> >>>On September 15, 2002 at 20:25:46, martin fierz wrote: >>> >>>>> 8/8/4K1p1/4PpB1/3p1P1P/5n2/4k3/8 b - - >>>I agree, with almost everything. However, I think, that typically the method of >>>the chess engine works as well. Indeed, I am partitially using ideas you > ^^ >The "as" should not be there. oh :-) >>you are right about bf6 of course. i thought about it but you hadnt mentioned it >>in your previous post and so i just followed your mainline. > >Not, that it is important, but I hash shown it in the original post, and the >last position was to find Kxf4. > >>this is not the only >>winning path though, but it is very convincing. > >Which would be an alternate winning path, in the line with Bf6 ? not in your main line - the line with 1. ...d3 2. Kf7 Nd4 (full line in my previous post) is an easy win for black too, and could also be preferred, because in this line white does not get a passed pawn to the 7th rank, compared to the position with 1 ....Nxg5. as i say, in my human lookahead i am most certain that the KNPP-KB is a win (100%), and very certain that the 2 ...Nd4 line is a win (99%). as long as you don't go for the line in the game, both are ok. which makes me disagree with peter who says that the key is in the eval of the KNPP-KB ending - the idea of a big penalty for the pawn on the 7th should find the "other" winning line, and be ok too. >>i agree that i wouldnt feel too good about giving huge scores, like 5 in a +8 >>position. you could also try the other way round: my eval of the KNPP-KB >>position is "100% win", not +3. you could try to fix that one alternatively to >>fixing the position with the pawn on the 7th. > >BTW. From another recent stupid Yace game (you can find it by clicking on the >link in the original post of this threat), a KNPPKB position: > >[D] 8/8/3k4/8/5p2/2n5/pK2B3/8 w - - > >Well, Yace had more than +3 in that postion :-( I am using pawn = 0.8 yeah. of course, KNPP - KB is not always a win :-) when i look at this particular position i notice two additional features: 1. i have connected pawns and 2. my king is protecting my pawns. the 3rd feature, the white king is misplaced, is probably irrelevant given the first two - together they should ensure the win. without feature 2, you could construct a position where the white king is sitting in the pawns and will sac a bishop for the 2 pawns. with 1. & 2. combined, this is usually not possible. as you can see, your position would not fulfill my requirements, so i wouldnt assess it as a clear win. thinking about it, you could construct something like w: Kd5 Ba3, b: Kd7, Pd6,e5 Nh1 which is a draw. hmm. so perhaps you need feature 3 after all :-) or some form of knowledge that the knight can come to the rescue. anyway, it gets a bit complicated :-( >But, if Crafty didn't >have the more sophisticated knowledge about this type of endgames (and giving >sort of high positional scores, by making the eval 0), it would have never >blundered. It would have stayed with something like -3, but drawn the game. if you forget the exceptions of the exceptions, this can happen :-) i'd try to be very strict in implementing such stuff, like suggested above, if you include features 1&2&3 then you should be fine. at the very least, with this particular material balance, you could give a bonus for connected pawns. i still like the idea of pawn on 7th rank = +3 in positions where the total eval is >+7 or so (or +4 with >+8 or whatever). should make you find the 2...Nd4 line, right? aloha martin
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.