Author: Robert Hyatt
Date: 13:18:29 06/25/98
Go up one level in this thread
On June 25, 1998 at 14:04:47, Johanes Suhardjo wrote: >On June 18, 1998 at 13:35:47, Bruce Moreland wrote: > >> >>On June 18, 1998 at 06:28:48, Ernst A. Heinz wrote: >> >>>On June 18, 1998 at 04:16:15, blass uri wrote: >>>>On June 17, 1998 at 14:49:15, Vincent Diepeveen wrote: >>>> >>>>>black to move >>>>> >>>>>position white Kh3,Ra4,a2,f4,g3,h4 >>>>> black Kc4,Rc3,b4,e6,f5,g6 >>>>> black to move wins easily >>>>> >>>>>Believe it or not, diep managed to lost this as it didn't play Ra3, >>>>>it played Re3. >>>>>This is so simple, but i haven't worked at endgame yet. >>>> >>>>Fritz5 is not better in this position it needed more than 80 seconds >>>>to find Ra3 is winning >>>>before that(after 1 minute on my pentium 200MMX) it prefered Re3 >>>>It intended to play Ra3 without seeing it is winning after some >>>>seconds but change its mind >>> >>>That's really surprising because it actually is so simple. >>> >>>"DarkThought" fails high on 1. Ra3! in iteration #9 after less >>>than 1 sec. >> >>I don't know what is going on here either. Mine gets Ra3 in < 1 second on a >>P6/200, with a score of +2, has +4 in 6 seconds, and has +7 in 38 seconds. >> >>Is this because Fritz is doing null move in K+P endings or something? >> >>bruce > >Wow! How did you guys do it? It is a 9-ply sequence, isn't it? Do your >programs finish 9 plies in less than 1 second? My program on a 170 MHz >Sun Ultra 1 can't find it even after 11 plies in zillions of seconds. It >was too involve in enjoying itself with the lines after e5 fxe f4 etc. > > Johanes Suhardjo (johanes@farida.cc.nd.edu) >-- >Maybe you can't buy happiness, but these days you can certainly charge it. actually, e5 is also probably winning. e5, fxe5 f4 and white has tons of problems as he has to get the rook over to stop the epawn, which loses the apawn, so it's probably won that way as well. Ra3 is more forceful and is trivially winning. so it's better, because the rooks have to come off, and black's king is in the right place to win the apawn, and the game, with no real chance (or time) for white to conjure up anything on the other side of the board... but it's a non-trivial position to evaluate, because as black trades rooks and goes after the a-pawn, white has some interesting threats on the other side that make black take time out to handle them. IE ra3 Rxa3 bxa3 h4 and black has to take time out to plzy gxh4 or the white pawn queens first and loses. This happens in another couple of places. Would be interesting to see how a program spots this accurately with a shallow search. It takes me 12 plies to see Ra3 is +3. If the eval is around zero, it is not clear that a program knows what is happening. Bruce is doing some extension stuff that might well pick this up quickly, as Cray Blitz gets this right at 5 plies and 0 seconds, but the PV is quite long and has the horizon moves by white correct, and it realizes that black *has* to stop to capture each pawn that is offered, but it discovers that those moves are "singular" and it extends them quite deeply. I'm not yet doing singular extensions in crafty, so I don't catch this so quickly, but I'm starting to play with them again to see if I like the result.
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.