Author: Vincent Diepeveen
Date: 15:48:07 03/17/04
Go up one level in this thread
On March 17, 2004 at 18:23:45, martin fierz wrote: >On March 17, 2004 at 10:41:00, Vincent Diepeveen wrote: > >[snip] > >>After Rxe6 it is very clear. White is dead won and the only question is whether >>you win it eyes closed or whether you can also win it when asleep. > >[snip] > >>If i take some 1800 rated kids from my chessclub i'm sure that after Rxe6 gets >>played, that they will win from me with white. If they have white and play Qh3 i >>still win from them. >> >>Rxe6 leads to either mate, some big material win, or a dead won endgame. >> >>Even deep study i didn't find anything to delay a quick loss from black. > >i'll admit to this: i didn't study this position. i looked at the forced endgame >position for 1 or 2 minutes. i see absolutely nothing clear here. white is >looking good, but exchange one rook and you can be in trouble. you know not to >do that, but 1800 kids? i don't think so! i'd bet anything that i'd beat 1800 >kids in that endgame (specially since they are kids, they play endgames >horribly...). perhaps this is winning for white, but i wouldn't judge it to be >winning clearly and easily. > >perhaps if i look at this a bit longer it will be clear to me why this is so >completely winning as you claim. perhaps not though :-) >in any case, this is deeper than just saying: i look at this endgame position >and it's clear to me that it's a win with my eyes closed/in my sleep. bs2830 was considered a few years ago an important testset. so it was misused by commercial programs to show how good they were. don't blame the programmers. blame the average RGCC/CCC reader. they just want the stuff solved. the time invested in all such testsets like bs2830 , bk, win at chess (the hard ones), larsen (the hard ones) and certain ECM positions (ecm is so big that it's hard to study all positions), all that time is real real big. Of course nowadays in 2004 testsets play a smaller role because even the biggest diehard fans who shouted that you didn't need a better evaluation function, just a piece square table and a 20 ply search, they are hard to find nowadays. There are amazingly some commercial programmers still shouting that though, but the effort they spent in evaluation function is way bigger than search nowadays. A few years ago this was different. I remember that i searched 5 ply with diep at my 486. That really *sucked*. I remember a 5 ply diep playing a 6 ply bionic. Everyone betted at bionic. Then there was a trivial position with a mate in 6. Diep found it. I was very proud. Those games from 1994 were such a simple level when you compare it with 2004. You could win *any* position with a piece down, or 2 pawns down. NO problem. Just profit at the right moment. I remember a testgame from diep 2 years later at my university against a program which never has hit the RGCC. Michiel Hochstenbach. In fact it outplayed diep. then came the endgame. and it started nullmoving !!!! Diep was allowed to run with a pawn. opponent nullmoved! of course when the pawn was at the 7th row suddenly a big shock: "he this pawn is going to promote!" So i won that game with diep. Tactics in those days, and testsets, were very relevant to improve your program. 5 ply is just not enough. I do not see them as such nowadays. >cheers > 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.