Author: Heiner Marxen
Date: 13:51:34 02/26/00
Go up one level in this thread
On February 26, 2000 at 14:04:14, guy haworth wrote: > >I agree with Eugene's comments. But to take it further: > >Peter Karrer has worked with EN's help to do KQQKQP~ and KQPKQP~, salami-style, >one pawn-formation at a time, where ~ denotes a variant of chess allowing P=Q >only. > >These tables only required KQQKQQ + available 5-man tables for subgames. > >Peter showed that only 0.09% of KQQKQP(d2) positions change value of P=N is >allowed, and yes, there are positions that change from a win for White to a win >for Black. Nearly 1 out of 1100 sound a lot to me. >Hence, I expect practical o.t.b. players will accept information this near >perfect 'pro tem'. Maybe. I could not accept anything containing any error at all. When I do problem solving, I consider this to be a mathematically exact job. Hence, I cannot accept any approximations: they would cause incorrect answers, which is unusable (for me). >Therefore, the way forward is to do KQQQKQQ when a big enough 64-bit That is a surprise for me: does KQQQKQQ really have any practical impact? Does that occur in real games more than very very rarely? Heiner >architecture machine is available. It's a factor of 12 smaller than the table >for KQRNKQR (but that's still 5x bigger than KRNKNN for example) and Eugene's >index-economies work best with lots of Queens on the board. > >In its favour is the lack of symmetry: you can't see Black lasting that long >except that Black can capture into the maximal wtm KQQKQQ ending. So maximal >btm Black losses could be deeper than maximal wtm Wins. > >Once you have KQQQKQQ, you can do KQQQKQP~ as Peter did, ditching a factor of 4 >because of the P but winning a factor of 24 because of the fixed Pan position. >So that gets you to an ending smaller than KRNKNN which is do-able now. > >So KQQQKQQ is the next 'big one': Guy
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.