Author: GuyHaworth
Date: 13:39:03 10/31/02
Go up one level in this thread
[D]5R2/8/8/5n1R/4k1rP/8/5K2/8 b - - 0 1
http://cm.bell-labs.com/cm/cs/who/ken/chesseg.html has this as a draw.
[D]8/8/2R5/1R6/4rn2/8/8/k4K2 b - - 0 1
This is a win, as Rob Hyatt said, in (DTC) 9 moves. The DTC-minimaxing line I
chose differs from RH's but the bR is similarly captured on m10:
... my ' = Rob's ! I guess ... "optimal move", but mine is DTC-optimal
... DTC-equi-optimals are in square brackets
1...Ra4' 2.Ke1' Ra7 [Ra8] 3.Kd2 [Rd6] Ka2' 4.Kc2 [Rc4, Re5] Ka1'
5.Rb2 [Rb4, Rc4] Nd5 [5 options] 6.Kc1' Nf4' 7.Rb1+ [Rc3] Ka2 {only legal}
8. Rc2' ka3 {only legal} 9.Ra1' Kb~ 10.Rxa7 1-0
The raw statistics for KRNKNN are available if misleading, as John Nunn points
out in the latest edition of 'Secrets of Pawnless Endings'. More telling is the
way the %-wins change as the wins in 1, 2, 3 etc are discounted from the
population: there is usually a peak of 'wins in 1'
However, 'raw', the stats are:
wtm/btm draws: 29.92% / 78.27%
wtm/btm 1-0s: 70.08% / 21.72%
wtm/btm 0-1s: epsilon% / 0.01% ... 0.005 > epsilon > 0
maxDTC depth = 243; maxDTM depth = 263
average DTM depths:
wtm/btm 1-0s: 25.50m / 39.35m
wtm/btm 0-1s: 0.09m / 1.50m
maxDTM/averageDTM:
wtm/btm 1-0s: 10.28 / 6.63
wtm/btm 0-1s: 68.07 / 4.66
'presence' [a home-grown measure to avoid extremes of maxDTX and '%-wins']:
... = average (depth * #-won-at-that-depth)/(total number of positions)
wtm/btm 1-0s: 1786.83 / 854.73
wtm/btm 0-1s: epsilon / 0.02
A review of Ken Thompson's KRNKNN DTM EGT results, produced independently and at
the same time as Eugene Nalimov's, appears in:
Haworth, G.McC. (2000) Deepest Chess Win Revisited. ICGA_J v23.2 pp.94-96.
and was prompted by some observations by John Roycroft. This includes some
distributions of (1-0) wins by depth in two log-scale (essential) graphs.
g
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.