Author: Heiner Marxen
Date: 07:43:24 06/13/03
Go up one level in this thread
On June 12, 2003 at 19:30:17, Lyn Harper wrote: >[D]7k/1p6/1P2p3/1P2P3/4P1p1/6P1/8/K7 w - - > > The only program I've had that will announce a mate in this position is > Ruffian, on an Athlon 2400. In these type of positions big hash tables come > into play, since only kings can move. Also note the black king can't get into > white half of the board, although the white king can get through. > I jacked up the hash to 200mb. >It's an exercise in related squares, from "How to Play the Chess Endings", I think it's mostly called "corresponding squares". > by Eugene Znosko-Borovsky. > Ruffian announced mate in 30,then went off the mate claim, only to come back >and stick to it at depth 60. That's quite a full width depth for a program. The >selective depth was 63. > As I remember, it took 8mins to announce the mate. >Future versions of Ruffian might not be free. Chest confirms the mate in 30: PV: Ka2 Kg8 Ka3 Kf8 Kb2 Kg7 Kb3 Kf7 Kb4 Ke8 Kc3 Kf8 Kc4 Ke8 Kd3 Kf7 Kd4 Ke7 Ke3 Kf7 Kf4 Kg6 Kxg4 Kh6 Kh4 Kg6 g4 Kh6 g5+ Kg6 Kg4 Kg7 Kh5 Kh7 g6+ Kg8 Kh4 Kh8 Kg4 Kg8 Kh5 Kg7 Kg5 Kg8 Kf6 Kh8 Kxe6 Kg7 Kf5 Kh8 e6 Kg7 e7 Kh6 e8=Q Kg7 Qf7+ Kh8 Qh7# (Athlon 1500+, 128 MB hash, 137 sec (2.3 min)) Interestingly, EGTBs (all 4, some 5) did not really help, but rather slowed it down by nearly 2 seconds. That may be due to the special logic Chest already has to handle an attacker with (blocked) pawns, only. The timing (over increasing depth) was sort of interesting, too: # 9 0.00s 0kN 1.57 35- 0 # 10 0.00s 0kN [ 2.87] 5.13 65- 0 # 11 0.00s 1kN [ 2.95] 7.27 188- 0 # 12 0.02s 3kN [ 2.73] 9.06 435- 0 # 13 0.04s [ 2.00] 10kN [ 3.45] 9.42 1190- 0 # 14 0.24s [ 6.00] 59kN [ 5.98] 8.54 5081- 0 # 15 0.53s [ 2.21] 129kN [ 2.18] 15.05 10526- 0 # 16 0.80s [ 1.51] 190kN [ 1.48] 25.47 14868- 0 # 17 0.88s [ 1.10] 216kN [ 1.14] 32.66 17240- 0 # 18 1.52s [ 1.73] 378kN [ 1.75] 30.45 29102- 0 # 19 5.67s [ 3.73] 1375kN [ 3.64] 22.53 89113- 11 # 20 49.28s [ 8.69] 11165kN [ 8.12] 22.53 591497- 226 # 21 82.83s [ 1.68] 18556kN [ 1.66] 31.36 1029369- 1046 # 22 92.49s [ 1.12] 20740kN [ 1.12] 40.63 1157853- 1605 # 23 121.45s [ 1.31] 27211kN [ 1.31] 47.96 1544032- 6935 # 24 136.98s [ 1.13] 30640kN [ 1.13] 58.46 1714222- 11629 # 25 136.98s [ 1.00] 30640kN [ 1.00] 67.50 1714233- 11629 # 26 136.98s [ 1.00] 30642kN [ 1.00] 81.14 1714351- 11633 # 27 136.99s [ 1.00] 30643kN [ 1.00] 92.99 1714370- 11634 # 28 136.99s [ 1.00] 30643kN [ 1.00] 105.85 1714397- 11636 # 29 136.99s [ 1.00] 30643kN [ 1.00] 113.42 1714398- 11636 # 30 137.00s [ 1.00] 30644kN [ 1.00] 125.93 1714456- 11640 The transposition table appears to be the most important single factor, here. Chest estimates a speed up factor of nearly 126... wow! I suspect that exact mate scores in the TT are important, here. Cheers, Heiner
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.