Author: Dann Corbit
Date: 08:10:09 02/15/99
In each of the following EPD positions, I took the 'best move' suggested by the test suite and a different 'best move' suggested by the program. For these choices, I made the move in question and then ran Crafty against the position for about 8 hours. Here are some early results. The acn are nonesense numbers since the counter has overflowed 2^31 positions. In each case, I describe the analysis as 1 ply more than the program performed, since we are analyzing the positions after performing the chosen move. Further, since we have made one ply forward, the more negative the score, the better the original estimate. LCT POS 13: 1r2r1k1/4ppbp/B5p1/3P4/pp1qPB2/5Q1P/n4PP1/4RRK1 w - - acd 17; acn -233119022; acs 30001; ce -83; bm Nxa2; pv Qg3 Ra8 Bb7 Nc3 Bxa8 Rxa8 Qh4 Re8 Bh6 Bxh6 Qxh6 e6 Kh2 Nxe4 Rd1 Qb2 dxe6 Rxe6; c1 "LCTII.POS.13 after Nxa2 -- white to move"; c0 "Barbero - Kouatly, Budapest 1987"; 3rr1k1/4ppbp/B5p1/3P4/pp1qPB2/2n2Q1P/P4PP1/4RRK1 w - - acd 17; acn -104698002; acs 30001; ce -50; bm Nxa2; pv Be3 Qf6 Qxf6 Bxf6 Bb7 Rf8 Bc6 Rc8 e5 Bxe5 Bh6 Bg7 Bxg7 Kxg7 Rxe7 a3 Ra1 Rfd8 Rd7 Rxd7 Bxd7; c1 "LCTII.POS.13 after Rbd8 -- white to move"; c0 "Barbero - Kouatly, Budapest 1987"; It appears after 18 ply from the original position (17 ply from the chosen move) that LCT and Crafty agree that the LCT II choice was better than the original Crafty choice. The margin is not overwhelming, 1/3 of a pawn. 2k5/p7/Pp1p1b2/1PPP1p2/5P1p/3K3P/5B2/8 b - - acd 24; acn 119679487; acs 30001; ce -91; bm c5; pv bxc5 Be1 Kc7 Ba5+ Kb8 Bd2 Kc8 Ke2 Bb2 Be1 Bf6 Ba5 Be7 Kd3 Bf6 Ke3 Be7 Kd2 Bf6 b6 axb6 Bxb6 c4 a7 c3+ Kd3 Kb7 Kc2; c1 "LCTII.FIN.08 after c5 -- black to move"; c0 "Polugaevsky - Unzicker, Kislovodsk 1972"; 2k5/p7/Pp1p1b2/1P1P1p2/2P2P1p/3K3P/8/4B3 b - - acd 25; acn -53936097; acs 30000; ce -193; bm c5; pv Kc7 Ke3 Kd7 Bb4 Kc7 Ke2 Bd4 Kf3 Ba1 Kf2 Bf6 Ke3 Ba1 Be1 Bf6 Kd3 Kc8 Bb4 Kc7 Bd2 Kb8 Ke3 Kc7 Bb4; c1 "LCTII.FIN.08 after Be1 -- black to move"; c0 "Polugaevsky - Unzicker, Kislovodsk 1972"; Here, with 25/26 plies considered, crafty's choice is ahead by a pawn. I wonder if LCT II might be wrong on this one... 3r1bk1/p4ppp/Qp6/4p3/1P1B4/Pq2P1P1/2r2P1P/R3R1K1 w - - acd 17; acn -104830512; acs 30001; ce -15; bm e5; pv Bxe5 Rdd2 Qf1 Qd5 Bd4 Qf3 Rec1 Qxf2+ Qxf2 Rxf2 Rxc2 Rxc2 Rf1 Ra2 Ra1 Rd2 Rc1 a5 Rc7 axb4 axb4 Bxb4 Bxb6; c1 "LCTII.POS.06 after e5 -- white to move"; c0 "Nimzowitsch - Capablanca, New York 1927"; 5bk1/p2r1ppp/Qp2p3/8/1P1B4/Pq2P1P1/2r2P1P/R3R1K1 w - - acd 17; acn -77195556; acs 30001; ce -26; bm e5; pv Rec1 f6 Rxc2 Qxc2 f4 Re7 Qb5 Qb3 Qe2 e5 fxe5 fxe5 Qb2 Qd5 Bc3 Qe4 Qe2 Rf7 Rd1; c1 "LCTII.POS.06 after Rd7 -- white to move"; c0 "Nimzowitsch - Capablanca, New York 1927"; Here, after 18 plies, the move crafty originally chose is again ahead, but only by .11 pawns. Hardly decisive. I would like to get expert opinions as to these positions, especially FIN.08. Is Crafty fooled by a Zugzwang position? Is something important overlooked? A few more positions will follow (5 total will be analyzed in detail, one of which will have two alternate 'best moves' considered).
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.