Author: KarinsDad
Date: 06:35:41 01/26/99
Go up one level in this thread
On January 25, 1999 at 23:20:28, Dann Corbit wrote: >On January 25, 1999 at 22:23:33, KarinsDad wrote: > >>On January 25, 1999 at 21:51:35, Dann Corbit wrote: >> >>>On January 25, 1999 at 19:20:03, KarinsDad wrote: >>> >>>>I am looking for games and logs of computer vs. computer where the score went >>>>slowly in one direction and then quickly changed heavily in the opposite >>>>direction (i.e. one program appeared to have a winning advantage and then >>>>blundered into a bad position, the analysis of which was beyond it's event >>>>horizon). >>>Can you quantify that? A swing of -(?) in eval? One half pawn? One pawn? A >>>rook... >>>[snip] >> >>I think a swing of say +1 to -2 (or even +2 to -1) where effectively a minor >>piece is lost, give or take. I'm not picky, I just want to see if my program >>will detect problems of this nature (pro or con) before they occur. >Here is an example where reanalysis changed the ce by amounts up to 10. Of the >same position. Using the same program. At similar time controls. Shrug... >previous ce =-66, ce = 917, previous pv = Kc1 Be3+ Kc2 e5 d6 e4 d7 Bg5 Ng2 Bf6 >Nf4 e3 Nd5 e2 Nxf6 e1=Q d8=Q Qe2+ Kc1 Qc2+ Kxc2, pv = Kc1 Be3+ Kc2 e5 d6 e4 d7 >Bg5 Ng2 Bd8 Ne3 Bb6 Nc4 e3 Nxb6 e2 d8=Q e1=N+ Kc3 Nd3 Kxd3 Kb1, epd = >8/8/4p3/2bP4/8/8/p7/k2KN3 w >previous ce =-66, ce = 533, previous pv = Kd7 Rf4 e7 Rxe4 e8=Q Rxe8 Kxe8 Kc2 Kf8 >Kd2 Kf7 Kc1 Kg8 Kb1 Kh8 Ka1 Kg7, pv = Kd7 Rf4 e5 Re4 Kd6 Rd4+ Kc6 Re4 Kd5 Ra4 e7 >Ra8 Ke6 Kc2 Kf7 Ra7 Kf8 Ra1 e8=Q Rf1+ Qf7 Re1, epd = 8/8/3KPr2/8/4P3/8/8/1k6 w - >- >previous ce =-66, ce = 350, previous pv = Kh2 Qe2 Rc1 Qxf2+ Kh3 Qe2 c8=Q Qg4+ >Kh2 Qe2+ Kg1 Qxe3+ Kh2 Qe2+ Kh3 Qg4+ Kg2 Qf3+ Kh3 Qg4+, pv = Kh2 Qe2 Rc1 Qxf2+ >Kh3 Qf3 c8=Q Qg4+ Kh2 Qe2+ Kg1 Qxe3+ Kh2 Qe2+ Kh3 Qg4+ Kg2 Qf3+ Kh3, epd = >8/2P3p1/6kp/2R2p2/3Pp2P >previous ce =344, ce = -15, previous pv = Rc5 Qd1 c7 Qf3+ Kg1 Qd1+ Kh2 Qe2 Rc1 >Qxf2+ Kh3 Qe2 c8=Q Qg4+ Kh2 Qe2+ Kg1 Qxe3+ Kh1 Qf3+ Kh2 Qe2+ Kh3 Qg4+ Kg2 h5, pv >= Rc5 Qd1 c7 Qf3+ Kg1 Qd1+ Kh2 Qe2 Rxf5 Qc2 Rf4 h5 d5, epd = >8/6p1/2P3kp/3R1p2/3Pp2P/4P1P1/5PK1 >previous ce =-244, ce = 68, previous pv = a4 Bxa4 Nc3 Qxc3+ Kxa4 cxb5+ Kxb5 Qb3+ >Ka5 Qb6+ Ka4 Rxc5 Qxc5 Qxc5 Rhc1 Qd4+ Kb5 Qb2+ Kc6 Qxh2 Re1, pv = a4 cxb5 Qc1 >Qxa1 Qxc2 Rxc5 Qd1 Rc8 h3 bxa4 Kb4 Rb8+ Kc5, epd = >6k1/p4ppp/2p5/KBP1r3/8/5P2/Pqb4P/RN4QR w - - >previous ce =210, ce = -66, previous pv = Ne8 Rf1+ Ke4 Re1+ Kd5 Rd1+ Kc6 g1=Q >Rg7+ Kf5 Rxg1 Rxg1 Nc7 Rg7 e6 Rxc7+ Bxc7 Rxa7, pv = Ne8 Rf1+ Ke4 Rf4+ Kd5 Rg4 >Rg7+ Kf5 Rf7+ Kg6 Rg7+ Kf5, epd = 1B6/P3R3/5N2/4P1k1/2P5/5K2/r5p1/6r1 w - - >previous ce =157, ce = -66, previous pv = Re3+ Kf4 Rb3+ Ke4 Bg2+ Kf5 Rxb2 Rxc7 >Bc3 Rd1 Bxd4 Rxc6 Bxc6 Rxd4 Rxa2, pv = Re3+ Kf4 Ra3+ Ke4 Re3+, epd = >5k2/ppp4R/2n2N2/8/3PK3/3r3b/PB1b3P/R7 b - - >previous ce =-162, ce = 54, previous pv = Bc4 Qd6 a3 Ba5 Qh5 Re7 Kg1 Nf6 Qg5 Ne4 >Qf5, pv = Bc1 Nf4+ gxf4 Qg6+ Kh1 Re7 Bb2 Nxd4 Bf3 c5 a3 Ne6 Qe2, epd = >3r2k1/ppp2ppp/2n4q/3n4/1b1P4/4r1P1/PB2BPKP/RN1Q1R2 w - - >previous ce =-66, ce = 146, previous pv = Kf5 Ke3 a5 Kd4 Kf4 Bd5 h5 Kc5 b6 axb6 >h6 Bg8 axb6 Kxb6 Ke5 Bh7 Kf6 Be4, pv = Kf5 Ke3 a5 Kd4 Kf4 Bd5 b6 Kc5 bxa7 Kd6 h5 >Ke6 Kg5 Kf7 h6 Kg8 Kf4 Kh7 Ke5, epd = 8/p7/8/1P2K3/P6P/5b2/5k2/8 w - - >previous ce =-110, ce = 71, previous pv = Kf6 a4 a6 a5 h6 Qc2 h5, pv = Kf6 a3 a6 >a4 a5, epd = 8/p6p/2p5/2Pp1Pk1/3P2n1/5pP1/P2Q1P2/4KBq1 b - - >previous ce =99, ce = -53, previous pv = c6 Nc7 Qb6 Na8 Qb3 Qc7 a7 Rbc8 Rc1 Bf8 >Rb1 Bd6 bxc8=Q Rxc8, pv = c6 Nc7 Qb6 Na8 Qb3 Rbd8 a7 Qc7 Bc3 Nb6 Rb1 Qxc6 Qxb6, >epd = 1r2r1k1/1P2qpb1/P3nnb1/1QPp2p1/3Pp1P1/4P3/1B1N1P2/R3K2R w KQ - >previous ce =-66, ce = 79, previous pv = Rxd3 Bxd3 Bb3 Rc1 Qd8 Rc3 Be6 Bc2 Kf8 >Kf1 Qh4 Bd3 Kg8 a4 Kf8, pv = Rxd3 Bxd3 Qd8 Rd1 h5 Bc2 Qh4 Bc3 Kh7 Rb1 Bc4 Bb3 >Bd3 Rd1 Be2, epd = 6k1/pp3pbp/2p1b1p1/2P1p1q1/1P2P3/P2QPR1P/1B1r2P1/R4BK1 b - - >previous ce =-110, ce = 33, previous pv = Qe7 g3 Bf8 Bd3 Bg7 Kh2 Be6 Kg2 Kf8 g4 >Qg5 Kf2 b6 cxb6 axb6, pv = Qe7 Rb2 Qe6 Rbf2 b6 Kh1 Qe8 cxb6 axb6 a4 Bg5 a5 Bh4 >axb6 Bxf2 Rxf2, epd = 6k1/pp3p1p/2p3pb/2P1p3/1P2P2q/PbB1PR1P/3R2P1/5BK1 b - - > > >I have an EPD test suite that may suit you. If you get all the answers right at >any time control, I guarantee you a world-beater. I should be so lucky. Thanks Dann! One of my theories is that both computers and humans accidentally walk into shark infested waters (i.e. bad positions) without seeing the danger signals. Humans can see the danger signals more readily than computers and I'm trying to figure out why that is (other than the obvious chunking of patterns). Hopefully, some of these positions will give me clues in and of themselves to determine the answer to some of that. In other words, what it is about the position that determines that you are heading towards a cliff. Thanks again :) KarinsDad
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.