Author: Mike S.
Date: 12:09:15 12/08/02
Go up one level in this thread
On December 08, 2002 at 08:34:16, Bob Durrett wrote: >I am interested in knowing whether or not modern chess engines recognize >advantages of these types. It is probably inavoidable, that the simple positions I tried to design are mixtures of various advantages and disadvantages. Hopefully the main element is the one in your question, each: >(1) If the White side has a substantial space advantage, but nothing else, will >the chess engine evaluate the positions with that advantage and give it a high >score? [D]rnbqkbnr/pppppppp/8/PPPPPPPP/8/8/8/RNBQKBNR w KQkq - 0 1 Analysis by Fritz 7: 1.h6 gxh6 2.Nf3 hxg5 3.Nxg5 Nh6 4.Bd3 +- (1.53) Depth: 5/32 00:00:00 93kN ± (1.31) Depth: 6/25 00:00:00 124kN (...) 1.h6 g6 2.e6 dxe6 3.Bb2 e5 4.fxg6 hxg6 5.h7 Nd7 6.c6 bxc6 7.hxg8Q +- (1.44) Depth: 10/39 00:00:18 5080kN +- (2.34) Depth: 11/43 00:00:48 14730kN Analysis by Chess Tiger 14.0: 1.h6 gxh6 2.Qh5 e6 3.fxe6 dxe6 4.d6 hxg5 5.Bxg5 Qd7 6.c6 +- (2.18) Depth: 8 00:00:00 55kN +- (1.66) Depth: 9 00:00:02 308kN (...) 1.h6 g6 2.fxg6 fxg6 3.e6 dxe6 4.Bb2 Nf6 5.d6 cxd6 6.cxd6 exd6 7.Bxf6 +- (2.62) Depth: 11 00:00:26 3734kN +- (3.06) Depth: 12 00:00:52 8330kN Interesting, that both start with 1.h6 and later come back to it, consindering some other moves in between (I snipped them to save place). The engines immediatly choose a move which weakens Black's king position. Analysis by Ruffian 1.0.1: 1.a6 ² (0.63) Depth: 1/3 00:00:00 1.Nd2 ± (0.72) Depth: 1/7 00:00:00 1.Nc3 e6 ± (0.80) Depth: 1/7 00:00:00 ± (0.71) Depth: 2/8 00:00:00 1.h6 d6 2.hxg7 Bxg7 3.f6 exf6 4.gxf6 ± (0.75) Depth: 2/8 00:00:00 ² (0.60) Depth: 3/10 00:00:00 1.Nf3 e6 2.fxe6 Bxc5 ² (0.64) Depth: 3/11 00:00:00 1.Qd4 e6 2.fxe6 dxe6 3.Bc4 exd5 4.Bxd5 Nd7 5.Qf4 Ndf6 ± (0.72) Depth: 3/11 00:00:00 ² (0.59) Depth: 9/25 00:00:21 6014kN 1.h6 a6 2.e6 dxe6 3.fxe6 fxe6 4.hxg7 Bxg7 5.Qh5+ Kf8 6.Ra3 Qxd5 7.Rf3+ Nf6 8.gxf6 Qxh5 9.Rxh5 exf6 ± (0.81) Depth: 9/26 00:00:35 10250kN +- (2.07) Depth: 11/34 00:02:02 39242kN >(2) If the White side has a substantial time (or development) advantage, but >nothing else, will the chess engine evaluate the positions with that advantage >and give it a high score? This was the starting position in a large experiment, published in the CSS magazine 4/2002: [D]rnbqkbnr/pppppppp/8/8/3PP3/2NB1N2/PPPBQPPP/R3K2R b KQkq - 0 7 In a tournament (double round robin) Black (!) managed to achieve a 20% score. - So, not even such a large developement advantage is a guarantee to win in computer chess. Analysis by Fritz 7: 7...Nc6 8.0-0-0 a6 9.d5 Nb4 10.e5 Nxd3+ 11.Qxd3 d6 12.Rhe1 dxe5 13.Nxe5 ± (1.31) Depth: 8/27 00:00:00 192kN ± (1.37) Depth: 12/39 00:00:47 14735kN 7...e6 8.0-0 Ne7 9.Bg5 h6 10.Bf4 a6 11.Rfe1 Nbc6 12.Qe3 Nb4 ± (1.34) Depth: 12/39 00:01:15 23547kN ± (1.34) Depth: 13/40 00:03:33 68509kN Analysis by Chess Tiger 14.0: 7...d6 8.e5 e6 9.Ng5 h6 10.Qh5 g6 ± (0.80) Depth: 6 00:00:00 54kN ± (0.90) Depth: 7 00:00:00 102kN 7...e6 8.Bg5 Be7 9.Bf4 d6 10.a3 Nc6 ± (0.88) Depth: 7 00:00:01 180kN 7...h6 8.d5 e6 9.e5 Na6 10.0-0 Nb4 11.Be4 c6 12.d6 f5 13.exf6 Nxf6 ± (0.82) Depth: 7 00:00:01 212kN ± (1.14) Depth: 12 00:00:41 4998kN Analysis by Ruffian 1.0.1: (...) 7...Nc6 8.0-0-0 d6 9.Ng5 Nf6 10.Bb5 ± (1.17) Depth: 4/7 00:00:00 ± (1.17) Depth: 6/13 00:00:00 29kN 7...d6 8.Bc4 Nc6 9.Ng5 e6 10.Qf3 Qf6 11.0-0-0 Qxf3 12.Nxf3 ± (1.15) Depth: 6/14 00:00:00 87kN ± (1.13) Depth: 8/19 00:00:03 846kN 7...Nc6 8.0-0-0 g6 9.Bc4 Bg7 10.Bg5 f6 11.Be3 f5 12.e5 ± (1.10) Depth: 8/19 00:00:04 1381kN ± (1.19) Depth: 10/25 00:00:17 5696kN 7...e6 8.0-0 Ne7 9.e5 Nec6 10.Qe4 Be7 11.Bf4 g5 12.Bd2 ± (1.17) Depth: 10/25 00:00:39 13788kN ± (1.17) Depth: 10/25 00:00:39 13788kN >(3) If the White side has a substantial pawn structure advantage, but nothing >else, will the chess engine evaluate the positions with that advantage and give >it a high score? It'S not unusual that the side which suffers from doubled pawns, has some compensation by (half-)open files... Therefore I removed the major pieces from the next position: [D]6k1/ppp1b1p1/2p1p1p1/3nPp2/3P4/1P3N2/PBP2PPP/6K1 w - - 0 1 Analysis by Fritz 7: 1.h3 Kf7 2.a3 a5 3.Kf1 c5 4.dxc5 Bxc5 5.c4 Nf4 6.Bc3 Bxa3 7.Bxa5 ² (0.66) Depth: 8/19 00:00:00 135kN ² (0.69) Depth: 12/25 00:00:10 3982kN 1.Kf1 Kf7 2.a3 b5 3.b4 g5 4.h3 g4 5.hxg4 fxg4 6.Nd2 g3 7.c4 Nb6 ± (0.72) Depth: 12/25 00:00:13 4967kN ² (0.69) Depth: 14/28 00:00:46 16579kN Analysis by Chess Tiger 14.0: 1.c4 Nf4 2.Ne1 c5 3.Kf1 cxd4 4.Bxd4 c5 5.Be3 g5 ² (0.46) Depth: 9 00:00:00 86kN ² (0.34) Depth: 9 00:00:00 106kN 1.Kf1 Kf7 2.Ne1 c5 3.dxc5 Bxc5 4.Nd3 Be7 5.c4 Nb4 6.Nxb4 Bxb4 ² (0.38) Depth: 9 00:00:00 168kN ² (0.40) Depth: 10 00:00:01 319kN 1.c4 Nf4 2.Ne1 Bb4 3.Nc2 Bf8 4.b4 Kf7 5.Kf1 Ke7 6.Ne3 g5 ² (0.42) Depth: 10 00:00:02 373kN ² (0.44) Depth: 12 00:00:04 799kN 1.Ne1 c5 2.dxc5 Bxc5 3.Nd3 Be7 4.c4 Nb4 5.Nxb4 Bxb4 6.Kf1 Kf7 7.Ke2 g5 8.a3 Bc5 9.b4 Bb6 ² (0.50) Depth: 12 00:00:10 1942kN ² (0.42) Depth: 16 00:01:27 16836kN Analysis by Ruffian 1.0.1: (...) 1.Kf1 Kf7 = (0.08) Depth: 1/2 00:00:00 = (-0.12) Depth: 2/3 00:00:00 1.c4 Nb4 2.a4 = (-0.06) Depth: 2/5 00:00:00 = (0.02) Depth: 3/6 00:00:00 1.Kf1 Kf7 2.Ne1 Bb4 3.Nd3 Bc3 4.Bxc3 Nxc3 5.a4 b5 6.a5 Ke7 7.f4 Ne4 8.Ke2 = (0.06) Depth: 3/6 00:00:00 = (0.01) Depth: 14/26 00:00:48 22835kN To judge about these results, it would be necessary have the engines play a few moves and see what the resulting pawn structure is (what is what they evaluate actually); I'm not sure if this above is a good test position... But at least Black has 2 doubled pawns and a backward pawn. It's not very easy to build "non-tactical" positions when they always jump to ply depth 10 immediatly :o) >(4) If the White side has a substantial material advantage, but nothing else, >will the chess engine evaluate the positions with that advantage and give it a >high score? [D]bn1rkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1 Analysis by Fritz 7: 1.Nc3 a6 2.e4 e6 3.d4 b5 4.Bd3 Bd6 5.Nf3 Kf8 6.Ne5 Bb4 +- (10.44) Depth: 8/19 00:00:00 86kN +- (10.34) Depth: 12/29 00:00:09 2463kN Analysis by Chess Tiger 14.0: 1.Nc3 c6 2.e4 e6 3.d4 d5 4.e5 b5 5.a3 Bb7 6.Bg5 Ne7 7.Nf3 +- (11.78) Depth: 8 00:00:00 57kN +- (11.80) Depth: 13 00:00:25 3256kN Analysis by Ruffian 1.0.1: (...) 1.e4 e5 +- (9.79) Depth: 1/1 00:00:00 +- (9.58) Depth: 2/4 00:00:00 1.d4 Nf6 2.Bf4 Nd5 3.Be5 +- (9.59) Depth: 2/4 00:00:00 +- (9.79) Depth: 5/9 00:00:00 1.Nc3 a6 2.d4 e6 3.e4 b5 4.Bg5 Be7 5.Nf3 b4 6.Bxe7 Nxe7 +- (9.88) Depth: 5/9 00:00:00 +- (10.00) Depth: 11/21 00:00:12 3842kN >(5) If the White side has a substantial initiative advantage, but nothing else, >will the chess engine evaluate the positions with that advantage and give it a >high score? For the illustration of initiative, a position would be good where one side has a series of - more or less strong - threats, and the opponent is forced to respond, preferably he'd have only one good move to defend... I can't design such a position in short time. Maybe opening experts know positions where that happens (2 or 3 threats in a row, with single-move defenses)? Initiative is an interesting topic. I can image that no programmer has a special evaluation for it. It seems to me, that it is a typical general human idea for something, engines handle at a "lower level" so to speak. IOW, if it's possible to continuosly play threats, resulting in more or less forced continuations, leading to an increased advantage, then engines will play like that and evaluate the goal they can achieve by that, no the initiative itself. Some engines have style settings, including aggressivity. One would expect that a higher aggressivity results in a style which is directed more at initiative than a "solid" style. Maybe engines play more threats then, even if they don't lead to advantages... (I never experimented with style settings)? Regards, Mike Scheidl
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.