Author: Howard Exner
Date: 05:00:59 09/04/98
Go up one level in this thread
On September 03, 1998 at 02:53:23, blass uri wrote: > >On September 01, 1998 at 07:17:44, Howard Exner wrote: > >>Here are some positions taken from CCL games that proved difficult for the >>computers to solve. They are 40/2 standard time control games played >>on seperate P200's. How do other programs navigate these positions? >> >>1. Hiarcs6 played f7f5? which is a blunder because Qd5+ would easily win. It >>seems that some positions that are easy for humans to solve can still pose >>problems for computers.Why would programs find this position hard to solve? >>8/5pk1/4p3/7Q/8/3q4/KP6/8 b - - id "Genius 5.0 - Hiarcs 6.0"; bm Qd5; > >Junior5(16 bit) has no problem to find Qd5+ with evaluation of more than 2 pawns >advantage for black. It found it in a very short time(I think less than a >second) It takes Rebel 8.0 0:50 to find Qd5+. Before that it wants to play pf5. >> >>2. Nimzo played the losing move, Kxb3 when Kxa4! would have been a routine draw. >>8/8/8/5P2/Pk5K/1N6/6b1/8 b - - id "Genius 5.0 - Nimzo 98"; am Kxb3; > >Junior5 needed 6 minutes and 21 seconds to find Kxa4 on my pentium200MMX >The evaluation was 2.01 pawns advantage for white. >I am not sure if it is a draw(If you have database of 5 pieces you can check it) Rebel 8 plays Kxa4 immediately and never wavers. It must have some knowledge for this one. The draw looks straight forward but a 5 piece database could confirm this for sure. >> >>3. This position illustrates the trapped piece theme. Rxa2, played by Genius >>5.0, should be avoided because of the game continuation ... >>1. ... Rxa2 2. Rxa2 Bc4 3.Qd2 Bxa2 4. b3 and the bishop on a2 was eventually >>lost. > >Junior5 did not think to play Rxa2 and prefered Nd7 Rebel 8 never considers Rxa2 either. > >>1q3rk1/p3b1pp/1n2p3/r2bPp2/1p6/4BN2/PPB1QPPP/R2R2K1 b - - id "Hiarcs 6.0 - >>Genius 5.0"; am Rxa2; >> >>4. Hiarcs played Ke2 but Kxe7 is white's only practical chance of winning. Pawns >>on both wings generally favour the Bishop over the Knight. Allowing opposite >>bishops to remain (even with knights on the board) is an error in judgement in >>this position. >>8/4bkp1/2N2n2/8/8/2pB3P/5PP1/5K2 w - - id "Hiarcs 6.0 - Nimzo 98; bm Nxe7; > >Junior prefered Bc4+ Kf8 Ke2 and I am not sure about the right move. This one I think will be difficult for most programs. They may not like Nxe7 because they give too much importance to black's passed c pawn. In this case the pawn is a liability, not an asset. Black's hands seemed tied because of the three to one kingside pawn majority, while the c pawn is easily stopped. > >>7. Avoid the move 1.Rh5 because 1. ... Rh6! 2. Bxf5 Rxh5 3.Bxh3 and black is >>winning because of the passed h pawn combined with the advance of the king. >>Nimzo did play Rh6 and lost. A better try is the simple 1. Bxf5 Bxf5 2. Rxh4. >>7R/5k2/1r6/1pN2n2/1Pp4p/2P4b/1K6/1B6 w - - id "Nimzo 98 - Shredder 2.0"; am Rh5; > >Junior needed 5 minutes and 35 seconds to find the mistake Rh5 instead of the >right move 1.Bxf5 >and it needed 40 minutes and 43 seconds to reject it again for Bxf5. > >It did not reject Rh5 because it believed Rh5 is a losing move but because it >evaluated Bxf5 as a bigger advantage for white(The advantages are small) >> >>8. Shredder2 should play Qb8. Instead Shredder played 1. ... Ng4? 2. f4 Nxe3 3. >>Qxb2 Nxf1 4. fxe5 and now the Knight on f1 is trapped. >>3r2k1/2qp1ppp/2p1pn2/4b3/4P3/2PBB3/Pr3PPP/R1Q2R1K b - - id "Nimzo 98 - Shredder >>2.0"; am Ng4; > >Junior needed 14 minutes and 34 seconds to reject Ng4. >It did not like Qb8 but prefered Qb7 >It did not think to play 3...Nxf1 when it considered 1...Ng4 but prefered >3...Bxf4 3. ... Bxf4 looks good enough to hold on. There is a line where white can win an extra pawn but it ends up in a drawish opposite bishop ending. > >Uri
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.