Author: Uri Blass
Date: 01:12:42 07/30/01
Go up one level in this thread
On July 29, 2001 at 22:42:39, Slater Wold wrote: >On July 29, 2001 at 21:03:11, Christophe Theron wrote: > >>On July 29, 2001 at 18:34:48, Slater Wold wrote: >> >>>When you see this during a game: >>> >>> >>>Analysis by Deep Junior 6.0: >>> >>>21...Rd8 22.Bxh7+ Kxh7 23.Qh4+ Kg6 24.Qxd8 Bb7 25.Rxf7 >>> +- (3.26) Depth: 6 00:00:00 5kN >>>21...Re7 22.Qd6 Re8 23.Rxf7 h5 24.Qd3 >>> ± (1.29) Depth: 6 00:00:00 16kN >>>21...h6 22.Bc2 >>> ± (0.75) Depth: 6 00:00:00 17kN >>>21...h6 22.Qd3 Re7 23.Bh7+ Kh8 24.Rxf7 Rxf7 25.Rxf7 >>> ± (0.97) Depth: 9 00:00:00 41kN >>>21...g6 22.Rxf7 >>> ± (0.75) Depth: 9 00:00:00 52kN >>>21...g6 22.Rxf7 Qb3 23.Qd3 Qxd3 24.Bxd3 Ra7 25.R1f6 >>> ± (1.14) Depth: 12 00:00:00 644kN >>>21...Qxa3 22.Bxh7+ >>> ± (0.75) Depth: 12 00:00:02 3306kN >>>21...Qxa3 22.Bxh7+ >>> ± (0.75) Depth: 14 00:00:02 3361kN >>>21...Qxa3 22.Bxh7+ Kxh7 23.Rxf7 Bb7 24.Qh4+ Kg8 25.Rxg7+ Kxg7 26.Qf6+ Kg8 >>>27.Qf7+ Kh8 28.Qxe8+ Kg7 29.Rf7+ Kh6 30.Rf6+ Kh7 >>> +- (#12) Depth: 16 00:00:44 85447kN >>>21...h5 22.Qd3 Bb7 23.Bxb7 Rf8 24.Bxa6 Qc6 25.Bb5 Qc5 26.Bxd7 Qxe5 27.Qd4 Qxd4 >>>28.exd4 >>> +- (7.58) Depth: 16 00:01:09 85447kN >>>21...h6 22.Qd3 Kh8 23.Rxf7 Rg8 24.Rf8 g6 25.Bc6 Rxf8 26.Rxf8+ Kg7 27.Bxa4 Kxf8 >>>28.Qxg6 >>> +- (6.29) Depth: 16 00:01:23 107286kN >>>21...g6 22.Rxf7 b5 23.Bd3 >>> +- (6.16) Depth: 16 00:01:33 122505kN >>> >>>Then you've been dupped by GT 2.0. >>> >>>Here's the position: >>>[D]2b1r1k1/3p1ppp/rp2p3/p1Q1P3/q1P1B3/P3P3/5RPP/5RK1 w - - 0 1 >>> >>>Just thought some would enjoy this! >>> >>> >>>Slate >> >> >>I fear there is a problem in your FEN. I see 22.Qd3 in the analysis, but the >>white Queen cannot go to d3 on your diagram... >> >> >> Christophe > >Weird. I did not even notice that. I clipped this while playing, and saved it >for later. This is _EXACTLY_ what I clipped: You posted the wrong position You posted the position before 21.Qd4 and not after 21.Qd4 Here is the right position with Junior7's analysis: New game - Gambit Tiger 2.0 2b1r1k1/3p1ppp/rp2p3/p3P3/q1PQB3/P3P3/5RPP/5RK1 b - - 0 1 Analysis by Junior 7: 21...h5 22.Rxf7 Qxa3 ± (1.10) Depth: 3 00:00:00 21...h6 22.Rxf7 Qxa3 ± (1.03) Depth: 3 00:00:00 21...Re7 22.Qd3 ± (0.75) Depth: 3 00:00:00 21...Re7 22.Bxh7+ Kf8 23.Qd3 Bb7 +- (1.85) Depth: 6 00:00:00 4kN 21...Rf8 22.Qd3 g6 23.Rf6 ± (1.00) Depth: 6 00:00:00 5kN 21...Qb3 22.Rxf7 Qxa3 23.Kh1 ± (0.82) Depth: 6 00:00:00 9kN 21...Qxa3 22.Rxf7 a4 23.Kh1 ± (0.81) Depth: 6 00:00:00 11kN 21...Qxa3 22.Rxf7 a4 23.Bc2 Ra5 24.Qxb6 Rxe5 ± (0.82) Depth: 9 00:00:00 38kN 21...Qxa3 22.Rxf7 a4 23.h4 Ra5 24.h5 Rc5 ± (1.02) Depth: 12 00:00:01 623kN 21...Rf8! ² (0.52) Depth: 12 00:00:03 2165kN 21...Rf8! 22.Qd6 g6 23.Qd3 Ra7 24.Rf6 Bb7 25.Bxb7 ² (0.49) Depth: 12 00:00:04 2594kN 21...Rf8 22.Qd6 g6 23.Qe7 f5 24.exf6 Rf7 25.Qe8+ Rf8 26.Bxg6 Qxa3 27.Qe7 Qxe7 28.fxe7 Rxf2 29.e8Q+ +- (3.53) Depth: 15 00:00:29 17907kN 21...Rf8 22.Qd6 g6 23.Qe7 f5 24.exf6 Rf7 25.Qe8+ Rf8 26.Bxg6 Rxe8 27.f7+ Kg7 28.fxe8Q Qxc4 29.Rc2 Qd5 30.Qf7+ Kh6 +- (3.73) Depth: 17 00:03:35 132916kN 21...Rf8 22.Qd6 +- (4.03) Depth: 18 00:05:46 202636kN (Blass, Tel-aviv 30.07.2001) <snipped> >And as you can see in the analysis, it is searching move 21. > >And played 21...Qb3, and that doesn't even show up in it's analysis. > >Weird eh? Nothing is weird. In yout analysis Junior jumped from depth 14 to 16 when in the game it got depth 15. Different depthes mean often different moves and even the same depth may mean different moves if the jump is not the same(jumping from depth 14 to 16 does not always produce the same move as jumping from depth 15 to 16). If you want to reproduce Qb3 you should give your Junior playing condition and not analysis condition. You should try to play at the same time control(the previous moves are also important because without them it is possible that Junior cannot reproduce the same jump to depth 15) > >Anyway, the point was, DJ analysis went from .75 for white to 5.65 for white in >one move. GT 2.0 is the only program that can do this to another >program......that I have ever seen. I am sure that gambittiger is not the only program that can do it. Gambittiger is not the first program to win a game by a mate attack and I believe that there ae bigger changes in the evaluation in one move. I saw cases when tiger believed it has almost a queen advantage and drew the game(tiger had an unstoppable passed pawn in queen endgame but underestimated the chances of the opponent for a perpetual check). 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.