Author: Stephen Ham
Date: 07:30:19 04/15/04
Go up one level in this thread
On April 14, 2004 at 20:46:20, Sam Lloyd wrote: >On April 14, 2004 at 11:12:08, Stephen Ham wrote: > >>Hi Terry, >> >>I agree with you and am glad that 13 a3! interests you as well. >> >>As you know, I'm a correspondnece chess player with a technical style of play >>(tactics aren't my strength - especially against machines!). So I've had fun >>testing engines to see if any of them will select my purely human moves. So far, >>there's been very little acceptance of my positional/technical move tests by the >>engines. The closest correlation seems to be with Rebel 12. In one of my tests, >>it was the only engine to select the move that I played. When I fed my move into >>the other engines that didn't select my move, their evaluations eventually rose >>to a favorable figure. So it's clear that engines in general won't find the >>"positional" moves that I think are best. But when forced to play the move, they >>eventually learn to like the resulting positions, although the evaluation may >>not be as high as the original PV for that engine. >> >>So I think it would be fun for you and other engine experts to test 13 a3! to >>see if any engine selects it. Since the justifications behind 13 a3! are deep >>(they certainly are for humans), you may need to allow 24-hours or so for the >>test. >> >>Thanks again for mentioning this game, Terry. It is indseed rich in test >>positions, both tactically and of a positional/technical nature. I'll try to >>Rebel 12 on this position and will report what it selects. >> >>Stephen >> >Rebel 12 finds 13. a3 instantly, and stays with it. > >Hiarcs 9 takes a little longer, but as I suspected found 13. a3 and stays with >it as well. > >My hardware is prehistoric by todays standards. P-III 500 128MB RAM with 64MB >for HTs. > >Hiarcs 9 - McCracken,T >[D]r1bq1rk1/3nbppp/2n1p3/2ppP3/pp3B1P/3P1NP1/PPP2PB1/R2QRNK1 w - - 0 1 > >Analysis by Hiarcs 9: > >13.Qd2 a3 > ³ (-0.44) Depth: 1/2 00:00:00 >13.Qd2 Bb7 14.c3 > ³ (-0.63) Depth: 2/6 00:00:00 >13.h5 a3 14.bxa3 Rxa3 15.h6 g6 16.Bc1 Ra8 17.Bb2 Bb7 > ³ (-0.44) Depth: 2/6 00:00:00 >13.h5 a3 14.bxa3 Rxa3 15.h6 g6 16.Bc1 Ra8 17.Bb2 Bb7 > ³ (-0.44) Depth: 2/6 00:00:00 >13.h5 a3 14.bxa3 Rxa3 15.h6 g6 16.Bc1 Ra8 17.Bb2 Bb7 > ³ (-0.44) Depth: 3/6 00:00:00 >13.h5 a3 14.bxa3 Rxa3 15.h6 g6 16.Bc1 Ra8 17.Bb2 Bb7 > ³ (-0.44) Depth: 4/6 00:00:00 >13.h5 a3 14.bxa3 Rxa3 15.h6 g6 16.Bc1 Ra8 17.Bb2 Bb7 > ³ (-0.44) Depth: 5/6 00:00:00 >13.h5 a3 14.bxa3 Rxa3 15.h6 g6 16.Bc1 Ra8 17.Bb2 Bb7 > ³ (-0.44) Depth: 6/6 00:00:00 >13.h5 a3 14.bxa3 Rxa3 15.h6 g6 16.Bc1 Ra8 17.Bb2 Bb7 > ³ (-0.44) Depth: 7/8 00:00:00 1kN >13.h5 a3 14.bxa3 Rxa3 15.h6 g6 16.Bc1 Ra8 17.Bb2 Bb7 > ³ (-0.44) Depth: 8/8 00:00:00 1kN >13.h5 a3 14.bxa3 Rxa3 15.h6 g6 16.Bc1 Ra8 17.Bb2 Bb7 > ³ (-0.44) Depth: 9/12 00:00:00 4kN >13.h5 f6 14.exf6 Rxf6 15.Qe2 a3 16.b3 Nb6 17.Ne5 Nxe5 18.Bxe5 > ³ (-0.43) Depth: 10/26 00:00:58 2162kN >13.h5 a3 14.bxa3 Rxa3 15.h6 g6 16.Bc1 Ra5 17.Bb2 Bg5 18.Nxg5 Qxg5 19.c4 > ³ (-0.56) Depth: 11/27 00:02:24 5363kN >13.a3 Bb7 > ³ (-0.55) Depth: 11/27 00:03:13 7211kN >13.a3 bxa3 14.bxa3 Ba6 15.h5 c4 16.h6 g5 17.Nxg5 Bxg5 18.Qg4 Ncxe5 19.Qxg5+ > ³ (-0.46) Depth: 11/28 00:04:38 10347kN >13.a3 bxa3 14.bxa3 Rb8 15.c4 Nb6 16.Ne3 Na5 17.Rb1 Nb3 18.Nd2 dxc4 19.Ndxc4 Ba6 > ³ (-0.32) Depth: 12/30 00:12:30 28289kN >13.a3 bxa3 14.bxa3 Re8 15.c4 Nb6 16.Ne3 Rb8 17.Rb1 > ³ (-0.29) Depth: 13/37 00:37:12 83484kN > >(McCracken, None 14.04.2004) > >Terry Thanks, Terry and Richard, for running those tests. I confess that I'm amazed by the results. Both you, with Rebel and Hiarcs, and Richard's Baron found 13 a3. I must admit that if I had to bet money, I would have bet against any engine, other than possibly Rebel, in finding 13 a3. And I would have bet that Rebel would require overnight analysis to find it. Again, the results are astounding. What's interesting too is that these engines, while finding the best move, probably don't comprehend the position very well, since they all seem to give a very slight edge to Black. Instead, most humans, especially after seeing the motifs in the Fischer game and similar motifs in the Closed Sicilian, would favor White with a clear advantage. Perhaps this is yet another example of engines finding the correct move, but their evaluations are wrong. I see this with Shredder all the time. All the best, Stephen
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.