Test position: Tough sac 19.Bg6!! (Hector-So)

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

Albert Silver
Posts: 3019
Joined: Wed Mar 08, 2006 9:57 pm
Location: Rio de Janeiro, Brazil

Test position: Tough sac 19.Bg6!! (Hector-So)

Post by Albert Silver »

[D]r1b3k1/ppp1q1pp/2n1p3/3pP2P/8/2NB2Q1/PPP3P1/2K2R2 w - - 0 19

Hector found the very strong 19.Bg6!! in this position. The main line goes 19.Bg6!! hxg6 20.hxg6 Nd4 21.Rf7 Qd8 22.Qg4!

For the record, Hector missed 22.Qg4! and played 22.Qf4? but his concept was immaculate.
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
Top
User avatar
Eelco de Groot
Posts: 4567
Joined: Sun Mar 12, 2006 2:40 am
Full name:   

Re: Test position: Tough sac 19.Bg6!! (Hector-So)

Post by Eelco de Groot »

[Event "19th Sigeman & Co"]
[Site "Malmo SWE"]
[Date "2011.06.13"]
[Round "5"]
[White "Hector,J"]
[Black "So,W"]
[Result "0-1"]
[WhiteElo "2588"]
[BlackElo "2667"]
[EventDate "2011.06.09"]
[ECO "C13"]

1. e4 e6 2. d4 d5 3. Nc3 Nf6 4. Bg5 Be7 5. e5 Nfd7 6. h4 O-O 7. Qg4 f6 8.
Nf3 Rf7 9. exf6 Nxf6 10. Qg3 Nc6 11. O-O-O Qf8 12. h5 Bb4 13. Bxf6 Rxf6 14.
Bd3 Bd6 15. Ne5 Rxf2 16. Rdf1 Bxe5 17. dxe5 Rxf1+ 18. Rxf1 Qe7 19. Bg6 hxg6
20. hxg6 Nd4 21. Rf7 Qd8 22. Qf4 Nf5 23. g4 Nh4 24. Rxg7+ Kxg7 25. Qf7+ Kh6
26. Qh7+ Kg5 27. Qh5+ Kf4 28. Ne2+ Ke3 29. g5 Nf3 30. Kd1 Qxg5 31. Qh7 Kf2
0-1
Albert Silver wrote:[D]r1b3k1/ppp1q1pp/2n1p3/3pP2P/8/2NB2Q1/PPP3P1/2K2R2 w - - 0 19

Hector found the very strong 19.Bg6!! in this position. The main line goes 19.Bg6!! hxg6 20.hxg6 Nd4 21.Rf7 Qd8 22.Qg4!

For the record, Hector missed 22.Qg4! and played 22.Qf4? but his concept was immaculate.
Rainbow Serpent does not find it, only for a moment there is a Fail High, but if Bg6 forced sees where that leads after some Fail Lows, and Fail Highs when the search window drops too much. The routine for calculating the new searchwindow is a bit impatient :? Eventually it gets a score for hxg6 inside the searchwindow but then thinks not accepting the offer refutes it. It needs more hardware I think to properly search the moves, 256 Mb is not enough. The variation or score may not be accurate enough, I will let it run a bit longer, I suppose the analysis of 19...Nd8 could be improved upon, f.e. is 20. Bxh7+?! the right idea?

r1b3k1/ppp1q1pp/2n1p3/3pP2P/8/2NB2Q1/PPP3P1/2K2R2 w - -

Engine: eXperimental Tommy Cooper - Rainbow Serpent Build 209 (Athlon 2009 MHz, 256 MB)
Latest mini-change: Forward mobility implementation, but poorly tuned
by Tord Romstad, Marco Costalba, Joona Kiiski

1/01 0:00 +3.31 19.Bxh7+ (1.179) 3

2/02 0:00 +3.31 19.Bxh7+ Kxh7 (1.976) 6

3/04 0:00 +2.62 19.Bxh7+ Kxh7 20.Rh1 Kg8 (3.096) 9

4/04 0:00 -0.56 19.Bxh7+ Kxh7 20.Nxd5 exd5 (23.608) 62

4/06 0:00 -0.08 19.Rf6 Nxe5 20.Rf2 Nxd3+ 21.cxd3 Kh8 (34.046) 87

4/06 0:00 +2.30 19.h6 Nb4 20.Qf2 Nxd3+ 21.cxd3 Bd7 (55.167) 125

4/06 0:00 +2.34 19.a3 Kh8 20.h6 Bd7 21.hxg7+ Qxg7 (80.054) 165

5/06 0:00 +0.84 19.a3 Bd7 20.a4 Rf8 21.Bxh7+ Kxh7 (197.054) 280

5/06 0:00 +1.93 19.h6 Nb4 20.a3 Nxd3+ 21.cxd3 Kh8 (247.686) 310

5/06 0:00 +1.94 19.Nb5 Nb4 20.Qf2 Nxd3+ 21.cxd3 a6 (257.998) 311

6/06 0:00 +1.29 19.Nb5 Nb4 20.Qf3 Nxd3+ 21.cxd3 Bd7 (291.221) 326

6/06 0:00 +1.55 19.h6 Nb4 20.Qf3 Nxd3+ 21.cxd3 Bd7 (332.258) 348

7/08 0:01 +0.98 19.h6 Bd7 20.Nb5 Nb4 21.Nxc7 Nxd3+
22.cxd3 Rf8 (783.362) 424

7/06 0:01 +1.34 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.a3 Nc6 (821.991) 431

8/08 0:02 +0.88 19.Nb5 Nb4 20.h6 Nxd3+ 21.Qxd3 Bd7
22.Nxc7 Rf8 (1.006.943) 444

8/08 0:02 +0.94 19.h6 Bd7 20.Nb5 Nb4 21.Nxc7 Nxd3+
22.Qxd3 Rf8 (1.151.003) 449

8/13 0:02 +1.09 19.Kb1 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.h6 Bd7 23.h7+ Kh8 24.Rf7 Qxf7
25.Qxf7 (1.208.496) 449

8/06 0:02 +1.09 19.Qf4 Nb4 20.a3 Nxd3+ 21.cxd3 Bd7 (1.292.624) 454

8/16 0:02 +1.16 19.Rf3 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.h6 Bd7 23.h7+ Kh8 24.Rf7 Qf8
25.a3 Nc6 26.Rxd7 Nxe5 (1.354.773) 456

9/10 0:03 +0.75-- 19.Rf3 Bd7 20.Nb5 Nb4 21.Bxh7+ Kxh7
22.Nxc7 Nxa2+ 23.Kd2 Kh8 (1.708.811) 463

9/06 0:04 +0.80 19.Qf4 Nb4 20.Be2 Bd7 21.a3 Nc6 (1.906.359) 471

10/12 0:07 +0.76 19.Qf4 Bd7 20.a3 Rf8 21.Qxf8+ Qxf8
22.Bxh7+ Kxh7 23.Rxf8 Nxe5 24.Rf1 b6 (3.581.071) 496

11/12 0:10 +0.48-- 19.Qf4 Bd7 20.a3 Rf8 21.Qxf8+ Qxf8
22.Bxh7+ Kxh7 23.Rxf8 Nxe5 24.Rf1 b6 (5.359.645) 508

11/12 0:12 +0.51 19.h6 Bd7 20.Rh1 g6 21.a3 Rf8
22.Bxg6 hxg6 23.Qxg6+ Kh8 24.Qg3 Rf5 (6.496.127) 515

11/15 0:14 +0.54 19.Rh1 Bd7 20.h6 g6 21.Nb5 Be8
22.Rf1 a6 23.Nc3 Nd4 24.Qf4 Nf5
25.g4 g5 26.Qf3 (7.372.213) 510

12/08 0:19 +0.36 19.Rh1 Bd7 20.Nb5 Nb4 21.Bxh7+ Kxh7
22.Nxc7 Rc8 (9.804.798) 510

13/14 0:28 +0.06 19.Rh1 Bd7 20.Kb1 Rf8 21.h6 g6
22.Ne2 Qf7 23.c3 a6 24.Qg5 Na5
25.Nd4 c5 (14.800.909) 510

14/16 0:36 0.00 19.Rh1 Bd7 20.Kb1 Rf8 21.h6 g6
22.Ne2 Qf7 23.c3 a6 24.Qg5 Na5
25.Qe3 Bb5 26.Bxb5 axb5 (18.742.009) 513

15/16 0:55 -0.22 19.Rh1 Bd7 20.Kb1 Rf8 21.h6 g6
22.Nd1 Qf7 23.Ne3 Qf4 24.Qh3 Nxe5
25.Rf1 Nxd3 26.Rxf4 Nxf4 (28.348.466) 511

15/10 0:59 -0.21 19.h6 Bd7 20.Rf4 Nxe5 21.Qe3 Nxd3+
22.cxd3 Rf8 23.Nxd5 Qd6 (30.468.832) 514

15/10 1:03 +0.28++ 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Kb1 Bd7 23.Nxc7 Rf8 (32.822.727) 513

15/14 1:05 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nd6 cxd6 23.exd6 Qxd6 24.Qe8+ Kh7
25.Qg6+ Kg8 (33.475.214) 513

16/14 1:07 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nd6 cxd6 23.exd6 Qxd6 24.Qe8+ Kh7
25.Qg6+ Kg8 (34.760.978) 514

17/14 1:10 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nd6 cxd6 23.exd6 Qxd6 24.Qe8+ Kh7
25.Qg6+ Kg8 (36.491.641) 515

17/13 1:13 +0.12++ 19.Bg6 hxg6 20.hxg6 Bd7 21.Rh1 Nd4
22.Rh8+ Kxh8 23.Qh3+ Kg8 24.Qh7+ Kf8
25.Qh8+ (38.053.996) 515

17/14 1:23 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nd6 cxd6 23.exd6 Qxd6 24.Qe8+ Kh7
25.Qg6+ Kg8 (43.568.785) 519

18/14 1:28 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nd6 cxd6 23.exd6 Qxd6 24.Qe8+ Kh7
25.Qg6+ Kg8 (46.121.849) 520

19/14 1:42 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nd6 cxd6 23.exd6 Qxd6 24.Qe8+ Kh7
25.Qg6+ Kg8 (53.622.426) 521

20/14 1:59 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nd6 cxd6 23.exd6 Qxd6 24.Qe8+ Kh7
25.Qg6+ Kg8 (62.550.325) 521

21/14 2:23 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nd6 cxd6 23.exd6 Qxd6 24.Qe8+ Kh7
25.Qg6+ Kg8 (74.795.977) 521

22/25 2:53 +0.12++ 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.h6 Bd7 23.Nd6 cxd6 24.exd6 Be8
25.h7+ Kh8 26.dxe7 Bxg6 27.Rf8+ Kxh7
28.Rxa8 Nxc2 29.Rxa7 Nb4 30.Ra8 e5
31.Kd2 (90.685.176) 521

22/10 7:14 0.00 19.Nb5 Bd7 20.Nxc7 Rf8 21.Rh1 Nb4
22.Bxh7+ Kxh7 23.a3 Nxc2 (241.822.826) 556

23/10 7:27 0.00 19.Nb5 Bd7 20.Nxc7 Rf8 21.Rh1 Nb4
22.Bxh7+ Kxh7 23.a3 Nxc2 (248.356.049) 555

24/10 7:51 0.00 19.Nb5 Bd7 20.Nxc7 Rf8 21.Rh1 Nb4
22.Bxh7+ Kxh7 23.a3 Nxc2 (260.573.110) 552

25/30 10:07 0.00 19.Nb5 Bd7 20.Nxc7 Rf8 21.Rxf8+ Qxf8
22.Nb5 Nb4 23.Nd6 Nxd3+ 24.cxd3 Qf1+
25.Kd2 Qa1 26.Qf4 Qxb2+ 27.Kd1 Qb1+
28.Ke2 h6 29.Qf7+ Kh8 30.Qxd7 Qc2+
31.Kf3 Qd1+ 32.Kf2 (332.717.859) 547

26/08 12:17 0.00 19.Nb5 Bd7 20.Nxc7 Rf8 21.Rxf8+ Qxf8
22.Nb5 Nb4 (398.442.616) 540

27/12 20:47 0.00 19.Nb5 Bd7 20.Nxc7 Rf8 21.Rxf8+ Qxf8
22.c3 a6 23.Kb1 Kh8 24.Ka1 Qd8 (683.477.593) 548

28/12 29:27 -0.16 19.Nb5 Bd7 20.Nxc7 Rf8 21.Rxf8+ Qxf8
22.a3 a6 23.h6 g6 24.Qe3 d4 (973.728.493) 551

29/18 59:41 -0.31 19.Nb5 Bd7 20.Nxc7 Rf8 21.Rxf8+ Qxf8
22.Nb5 Be8 23.c3 Bxh5 24.Nd6 Qd8
25.b4 Qd7 26.Kb2 a6 27.b5 axb5 (1.989.686.287) 555

But now after 19. Bg6 I get the same score, -0.31:


[D]r1b3k1/ppp1q1pp/2n1p1B1/3pP2P/8/2N3Q1/PPP3P1/2K2R2 b - -

Engine: eXperimental Tommy Cooper - Rainbow Serpent Build 209 (Athlon 2009 MHz, 256 MB)
Latest mini-change: Forward mobility implementation, but poorly tuned
by Tord Romstad, Marco Costalba, Joona Kiiski

1/01 0:00 +0.96 19...hxg6 (1.698) 113

2/04 0:00 -0.12 19...hxg6 20.hxg6 Bd7 21.Rf7 (16.847) 366

3/07 0:00 +0.96 19...hxg6 20.Qxg6 Nxe5 21.Qg3 d4
22.Qxe5 dxc3 (18.259) 396

4/10 0:00 -0.12 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qf4 Nf5 23.Qb4 Nh6 24.Rf4 (19.725) 428

5/07 0:00 +0.20 19...hxg6 20.hxg6 Nd4 21.Rh1 Nf5
22.Qh2 Nh6 (23.288) 375

6/09 0:00 +0.20 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Kb1 Nf5 23.Qh3 Nh6 (52.635) 421

7/08 0:00 -0.20 19...hxg6 20.hxg6 Bd7 21.Rf7 d4
22.Rxe7 Nxe7 23.Qf4 (79.510) 464

8/11 0:00 +0.32++ 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qf4 Nf5 23.Qb4 Nh6 24.Rf4 Nf5 (146.446) 494

8/08 0:00 -0.30 19...hxg6 20.hxg6 b6 21.Rf7 Qxf7
22.gxf7+ Kxf7 23.Nb5 (267.135) 503

9/09 0:01 +0.02++ 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Kb1 Nf5 23.Qg4 Nh6 (539.228) 515

9/09 0:01 +0.34++ 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Kb1 Nf5 23.Qg4 Nh6 (654.242) 523

9/11 0:01 +0.28 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Rf4 Nf5 23.Qh2 Nh6 24.g4 Qg5 (705.380) 525

10/11 0:02 +1.05++ 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Rf4 Nf5 23.Qh2 Nh6 24.g4 Qg5 (1.073.380) 532

10/15 0:02 +1.10 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qf2 Nf5 23.Qc5 b6 24.Qb4 Nh6
25.Rf1 Nf5 26.Kb1 Qg5 (1.265.543) 525

11/13 0:02 +1.16 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Nh6 24.Rf4 Bd7
25.Qh3 Qg5 (1.423.968) 533

12/09 0:04 +1.46 19...hxg6 20.hxg6 Nd4 21.Qg4 Nf5
22.Qh5 Nh6 23.Rf3 Bd7 (2.217.698) 521

13/19 0:10 +1.36 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Rf4 Nf5 23.Qh3 Nh6 24.Qh5 Bd7
25.g4 Qe7 26.g5 Rf8 27.Rxf8+ Qxf8
28.gxh6 gxh6 (5.761.595) 548

14/19 0:11 +1.65++ 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Rf4 Nf5 23.Qh3 Nh6 24.Qh5 Bd7
25.g4 Qe7 26.g5 Rf8 27.Rxf8+ Qxf8
28.gxh6 gxh6 (6.281.791) 547

14/19 0:14 +1.92 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.b3 Rf8 27.Rd1 Qh6
28.Qxh6 gxh6 (7.701.405) 548

15/32 0:24 +1.73 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.Rh4 Rf8 27.a3 Kf7
28.Rh5 Qg6 29.Qh3 Ke8 30.Rh8 Qf7
31.Rxf8+ Qxf8 32.Qh5+ Kd8 (13.391.545) 547

16/32 0:25 +1.73 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.Rh4 Rf8 27.a3 Kf7
28.Rh5 Qg6 29.Qh3 Ke8 30.Rh8 Qf7
31.Rxf8+ Qxf8 32.Qh5+ Kd8 (13.979.627) 547

17/32 0:27 +1.73 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.Rh4 Rf8 27.a3 Kf7
28.Rh5 Qg6 29.Qh3 Ke8 30.Rh8 Qf7
31.Rxf8+ Qxf8 32.Qh5+ Kd8 (15.113.408) 546

18/21 0:30 +1.73 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Qf4 Qxf4+ 26.Rxf4 Bd7 27.Kd2 Rf8
28.g3 b6 29.Rxf8+ Kxf8 (16.687.874) 546

19/32 0:36 +1.85++ 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.Rh4 Rf8 27.a3 Kf7
28.Rh5 Qg6 29.Qh3 Ke8 30.Rh8 Qf7
31.Rxf8+ Qxf8 32.Qh5+ Kd8 (20.055.516) 545

19/21 0:43 +1.97 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.g3 Rf8 27.Rd1 Qh6
28.Qg2 Bc6 29.Rh1 d4 (23.906.961) 547

20/25 0:57 +1.81 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.Rh4 Rf8 27.a3 Kf7
28.Rh5 Qg6 29.Qh3 Ke7 30.Qh4+ Ke8
31.Rg5 Qh6 (31.087.873) 545

21/39 1:13 +1.70 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.a3 Rf8 27.Rh4 Kf7
28.Rh5 Qg6 29.Qh3 Ke8 30.Rh8 Qf7
31.Rh7 Qf5 32.Rxg7 Qxh3 (40.157.461) 543

22/39 1:15 +1.70 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.a3 Rf8 27.Rh4 Kf7
28.Rh5 Qg6 29.Qh3 Ke8 30.Rh8 Qf7
31.Rh7 Qf5 32.Rxg7 Qxh3 (40.965.130) 542

23/39 1:20 +1.70 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.a3 Rf8 27.Rh4 Kf7
28.Rh5 Qg6 29.Qh3 Ke8 30.Rh8 Qf7
31.Rh7 Qf5 32.Rxg7 Qxh3 (43.681.156) 541

24/39 1:25 +1.70 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.a3 Rf8 27.Rh4 Kf7
28.Rh5 Qg6 29.Qh3 Ke8 30.Rh8 Qf7
31.Rh7 Qf5 32.Rxg7 Qxh3 (46.208.129) 541

25/17 1:48 +1.78 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Rf4 Bd7 26.g4 Rf8 27.Nxd5 exd5 (58.878.218) 541

26/41 2:15 +1.94++ 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Kb1 Bd7 26.a3 Rf8 27.Rh4 Kf7
28.Rh5 Qg6 29.Qh3 Ke8 30.Rh8 Qf7
31.Rh7 Qf5 32.Rxg7 Qxh3 (73.528.143) 543

26/20 3:26 +2.09 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qh2 Qg5+ 23.Rf4 Qxg6 24.Rxd4 Qg5+
25.Qf4 Qxf4+ 26.Rxf4 Bd7 27.Ne2 Rf8
28.c4 Rxf4 29.Nxf4 (114.625.142) 553

27/16 7:53 +1.73-- 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Nh6 24.g4 Nxf7
25.gxf7+ Kf8 26.Qh8+ Kxf7 27.Qxd8 (264.079.618) 557

27/02 10:53 +1.21-- 19...hxg6 20.hxg6 (359.472.799) 549

27/02 22:36 +0.27-- 19...hxg6 20.hxg6 (673.589.688) 496

27/09 24:31 -1.56-- 19...hxg6 20.hxg6 a5 21.Rh1 Nb4
22.Rh8+ Kxh8 23.Qh3+ Kg8 (720.840.427) 489

27/25 32:50 -2.84++ 19...hxg6 20.hxg6 Bd7 21.Rh1 Qe8
22.Qh4 Qxg6 23.Qh8+ Kf7 24.Rf1+ Ke7
25.Qxa8 Be8 26.Qxb7 Qg5+ 27.Kb1 Qxe5
28.Na4 Qd6 29.Nc5 Qxc5 30.Qxc7+ Bd7
31.Qg3 Ke8 (1.016.893.950) 516

27/23 33:25 -2.48++ 19...hxg6 20.hxg6 Bd7 21.Rh1 Qe8
22.Qh4 Qxg6 23.Qh8+ Kf7 24.Rf1+ Ke7
25.Qxa8 Be8 26.Qxb7 Qg5+ 27.Kb1 Qxe5
28.Qc8 d4 29.Nd1 Bg6 30.Qa8 Kd6 (1.035.566.265) 516

27/05 36:30 -1.75++ 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8 (1.140.050.246) 520

27/19 136:43 -1.64 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Qh4 24.Qh7+ Qxh7
25.Rf8+ Kxf8 26.gxh7 c6 27.h8Q+ Kf7
28.Nxd5 cxd5 (4.127.322.975) 503

27/07 177:08 -0.31 19...Nd8 20.Bxh7+ Kxh7 21.Rf4 Nf7
22.Rg4 Nh6 (5.451.741.328) 512
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
Top
User avatar
Eelco de Groot
Posts: 4567
Joined: Sun Mar 12, 2006 2:40 am
Full name:   

Re: Test position: Tough sac 19.Bg6!! (Hector-So)

Post by Eelco de Groot »


28/15 318:46 -1.91 19...Nd8 20.Nb5 Bd7 21.Nd6 h6 22.Nf7 Nxf7
23.Rxf7 Qxf7 24.Bxf7+ Kxf7 25.c4 Kg8
26.cxd5 exd5 (10.601.956.442) 554

28/16 320:40 -1.64 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Qh4 24.Qh7+ Qxh7
25.Rf8+ Kxf8 26.gxh7 c6 27.h8Q+ (10.661.230.824) 554

29/37 469:59 -2.00 19...hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Nh6 24.g4 Qe8
25.Rf4 c6 26.Ne2 Bd7 27.g5 Nf5
28.Nd4 Kf8 29.Qh8+ Ke7 30.Nxf5+ exf5
31.Qxg7+ Kd8 32.Qf6+ Kc7 (15.816.744.941) 560

29/15 470:25 -1.91 19...Nd8 20.Nb5 Bd7 21.Nd6 h6 22.Nf7 Nxf7
23.Rxf7 Qxf7 24.Bxf7+ Kxf7 25.c4 Kg8
26.cxd5 exd5 (15.831.264.993) 560
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
Top
User avatar
Eelco de Groot
Posts: 4567
Joined: Sun Mar 12, 2006 2:40 am
Full name:   

Re: Test position: Tough sac 19.Bg6!! (Hector-So)

Post by Eelco de Groot »

Black tried some other options but they also did not work. Stopped the engine now. The sacrifice is correct it seems, not with very high evaluations but better than any alternative move. Now we have to find that one ply earlier. Maybe the tuning of Forward Mobility can be improved further, but move 22.Qg4 is crucial for the variation. And pretty hard to find for a single core engine.

29/12 788:46 -1.66 19...h6 20.Nb5 Qg5+ 21.Qxg5 hxg5
22.Nxc7 Rb8 23.Bf7+ Kh7 24.Nxe6 Nxe5
25.Nf8+ (27.228.931.041) 575

30/12 805:58 -1.66 19...h6 20.Nb5 Qg5+ 21.Qxg5 hxg5
22.Nxc7 Rb8 23.Bf7+ Kh7 24.Nxe6 Nxe5
25.Nf8+ (27.867.798.400) 576

31/02 859:41 -2.02-- 19...h6 20.Nb5 (29.662.564.200) 575

31/03 906:10 -2.02 19...Bd7 20.Rf7 hxg6 (31.265.061.231) 575

32/14 1028:56-2.38-- 19...Bd7 20.Rf7 Qxf7 21.Bxf7+ Kxf7
22.h6 gxh6 23.Qf4+ Kg7 24.Qg4+ Kf7
25.Nxd5 Nxe5 26.Qf4+ (35.543.645.295) 575

best move: h7-h6 time: 1072:26.407 min n/s: 578.274 nodes: 37.209.910.451
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
Top
User avatar
Eelco de Groot
Posts: 4567
Joined: Sun Mar 12, 2006 2:40 am
Full name:   

Double Pass Futility Pruning in Quiescence Search.

Post by Eelco de Groot »

I am sure Peter Martan can improve on this with Rebel Q3 or my version of that, Rebel Q5T 8-) After another try the testposition now seems a bit easier than before, thanks to the "double pass" Futility Pruning in the quiescense search, as mentioned in the programming forum and based on an idea from John Merlino to only do recaptures after a certain negative depth in the quiescence search. This was also already implemented in Gaviota so it may have been around much longer. Maybe since the seventies :) in the time of Chess 4.6 :) I'm using a very soft version of that that just does heavier futility pruning if a capture move is not in the PV and not capturing on the same square as the previous move. It seems a logical way of limiting the number of full evals needed in the QS. But there is an immediate effect visible on all the evaluations, not what you maybe would anticipate; shorter time to depth/branching factor change.

I like the effect so far, I was thinking of implementing null move in quiescence search again but this is probably better. And it is a soft effect, you could try more pruning easily, this was just my second try, the first one had no effect at all, not even on speed so I hope the Visual C++ compiler actually manages to do this second pass of Fuitility Pruning in parallel with the other tests, in quiescence search that would make a big difference because it is called very often, quiescence search by far the most time consuming part of search IIRC.


[D]r1b3k1/ppp1q1pp/2n1p3/3pP2P/8/2NB2Q1/PPP3P1/2K2R2 w - -

Engine: eXperimental Tommy Cooper - Rainbow Serpent Build 212 (Athlon 2009 MHz, 256 MB)
Latest change: Double Pass Futility Pruning in Quiescence Search on non recapture moves
by Tord Romstad, Marco Costalba, Joona Kiiski

1/01 0:00 +3.31 19.Bxh7+ (1.179) 5

2/02 0:00 +3.31 19.Bxh7+ Kxh7 (1.976) 8

3/04 0:00 +2.62 19.Bxh7+ Kxh7 20.Rh1 Kg8 (3.096) 12

4/04 0:00 -0.56 19.Bxh7+ Kxh7 20.Nxd5 exd5 (23.603) 83

4/06 0:00 -0.08 19.Rf6 Nxe5 20.Rf2 Nxd3+ 21.cxd3 Kh8 (34.027) 109

4/06 0:00 +2.30 19.h6 Nb4 20.Qf2 Nxd3+ 21.cxd3 Bd7 (55.148) 160

4/06 0:00 +2.34 19.a3 Kh8 20.h6 Bd7 21.hxg7+ Qxg7 (80.051) 197

5/06 0:00 +1.27 19.a3 Bd7 20.Kb1 Rf8 21.Bxh7+ Kxh7 (214.132) 318

5/04 0:00 +2.07 19.h6 Nb4 20.Qf3 gxh6 (243.885) 339

6/08 0:00 +1.55 19.h6 Nb4 20.Qf4 Nxd3+ 21.cxd3 Bd7
22.hxg7 Qxg7 (334.479) 375

6/06 0:00 +1.81 19.Kb1 Nb4 20.Nb5 Nxd3 21.cxd3 a6 (376.133) 382

7/06 0:01 +1.33-- 19.Kb1 Nb4 20.Nb5 Nxd3 21.cxd3 a6 (482.688) 406

7/15 0:01 +1.48 19.h6 Nb4 20.Be2 Kh8 21.hxg7+ Qxg7
22.Qh4 Nc6 23.Bd3 Bd7 24.Rh1 Nxe5
25.Bxh7 Qxg2 26.Bd3+ (538.711) 415

8/08 0:01 +1.55 19.h6 Nb4 20.Be2 Bd7 21.a3 Nc6
22.Qf4 Rf8 (734.224) 435

9/10 0:03 +1.23-- 19.h6 Bd7 20.Kb1 Nb4 21.hxg7 Nxd3
22.cxd3 Qxg7 23.Qh4 Rf8 (1.748.303) 490

9/10 0:05 +0.81-- 19.h6 Bd7 20.Kb1 Nb4 21.hxg7 Nxd3
22.cxd3 Qxg7 23.Qh4 Rf8 (2.590.502) 503

9/12 0:07 +0.83 19.Rf2 Kh8 20.a3 Bd7 21.Nb5 Rf8
22.Rxf8+ Qxf8 23.Nxc7 Kg8 24.Nb5 g6 (3.598.804) 508

10/08 0:09 0.00 19.Rf2 Bd7 20.Nb5 Rf8 21.Nxc7 Rxf2
22.Qxf2 Nxe5 (4.773.486) 516

10/06 0:11 +0.39 19.h6 Bd7 20.a4 Nb4 21.hxg7 Qxg7 (5.760.106) 517

10/11 0:11 +0.42 19.Nd1 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kh8
22.Rf7 Nxa2+ 23.Kb1 Qxf7 24.Qxf7 (5.934.375) 516

10/13 0:13 +0.48 19.Ne2 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.h6 Bd7 23.h7+ Kh8 24.Rf7 Qxf7
25.Qxf7 (6.922.763) 519

11/08 0:15 -0.23 19.Ne2 Bd7 20.h6 Rf8 21.Rh1 g5 22.c3 Be8 (8.104.900) 516

11/12 0:20 +0.10 19.h6 Bd7 20.Qe3 Rf8 21.Rh1 g6
22.Nb5 Nb4 23.Nxc7 Nxd3+ 24.cxd3 Rc8 (10.862.803) 520

11/06 0:24 +0.13 19.Kd1 Bd7 20.Rh1 Rf8 21.Bxh7+ Kxh7 (12.540.070) 520

11/10 0:27 +0.14 19.a3 Bd7 20.Qf4 d4 21.Qe4 g6
22.hxg6 dxc3 23.gxh7+ Kh8 (14.163.267) 520

12/14 0:28 +0.28 19.a3 Bd7 20.Qf4 b6 21.Kb1 Rf8
22.Qxf8+ Qxf8 23.Bxh7+ Kxh7 24.Rxf8 Nxe5
25.Rf1 Be8 (14.919.729) 522

13/08 0:37 -0.08 19.a3 Bd7 20.Kb1 Rf8 21.Rh1 Nd4
22.Bxh7+ Kxh7 (19.914.167) 530

13/12 0:38 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nxc7 Qxc7 23.Qe8+ Kh7 24.Qg6+ Kg8 (20.190.375) 530

13/14 0:44 +0.01 19.h6 Bd7 20.Rh1 g6 21.Nd1 Nb4
22.Bxg6 Nxa2+ 23.Kb1 hxg6 24.Qxg6+ Kh8
25.Kxa2 Rg8 (23.539.722) 528

14/16 1:13 -0.11 19.h6 Bd7 20.Rh1 g6 21.Kb1 Rf8
22.Ne2 Be8 23.Qe3 Nb4 24.Qxa7 Qg5
25.Qxb7 Nxd3 26.cxd3 Qxg2 (38.535.786) 525

14/12 1:14 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nxc7 Qxc7 23.Qe8+ Kh7 24.Qg6+ Kg8 (38.972.990) 525

15/12 1:22 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nxc7 Qxc7 23.Qe8+ Kh7 24.Qg6+ Kg8 (42.990.335) 522

16/12 1:23 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nxc7 Qxc7 23.Qe8+ Kh7 24.Qg6+ Kg8 (43.757.196) 522

17/12 1:26 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nxc7 Qxc7 23.Qe8+ Kh7 24.Qg6+ Kg8 (45.205.862) 522

18/12 1:31 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nxc7 Qxc7 23.Qe8+ Kh7 24.Qg6+ Kg8 (47.683.328) 523

19/12 1:36 0.00 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.Nxc7 Qxc7 23.Qe8+ Kh7 24.Qg6+ Kg8 (50.433.259) 523

20/22 1:48 +0.12++ 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.h6 Bd7 23.Nd6 cxd6 24.exd6 Be8
25.h7+ Kh8 26.dxe7 Bxg6 27.Rf8+ Kxh7
28.Rxa8 Nxc2 29.e8Q Bxe8 (56.999.106) 523

20/22 1:49 +0.24++ 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.h6 Bd7 23.Nd6 cxd6 24.exd6 Be8
25.h7+ Kh8 26.dxe7 Bxg6 27.Rf8+ Kxh7
28.Rxa8 Nxc2 29.e8Q Bxe8 (57.605.514) 523

20/29 1:56 +0.48++ 19.Nb5 Nb4 20.Bxh7+ Kxh7 21.Qg6+ Kg8
22.h6 Bd7 23.Nd6 cxd6 24.exd6 Be8
25.h7+ Kh8 26.dxe7 Bxg6 27.Rf8+ Kxh7
28.Rxa8 Nxc2 29.e8Q Bxe8 30.Rxe8 Nd4
31.Re7 a5 32.Rxb7 (60.910.419) 524

20/14 3:27 0.00-- 19.Nb5 Bd7 20.Nxc7 Rf8 21.Rh1 Nb4
22.Bxh7+ Kxh7 23.Qg6+ Kg8 24.h6 Qf7
25.h7+ Kh8 (110.867.500) 534

20/14 4:07 -0.12-- 19.Nb5 Bd7 20.Nxc7 Rf8 21.Rh1 Nb4
22.Bxh7+ Kxh7 23.Qg6+ Kg8 24.h6 Qf7
25.h7+ Kh8 (132.095.770) 534

20/24 5:22 -0.35 19.Nb5 Bd7 20.Nxc7 Rf8 21.Nb5 Nb4
22.Rxf8+ Qxf8 23.Nd6 Nxd3+ 24.cxd3 Qf1+
25.Kc2 Qe2+ 26.Kc1 Qxh5 27.Nxb7 Qh1+
28.Kc2 Ba4+ 29.b3 Qa1 30.bxa4 Qxa2+ (175.989.191) 545

20/22 6:49 -0.35 19.h6 Bd7 20.Rf3 Rf8 21.hxg7 Qxg7
22.Bxh7+ Kh8 23.Qh4 Rxf3 24.Bd3+ Kg8
25.gxf3 Nxe5 26.Qd8+ Kf7 27.Qxc7 Qh6+
28.f4 Qxf4+ 29.Kb1 Qe3 (218.560.509) 534

20/22 7:51 -0.28 19.Bg6 hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Nh6 24.g4 Qe8
25.Rxc7 Qd8 26.Nb5 Qf8 27.Qh1 a6
28.Nd6 Qf4+ 29.Kd1 Qxe5 (250.340.823) 531

21/27 7:54 0.00++ 19.Bg6 hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Nh6 24.g4 Qe8
25.Rxc7 Qd8 26.Nb5 Qf8 27.g5 Qf4+
28.Kd1 Bd7 29.gxh6 Bxb5 30.Rxg7+ Kh8
31.Rh7+ Kg8 32.Rg7+ (251.973.968) 531

21/24 8:02 0.00 19.Bg6 hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Nh6 24.g4 Qe8
25.Rxc7 Qd8 26.Nb5 Qf8 27.g5 Qf4+
28.Kd1 Qf1+ 29.Kd2 Qf2+ 30.Kd1 Qf1+ (256.579.401) 531

22/24 8:39 0.00 19.Bg6 hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Nh6 24.g4 Qe8
25.Rxc7 Qd8 26.Nb5 Qf8 27.g5 Qf4+
28.Kd1 Qf1+ 29.Kd2 Qf2+ 30.Kd1 Qf1+ (275.361.051) 530

23/24 9:06 0.00 19.Bg6 hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Nh6 24.g4 Qe8
25.Rxc7 Qd8 26.Nb5 Qf8 27.g5 Qf4+
28.Kd1 Qf1+ 29.Kd2 Qf2+ 30.Kd1 Qf1+ (289.700.966) 529

24/14 9:50 +0.12++ 19.Bg6 hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Nh6 24.g4 Qe8
25.Rf2 Bd7 (312.245.884) 528

24/09 9:53 +0.24++ 19.Bg6 hxg6 20.hxg6 Bd7 21.Rh1 Rc8
22.Kb1 Ra8 23.Qh3 (313.933.279) 528

24/33 10:08 +0.47 19.Bg6 hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Nh6 24.g4 Qe8
25.Rf2 Bd7 26.g5 Nf5 27.Rxf5 exf5
28.Nxd5 Qe6 29.Nf6+ Kf8 30.Qh8+ Ke7
31.Qxg7+ Kd8 32.Qf8+ (321.750.114) 528

25/12 76:25 +0.95 19.Bg6 a6 20.Rf7 Qxf7 21.Bxf7+ Kxf7
22.h6 gxh6 23.Qf4+ Kg8 24.Qxh6 Bd7 (2.504.291.226) 546

26/04 97:30 +1.67 19.Bg6 h6 20.Nb5 Bd7 (3.222.441.227) 550

27/04 97:52 +1.67 19.Bg6 h6 20.Nb5 Bd7 (3.234.073.666) 550

28/04 98:46 +1.67 19.Bg6 h6 20.Nb5 Bd7 (3.262.461.110) 550

29/02 132:50 +1.47-- 19.Bg6 hxg6 (4.392.662.508) 551

29/08 148:14 +1.20-- 19.Bg6 hxg6 20.hxg6 Nd4 21.Rh1 Nf5
22.Qh2 Nh4 (4.915.237.592) 552

29/09 153:46 +1.47 19.Bg6 Bd7 20.Rf7 hxg6 21.Rxe7 Nxe7
22.hxg6 Rf8 23.Qg5 (5.093.144.748) 551

30/03 167:17 +1.68++ 19.Bg6 Bd7 20.Rf7 (5.449.834.138) 542

30/02 194:17 +1.37-- 19.Bg6 hxg6 (6.297.318.175) 540

30/02 269:53 +1.75++ 19.Bg6 Kh8 (8.518.596.808) 526

30/12 289:18 +1.28 19.Bg6 hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 23.Qh5 Nh6 24.Ne2 Qe8 (9.039.639.881) 520

31/08 297:30 +1.60++ 19.Bg6 hxg6 20.hxg6 Nd4 21.Rf7 Qd8
22.Qg4 Nf5 (9.228.031.590) 516

31/02 323:20 +1.12-- 19.Bg6 hxg6 (9.827.162.408) 506

31/05 380:21 +1.71 19.Bg6 h6 20.Nb5 Nd8 21.Nxc7 (11.599.577.451) 508

best move: Bd3-g6 time: 383:08.750 min n/s: 508.195 nodes: 11.682.763.399

Codechange in Stockfish 2.0, new code is:

Code: Select all

      // Extra Futility pruning
      if (   !PvNode
          && !isCheck
          && !moveIsCheck
          &&  move != ttMove
          &&  enoughMaterial
          && !move_is_promotion(move)
          && !pos.move_is_passed_pawn_push(move)
          && depth < -5*ONE_PLY
          && move_to&#40;move&#41; != move_to&#40;&#40;ss-1&#41;->currentMove&#41; // not a recapture
          && ply > 5&#41;
      &#123;
          futilityValue =  2*&#40;ss->eval + evalMargin&#41; - &#40;ss-2&#41;->eval - &#40;ss-2&#41;->evalMargin
                         + beta - bestValue
                         + pos.endgame_value_of_piece_on&#40;move_to&#40;move&#41;)
                         + &#40;move_is_ep&#40;move&#41; ? PawnValueEndgame &#58; VALUE_ZERO&#41;; // &#91;This second pass filter value should be at least FutilityMarginQS lower&#93;

          if &#40;futilityValue < alpha&#41;
          &#123;
              if &#40;futilityValue > bestValue&#41;
                  bestValue = futilityValue;
              continue;
          &#125;
      &#125;
this new block is simply added right after the other tests for futility here:

Code: Select all


    // Loop through the moves until no moves remain or a beta cutoff occurs
    while (   alpha < beta
           && &#40;move = mp.get_next_move&#40;)) != MOVE_NONE&#41;
    &#123;
      assert&#40;move_is_ok&#40;move&#41;);

      moveIsCheck = pos.move_is_check&#40;move, ci&#41;;

      // Futility pruning
      if (   !PvNode
          && !isCheck
          && !moveIsCheck
          &&  move != ttMove
          &&  enoughMaterial
          && !move_is_promotion&#40;move&#41;
          && !pos.move_is_passed_pawn_push&#40;move&#41;)
      &#123;
          futilityValue =  futilityBase
                         + pos.endgame_value_of_piece_on&#40;move_to&#40;move&#41;)
                         + &#40;move_is_ep&#40;move&#41; ? PawnValueEndgame &#58; VALUE_ZERO&#41;;

          if &#40;futilityValue < alpha&#41;
          &#123;
              if &#40;futilityValue > bestValue&#41;
                  bestValue = futilityValue;
              continue;
          &#125;
      &#125;

      // Detect non-capture evasions that are candidate to be pruned
      evasionPrunable =   isCheck
                       && bestValue > value_mated_in&#40;PLY_MAX&#41;
                       && !pos.move_is_capture&#40;move&#41;
                       && !pos.can_castle&#40;pos.side_to_move&#40;));

      // Don't search moves with negative SEE values
      if (   !PvNode
          && (!isCheck || evasionPrunable&#41;
          &&  move != ttMove
          && !move_is_promotion&#40;move&#41;
          &&  pos.see_sign&#40;move&#41; < 0&#41;
          continue;

      // Don't search useless checks
      if (   !PvNode
          && !isCheck
          &&  moveIsCheck
          &&  move != ttMove
          && !pos.move_is_capture_or_promotion&#40;move&#41;
          &&  ss->eval + PawnValueMidgame / 4 < beta
          && !check_is_dangerous&#40;pos, move, futilityBase, beta, &bestValue&#41;)
      &#123;
          if &#40;ss->eval + PawnValueMidgame / 4 > bestValue&#41;
              bestValue = ss->eval + PawnValueMidgame / 4;

          continue;
      &#125;

      // Extra Futility pruning
      if (   !PvNode
          && !isCheck
          && !moveIsCheck
          &&  move != ttMove
          &&  enoughMaterial
          && !move_is_promotion&#40;move&#41;
          && !pos.move_is_passed_pawn_push&#40;move&#41;
          && depth < -5*ONE_PLY
          && move_to&#40;move&#41; != move_to&#40;&#40;ss-1&#41;->currentMove&#41; // not a recapture
          && ply > 5&#41;
      &#123;
          futilityValue =  2*&#40;ss->eval + evalMargin&#41; - &#40;ss-2&#41;->eval - &#40;ss-2&#41;->evalMargin
                         + beta - bestValue
                         + pos.endgame_value_of_piece_on&#40;move_to&#40;move&#41;)
                         + &#40;move_is_ep&#40;move&#41; ? PawnValueEndgame &#58; VALUE_ZERO&#41;;

          if &#40;futilityValue < alpha&#41;
          &#123;
              if &#40;futilityValue > bestValue&#41;
                  bestValue = futilityValue;
              continue;
          &#125;
      &#125;

      // Update current move
      ss->currentMove = move;
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
Top
User avatar
Eelco de Groot
Posts: 4567
Joined: Sun Mar 12, 2006 2:40 am
Full name:   

Re: Double Pass Futility Pruning in Quiescence Search.

Post by Eelco de Groot »

I should have added that in Rainbow Serpent different evalMargins are used but I think the principle still stands and should work in Stockfish with some tinkering perhaps. Who knows, it might work better but it is not tested there by me.

Eelco
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
Top
peter
Posts: 3186
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: Double Pass Futility Pruning in Quiescence Search.

Post by peter »

Eelco de Groot wrote:I am sure Peter Martan can improve on this with Rebel Q3 or my version of that, Rebel Q5T 8-)
Hi Eelco!
:)
Rebel Q5T sounds very interesting!
Q3 doesn't solve this, as far as I saw till now, even if it has Bg6 soon in its calculations once and again, it doesn't keep it:

r1b3k1/ppp1q1pp/2n1p3/3pP2P/8/2NB2Q1/PPP3P1/2K2R2 w - - 0 1

Analysis by ProDeo 1.6Q3:

1.h6 Kh8 2.Df4 gxh6 3.Dxh6 Ld7 4.Th1
+- (2.18) Tiefe: 6 00:00:00 101kN
1.Lxh7+
+- (2.18) Tiefe: 6 00:00:00 143kN
1.h6 Kh8 2.Df4 gxh6 3.Dxh6 Ld7 4.Th1
+- (2.18) Tiefe: 6 00:00:00 143kN
1.h6 Ld7 2.a3 Kh8 3.Sb5 g5 4.Sxc7
+- (1.64) Tiefe: 7 00:00:00 440kN
1.h6 Ld7 2.a3 Kh8 3.Sb5 g5 4.Sxc7
+- (1.64) Tiefe: 8 00:00:00 1038kN
1.Kb1 Ld7 2.Sb5 Le8 3.a3 g6 4.hxg6
+- (1.73) Tiefe: 8 00:00:01 1514kN
1.Kb1 Ld7 2.Sb5 Le8 3.c4 dxc4 4.Le4 g6 5.hxg6
+/- (1.08) Tiefe: 9 00:00:01 2529kN
1.h6 Ld7 2.Sb5 Kh8 3.c3 g6 4.Sxc7 Tf8 5.Txf8+ Dxf8
+- (1.64) Tiefe: 9 00:00:02 3461kN
1.Lg6 +- (1.64) Tiefe: 9 00:00:03 5652kN
1.h6 Ld7 2.Sb5 Kh8 3.c3 g6 4.Sxc7 Tf8 5.Txf8+ Dxf8
+- (1.64) Tiefe: 9 00:00:03 5652kN
1.h6 Ld7 2.Sb5 Kh8 3.c3 g6 4.Sxc7 Tf8 5.Txf8+ Dxf8
+- (1.64) Tiefe: 10 00:00:03 7190kN
1.Lg6 +- (1.64) Tiefe: 10 00:00:04 7887kN
1.h6 Ld7 2.Sb5 Kh8 3.c3 g6 4.Sxc7 Tf8 5.Txf8+ Dxf8
+- (1.64) Tiefe: 10 00:00:04 7887kN
1.h6 Ld7 2.Df4 Le8 3.g4 a6 4.Dd2 Kh8
+- (1.41) Tiefe: 11 00:00:08 17767kN
1.h6 Ld7 2.Th1 g6 3.Lxg6 hxg6 4.Dxg6+ Kh8 5.Th2 Tg8 6.Dd3 Sxe5 7.De3
+/- (1.21) Tiefe: 12 00:00:38 87386kN
1.Lg6 +/- (1.21) Tiefe: 12 00:00:40 90622kN
1.h6 Ld7 2.Th1 g6 3.Lxg6 hxg6 4.Dxg6+ Kh8 5.Th2 Tg8 6.Dd3 Sxe5 7.De3
+/- (1.21) Tiefe: 12 00:00:41 90622kN
1.h6 Ld7 2.Th1 g6 3.Kd1 Tf8 4.Lxg6 hxg6 5.Dxg6+ Kh8 6.Se2 Tg8 7.Df6+ Dxf6 8.exf6 e5 9.h7 Txg2 10.f7
+/- (0.86) Tiefe: 13 00:01:54 264mN
1.Lg6 +/- (0.86) Tiefe: 13 00:01:55 266mN
1.h6 Ld7 2.Th1 g6 3.Kd1 Tf8 4.Lxg6 hxg6 5.Dxg6+ Kh8 6.Se2 Tg8 7.Df6+ Dxf6 8.exf6 e5 9.h7 Txg2 10.f7
+/- (0.86) Tiefe: 13 00:02:00 266mN
1.a3 Ld7 2.Df4 a6 3.g4 b5 4.h6 b4 5.Sb1
+/- (1.06) Tiefe: 13 00:02:39 362mN
1.a3 Ld7 2.Kb1 Tf8 3.Txf8+ Kxf8
+/= (0.47) Tiefe: 14 00:06:34 879mN
1.h6 Ld7 2.Th1 g6 3.Tf1 Tf8 4.a4 Txf1+ 5.Lxf1 Sb4 6.Df2 Lc6 7.Dg3 Kh8
+/- (0.96) Tiefe: 14 00:07:43 1043mN

Best in time is Junior with material values lowered drastically and the positional ones risen, I could call it Gambit Junior.
:)

r1b3k1/ppp1q1pp/2n1p3/3pP2P/8/2NB2Q1/PPP3P1/2K2R2 w - - 0 1

Analysis by Deep Junior 11.2 UCI5:

1.Df4 h6 2.Sb5 Dg5 3.Lg6
+- (2.10) Tiefe: 6 00:00:00 5kN
1.h6
+- (2.85) Tiefe: 6 00:00:00 10kN
1.h6 Kh8 2.Df4 Ld7 3.hxg7+ Dxg7 4.Df6 Sxe5
+- (2.88) Tiefe: 6 00:00:00 15kN
1.h6 Kh8 2.Df4 Ld7 3.Se2 Tg8 4.Tf3 gxh6
+- (2.74) Tiefe: 9 00:00:00 74kN
1.h6 Kh8 2.Th1 g6 3.Tf1 Ld7 4.Df4
+- (2.40) Tiefe: 12 00:00:00 2826kN
1.h6 Ld7 2.Th1 g6
+- (1.93) Tiefe: 15 00:00:02 51947kN
1.h6 Ld7 2.Se2 Tf8 3.Th1 g6 4.Lxg6 hxg6 5.Sf4 Txf4 6.Dxf4 Kh7 7.Tf1 d4 8.a3 Dc5
+/- (1.26) Tiefe: 17 00:00:13 329mN
1.h6 Ld7 2.Th1 g6 3.b3 Kh8 4.Th4 a5 5.Kb1 Tf8 6.Tf4 Txf4
+/- (0.79) Tiefe: 19 00:00:56 1374mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Tf4 Sf5 5.Dh3 Sh6 6.Dh5 De7 7.Kb1 Dc5 8.Se2 Ld7 9.Dh3
+/- (0.98) Tiefe: 19 00:01:25 2035mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Tf4 Sf5 5.Dh3 Sh6 6.Dh5 De7 7.Kb1 Dc5 8.Se2 Ld7 9.Sd4 Tf8 10.Dh4
+/- (0.98) Tiefe: 20 00:01:36 2311mN
1.Lg6 h6 2.Sb5 Dg5+ 3.Dxg5 hxg5 4.Lf7+ Kh7 5.Sxc7 Tb8 6.Lxe6 Lxe6 7.Sxe6 Kg8 8.Sxg5 Sxe5 9.Tf5 Te8 10.b3
+/- (1.11) Tiefe: 21 00:03:24 4950mN
1.Lg6 h6 2.Sb5 Dg5+ 3.Dxg5 hxg5 4.Lf7+ Kh7 5.Sxc7 Tb8 6.Lxe6 Lxe6 7.Sxe6 Kg8 8.Sxg5 Sxe5 9.Tf5 Te8 10.g3
+/- (1.11) Tiefe: 22 00:04:32 6544mN
1.Lg6 h6 2.Sb5 Dg5+ 3.Dxg5 hxg5 4.Lf7+ Kh7 5.Sxc7 Tb8 6.Lxe6 Lxe6 7.Sxe6 Kh6 8.Tf5 Kxh5 9.Txg5+ Kh6
+/- (1.10) Tiefe: 23 00:06:21 8982mN
1.Lg6 h6 2.Sb5 Dg5+ 3.Dxg5 hxg5 4.Lf7+ Kh7 5.Sxc7 Tb8 6.Te1 Kh6 7.Le8
+/- (1.03) Tiefe: 24 00:11:09 15598mN

As a matter of fact only Zappa Mex.2 "Aggressor Dissident" could do it in a similar time and with the probably best main variant:

r1b3k1/ppp1q1pp/2n1p3/3pP2P/8/2NB2Q1/PPP3P1/2K2R2 w - - 0 1

Analysis by Zappa Mexico IIAD:

1.h6 Sb4 2.Lxh7+ Kxh7 3.hxg7 Dxg7 4.Th1+ Kg8
+- (2.79) Tiefe: 10/26 00:00:01 662kN
1.h6 Sb4 2.Lxh7+ Kxh7 3.hxg7 Dxg7 4.Th1+
+- (2.79) Tiefe: 10/28 00:00:01 1035kN
1.h6 Ld7 2.Kb1 Kh8 3.Lxh7 gxh6 4.Ld3 Dg5
+- (2.80) Tiefe: 11/28 00:00:01 2109kN
1.h6 Ld7 2.Kb1 Kh8 3.Lxh7 gxh6 4.Ld3 Dg5
+- (2.80) Tiefe: 11/28 00:00:01 2533kN
1.h6 Ld7 2.a3 Tf8 3.Th1 g6 4.Lxg6 hxg6 5.Dxg6+
+- (2.46) Tiefe: 12/33 00:00:01 4263kN
1.h6 Ld7 2.a3 Tf8 3.Th1 g6 4.Lxg6 hxg6 5.Dxg6+
+- (2.46) Tiefe: 12/35 00:00:02 7489kN
1.h6 Ld7 2.Th1 g6 3.Lxg6 Kh8 4.Lh5 Tg8 5.Df2 Dg5+
+- (2.15) Tiefe: 13/38 00:00:02 11537kN
1.h6 Ld7 2.Th1 g6 3.Lxg6 Kh8 4.Lh5 Tg8 5.Df2 Dg5+
+- (2.15) Tiefe: 13/38 00:00:03 21529kN
1.h6 Ld7 2.Th1 g6 3.Lxg6 Kh8 4.Lh5 Tg8 5.Df2 Tf8 6.De1
+- (2.06) Tiefe: 14/38 00:00:04 24237kN
1.h6 Ld7 2.Th1 g6 3.Lxg6 Kh8 4.Lh5 Tg8 5.Df2 Tf8 6.De1
+- (2.06) Tiefe: 14/38 00:00:05 34931kN
1.h6 Ld7 2.Th1 g6 3.Lxg6 Kh8 4.Ld3 Tg8 5.Df2 Sxe5 6.Dd4 Df6
+- (2.18) Tiefe: 15/38 00:00:06 41262kN
1.h6 Ld7 2.Th1 g6 3.Lxg6 Kh8 4.Ld3 Tg8 5.Df2 Sxe5 6.Dd4 Df6
+- (2.18) Tiefe: 15/38 00:00:07 49397kN
1.h6 Ld7 2.Th1 g6 3.Lxg6 Kh8 4.Ld3 Tg8 5.De1 Sb4 6.Kb1 Sxd3 7.cxd3
+- (1.97) Tiefe: 16/39 00:00:08 63690kN
1.h6 Ld7 2.Th1 g6 3.Lxg6 Kh8 4.Ld3 Tg8 5.De1 Sb4 6.Kb1 Sxd3 7.cxd3
+- (1.97) Tiefe: 16/51 00:00:13 113mN
1.h6 Ld7 2.Th1 g6 3.Lxg6 hxg6 4.Dxg6+ Kh8 5.h7 Tf8 6.Dg3 Df7 7.Dh3 Le8 8.g4
+- (1.83) Tiefe: 17/53 00:00:16 148mN
1.h6 Ld7 2.Th1 g6 3.Lxg6 hxg6 4.Dxg6+ Kh8 5.h7 Tf8 6.Dg3 Df7 7.Dh3 Le8 8.g4
+- (1.83) Tiefe: 17/53 00:00:27 264mN
1.h6 Ld7 2.Th1 g6 3.Lxg6 hxg6 4.Dxg6+ Kh8 5.h7 Tf8 6.Dg3 Df7 7.Dh3 Sb4 8.a3 Sxc2
+- (1.53) Tiefe: 18/55 00:00:38 378mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.g3 Sh6 6.Dh5 c5 7.Tf4 d4 8.Se4
+- (1.78) Tiefe: 18/55 00:01:21 802mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.g3 Sh6 6.Dh5 c5 7.Tf4 d4 8.Se4
+- (1.78) Tiefe: 18/55 00:01:23 814mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.g3 Sh6 6.Dh5 Ld7 7.Tf1 De7 8.Th1 Tf8
+/- (1.33) Tiefe: 19/55 00:01:27 859mN
1.h6 Ld7 2.Th1 g6 3.Lxg6 hxg6 4.Dxg6+ Kh8 5.h7 Tf8 6.Kb1 Sxe5 7.Dg3 Dg7 8.Dh4 Sc4 9.Kc1
+/- (1.34) Tiefe: 19/57 00:01:56 1154mN
1.h6 Ld7 2.Th1 g6 3.Lxg6 hxg6 4.Dxg6+ Kh8 5.h7 Tf8 6.Kb1 Sxe5 7.Dg3 Dg7 8.Dh4 Sc4 9.Kc1
+/- (1.34) Tiefe: 19/57 00:02:46 1644mN
1.h6 Ld7 2.Tf3 Tf8 3.hxg7 Dxg7 4.Lxh7+ Kxh7 5.Dh2+ Dh6+ 6.Dxh6+ Kxh6 7.Txf8 a6 8.Th8+ Kg5 9.Tg8+ Kf4 10.Kd2 Sxe5 11.Se2+
+/- (1.17) Tiefe: 20/59 00:03:59 2372mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.Dh5 Sh6 6.Tf1 Ld7 7.g4 De7 8.g5 Tf8 9.Txf8+ Dxf8 10.gxh6
+- (1.58) Tiefe: 20/59 00:04:37 2789mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.Dh5 Sh6 6.Tf1 Ld7 7.g4 De7 8.g5 Tf8 9.Txf8+ Dxf8 10.gxh6
+- (1.58) Tiefe: 20/59 00:05:10 3081mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.Dh5 Sh6 6.Tf1 Ld7 7.g4 De7 8.g5 Tf8 9.Txf8+ Dxf8 10.gxh6
+- (1.68) Tiefe: 21/61 00:05:42 3393mN

And then I have to admit this was on 12 cores.

P.S. With some more MP- luck Zappa Aggressor Dissident just in time to edit made it even some faster here:

r1b3k1/ppp1q1pp/2n1p3/3pP2P/8/2NB2Q1/PPP3P1/2K2R2 w - - 0 1

Analysis by Zappa Mexico IIAD:

1.h6 Ld7 2.Kb1 Kh8 3.Lxh7 gxh6 4.Ld3 Dg5
+- (2.80) Tiefe: 11/28 00:00:01 2081kN
...
1.h6 Ld7 2.Th1 g6 3.Lxg6 hxg6 4.Dxg6+ Kh8 5.h7 Tf8 6.Dg4 Sxe5 7.Dh5 Tf5 8.Dh2 Sc4 9.Dxc7
+/- (1.29) Tiefe: 18/55 00:00:44 433mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.Dh5 Dh4 6.Dh7+ Dxh7 7.Tf8+ Kxf8 8.gxh7 Ke7 9.h8D
+- (1.45) Tiefe: 18/55 00:01:29 873mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.Dh5 Dh4 6.Dh7+ Dxh7 7.Tf8+ Kxf8 8.gxh7 Ke7 9.h8D
+- (1.45) Tiefe: 18/55 00:01:30 887mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.Dh5 Dh4 6.Dh7+ Dxh7 7.Tf8+ Kxf8 8.gxh7 Ke7 9.h8D d4 10.Se4
+- (1.59) Tiefe: 19/57 00:01:36 940mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.Dh5 Dh4 6.Dh7+ Dxh7 7.Tf8+ Kxf8 8.gxh7 Ke7 9.h8D d4 10.Se4
+- (1.59) Tiefe: 19/57 00:01:51 1078mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.Dh5 Sh6 6.Tf1 Ld7 7.g4 De7 8.g5 Tf8 9.Txf8+ Dxf8 10.gxh6
+- (1.58) Tiefe: 20/57 00:02:00 1165mN
1.Lg6 hxg6 2.hxg6 Sd4 3.Tf7 Dd8 4.Dg4 Sf5 5.Dh5 Sh6 6.Tf1 Ld7 7.g4 De7 8.g5 Tf8 9.Txf8+ Dxf8 10.gxh6
+- (1.58) Tiefe: 20/59 00:02:38 1532mN
1.Lg6 h6 2.Sb5 Dg5+ 3.Dxg5 hxg5 4.Sxc7 Tb8 5.Lf7+ Kh7 6.Sxe6 Sxe5 7.Lg6+ Sxg6 8.hxg6+ Kh6 9.Tf8 a6 10.Sd8 Kxg6 11.Sf7 Kf6
+- (1.91) Tiefe: 21/61 00:05:14 3064mN
1.Lg6 h6 2.Sb5 Dg5+ 3.Dxg5 hxg5 4.Sxc7 Tb8 5.Lf7+ Kh7 6.Sxe6 Sxe5 7.Lg6+ Sxg6 8.hxg6+ Kh6 9.Tf8 a6 10.Sd8 Kxg6 11.Sf7 Kf6
+- (1.91) Tiefe: 21/61 00:06:03 3517mN
Peter.
Top

Return to “Computer Chess Club: General Topics”