Endgame test: new file

[mercury]

Hello
I updated the file of my endgame test. You can download it in this page in my site echecs88. Most of the new positions (120-155) are from Arasan's site. Some positions are entirely new because they are from my games: 118; 120-123; 155 and two are modified from existing positions: 126; 141 .
There is also a rating list of a few engines with 2 CPU in this page.

[Howard E]

Thanks, I'll try it.

[mercury]

[D]8/4k1p1/8/5P1p/b1N4P/5PP1/1p1K4/1B1n4 w - - 0 1
Position 131
I wonder if there is a win for white after 1.g4

[bob]

[D]8/4k1p1/8/5P1p/b1N4P/5PP1/1p1K4/1B1n4 w - - 0 1
Position 131
I wonder if there is a win for white after 1.g4

Crafty believes other moves draw. Bc2 for example. But for g4 it believes white is at +2 and climbing slowly

[Eelco de Groot]

[D]8/4k1p1/8/5P1p/b1N4P/5PP1/1p1K4/1B1n4 w - - 0 1
Position 131
I wonder if there is a win for white after 1.g4

It seems to be a winning move at least to Stockfish 1.3, my own slow MSVC 32 bits build, finds 1.g4 at 14 ply, after ca. 1 million nodes. After 27 plies the PV line is:

27.01 20:25 +3.25 1.g4 hxg4 2.fxg4 Bb5 3.Na3 Ba4 4.g5 Bb3
5.Nb5 Nf2 6.Kc3 Bf7 7.Kxb2 Nd1+
8.Kc1 Ne3 9.Be4 Nc4 10.Kc2 Nd6
11.Nxd6 Kxd6 12.Kc3 Ke5 13.Bd3 Be8
14.f6 (866.018.549) 706

Now. after entering the first moves 1.g4 hxg4 2.fxg4 Bb5, with a setting that has the Passed Pawn Extension in PV nodes one halfply higher:


[D]8/4k1p1/8/1b3P2/2N3PP/8/1p1K4/1B1n4 w - -


27.01 12:48 +3.74 3.Na3 Ba4 4.g5 Bb3 5.Nc2 Nf2 6.Ne3 Bg8
7.Kc3 Nh3 8.Ng4 Kd6 9.f6 gxf6
10.Nxf6 Be6 11.Kxb2 Ke5 12.Kc3 Kf4
13.g6 Ke5 14.Nh5 Bg8 15.g7 Nf2
16.Bg6 (604.855.150) 786

28.01 31:12 +3.84 3.Na3 Ba4 4.g5 Bb3 5.Nc2 Nf2 6.Ne3 Bg8
7.Kc3 Nh3 8.Ng4 Kd6 9.Kxb2 Nf4
10.Bc2 Nh5 11.Kc3 Bf7 12.Kd4 Nf4
13.f6 g6 14.Ne3 Ne6+ 15.Kc3 Nc7
16.Bd3 (1.291.882.110) 690


This next analysis is with experimental code, based on Marco's Stockfish 1.3 version of Glaurung. It finds 1.g4 a couple of times but each time alternative moves end up just a little higher maybe because I search them a little too deep For analyzing it is better to check alternatives a bit more thorough so that is not so bad... Just having some fun with all the new functions added by Marco and Joona, and if I can remember a little bit how Glaurung's search and evaluation worked, whether it is easy to make some changes in Marco's and Joona's version:

[D]8/4k1p1/8/5P1p/b1N4P/5PP1/1p1K4/1B1n4 w - -

Engine: Ancalagon 1.3 Build 16 (Athlon 2009 MHz,256 MB)
by T. Romstad, M.Costalba, E. de Groot

2.00 0:00 +2.03 1.Ne5 Kd6 (110) 0

2.00 0:00 +2.04 1.g4 (279) 1

2.00 0:00 +2.04 1.f4 (336) 1

3.00 0:00 +2.07 1.f4 Kf6 2.Ne5 (896) 4

4.00 0:00 +1.94 1.f4 Bb3 2.Ne5 Kd6 (2.210) 10

4.00 0:00 +1.94 1.Nb6 (3.068) 14

5.00 0:00 +1.59 1.Nb6 Bb3 (5.603) 25

5.00 0:00 +1.60 1.f4 (6.164) 28

5.00 0:00 +1.60 1.Kd3 (6.966) 31

5.00 0:00 +1.61 1.g4 (8.014) 34

6.00 0:00 +1.78 1.g4 Bb3 2.Na5 Bd5 3.Kxd1 Bxf3+
4.Kc2 hxg4 5.Kxb2 (13.212) 56

6.00 0:00 +1.78 1.Kd3 (14.796) 63

6.00 0:00 +1.79 1.f4 (15.798) 67

7.00 0:00 +1.88 1.f4 Bb3 2.Na5 Ba4 3.Nb7 Kf6 4.Nc5 (22.928) 91

8.00 0:00 +2.13 1.f4 Bb5 2.Kxd1 Bxc4 3.Kc2 Kd6
4.Kxb2 Kd5 5.Kc3 (39.741) 141

9.00 0:00 +1.90 1.f4 Bb3 2.Nxb2 Nxb2 3.Kc3 Nc4
4.Kxb3 Nd2+ 5.Kb2 Nf1 6.f6+ gxf6 (55.278) 186

10.00 0:00 +1.66 1.f4 Bb3 2.Na5 Ba4 3.Nb7 Nf2 4.Nc5 Bc6
5.Kc2 Kd6 (103.642) 288

11.00 0:00 +1.37 1.f4 Bb3 2.Na5 Ba4 3.Nb7 Nf2 4.Kc3 Nd1+
5.Kd4 Bc6 6.Na5 Bf3 (234.058) 454

11.00 0:00 +1.37 1.g4 (307.244) 504

11.00 0:00 +1.38 1.Ne5 (336.873) 513

11.00 0:00 +1.38 1.Be4 (423.311) 553

11.15 0:01 +1.39 1.Kd3 (699.272) 630

12.01 0:01 +1.35 1.Kd3 Bb5 2.Kd4 Bxc4 3.Kxc4 Ne3+
4.Kb3 Nf1 5.g4 Nh2 6.gxh5 Nxf3
7.Kxb2 Nxh4 (735.570) 645

12.02 0:01 +1.35 1.Be4 (818.467) 663

12.03 0:01 +1.36 1.Ne5 (893.114) 672

12.04 0:01 +1.36 1.g4 (993.184) 683

12.07 0:01 +1.37 1.Bc2 (1.289.118) 717

13.01 0:02 +1.52 1.Bc2 Bxc2 2.Kxc2 Kf6 3.g4 Nf2
4.gxh5 Nd3 5.Nd2 Nf4 6.Ne4+ Kxf5
7.Kxb2 Ke5 8.Kc3 Nxh5 (1.501.258) 739

13.02 0:02 +1.53 1.g4 (1.614.540) 754

13.03 0:02 +1.53 1.Ne5 (1.669.798) 757

13.04 0:02 +1.54 1.Be4 (1.829.373) 770

14.01 0:02 +1.50 1.Be4 Bc6 2.Bb1 Bd5 3.Kxd1 Bxc4 4.f4 Kf6
5.Kc2 Bb5 6.Kxb2 Bd7 7.Kc3 Bxf5
8.Bd3 (2.196.540) 785

14.02 0:03 +1.51 1.Ne5 (2.538.399) 792

14.03 0:03 +1.51 1.g4 (2.779.651) 801

14.08 0:03 +1.52 1.Nb6 (3.195.968) 811

15.01 0:04 +1.56 1.Nb6 Bb3 2.g4 hxg4 3.fxg4 Kd6 4.h5 Nf2
5.g5 Ke5 6.Kc3 Bf7 7.Nd7+ Kf4 8.h6 Ne4+
9.Kxb2 gxh6 (3.448.931) 820

15.02 0:04 +1.57 1.g4 (3.743.093) 829

15.03 0:04 +1.57 1.Ne5 (3.926.813) 837

15.04 0:05 +1.58 1.Be4 (4.410.446) 847

16.01 0:06 +1.66 1.Be4 Bc6 2.Bb1 Bd5 3.Kxd1 Bxc4 4.f4 Kf6
5.Kc2 Bb5 6.Kxb2 Bd7 7.Kc3 Bxf5
8.Bxf5 Kxf5 9.Kd4 g6 (5.335.019) 857

16.02 0:06 +1.67 1.Ne5 (5.752.329) 866

16.03 0:07 +1.67 1.g4 (6.544.356) 870

17.01 0:08 +1.92 1.g4 (7.739.325) 873

18.01 0:15 +2.11 1.g4 Bb5 2.Na3 Bc6 3.Kxd1 Bxf3+
4.Ke1 Bxg4 5.Nc4 Kf6 6.Ne3 Ke5 7.Kd2 Kf4
8.Bd3 Bf3 9.Nc4 Kg4 10.Nxb2 Kxh4 (12.887.329) 846

19.01 0:24 +2.21 1.g4 Bb5 2.Na3 Ba4 3.gxh5 Bb3 4.Nb5 Nf2
5.Nc3 Nh3 6.Ne4 Ng1 7.f6+ gxf6 8.Ke3 Bg8
9.h6 Bh7 10.Nc3 f5 (21.141.245) 846

20.01 0:37 +2.43 1.g4 Bb5 2.Na3 Ba4 3.gxh5 Bb3 4.Nb5 Nf2
5.Nc3 Nh3 6.Ne4 Ng1 7.f6+ gxf6 8.Ke3 Bf7
9.h6 Bg6 10.Nd2 Bxb1 11.Nxb1 Ke6 (31.751.460) 854

21.01 1:07 +2.96 1.g4 (56.812.041) 846

22.01 1:36 +3.00 1.g4 hxg4 2.fxg4 Bb5 3.Kxd1 Bxc4
4.Kc2 Kf6 5.Kxb2 Be2 6.g5+ Ke5 7.Kc3 Bg4
8.f6 gxf6 9.g6 Be6 10.h5 f5 11.Bd3 Kf6
12.Kd4 Kg7 13.Ke5 (80.497.689) 837

23.01 2:58 +3.17 1.g4 hxg4 2.fxg4 Bb5 3.Kxd1 Bxc4
4.Kc2 Kf6 5.Kxb2 Be2 6.g5+ Ke5 7.Kc3 Bg4
8.f6 gxf6 9.g6 Be6 10.h5 f5 11.Bd3 Kf6
12.Kd4 Bb3 13.Ke3 Bg8 (145.972.313) 818

24.01 5:14 +3.19 1.g4 hxg4 2.fxg4 Bb3 3.Na5 Ba4 4.g5 Kf8
5.Nc4 Kg8 6.Ne5 Nf2 7.Nd3 Nxd3
8.Kxd3 Kf8 9.Kc3 Ke7 10.Kxb2 Bc6
11.h5 Bf3 12.h6 gxh6 13.gxh6 (254.050.236) 806

25.01 9:15 +3.23 1.g4 hxg4 2.fxg4 Bb5 3.Kxd1 Bxc4
4.Kc2 Kf6 5.Kxb2 Ke5 6.Kc3 Bg8 7.Bd3 Kf4
8.Be2 Bd5 9.Kd4 Be4 10.h5 Bc6 11.Kc5 Be4
12.Kd6 Ke3 13.Bc4 Kd4 (440.755.531) 794

26.01 14:10 +3.03 1.g4 hxg4 (666.346.719) 783

27.01 35:34 +3.41 1.g4 hxg4 2.fxg4 Bb5 3.Na3 Ba4 4.g5 Bb3
5.Nb5 Nf2 6.Kc3 Bf7 7.Kxb2 Nd1+
8.Ka3 Bc4 9.h5 Kf8 10.Nd4 Bf7 11.g6 Nc3
12.Bd3 Bd5 13.f6 gxf6 14.h6 (1.621.854.172) 759


Changes are mainly in Rootsearch and I'm trying to avoid inaccurate nullwindow-nullmove searches that only take milliseconds to return, no matter what the depth.

Regards, Eelco

[mcostalba]

Just having some fun with all the new functions added by Marco and Joona, and if I can remember a little bit how Glaurung's search and evaluation worked, whether it is easy to make some changes in Marco's and Joona's version:

This is nice to hear!

Easiness of reading and modifying is one of my main goals. Well Glaurung is very easy too, and I hope I have kept that feature also in Stockfish.

[mercury]

[D] 8/8/3k4/7p/p2b1p2/5PqP/1P6/2N2Q1K b - - 0 1
position 141
How long for engines to find 1...Qf2!
In 5 minutes only rybka 3 found it on my computer

[Eelco de Groot]

[D] 8/8/3k4/7p/p2b1p2/5PqP/1P6/2N2Q1K b - - 0 1
position 141
How long for engines to find 1...Qf2!
In 5 minutes only rybka 3 found it on my computer

Hello Ludovic,

I suppose it depends on how fast your computer is whether Ancalagon can match that:


[D]8/8/3k4/7p/p2b1p2/5PqP/1P6/2N2Q1K b - -

Engine: Ancalagon 1.3 WeakSquares 150 Build 19 (256 MB)
by T. Romstad, M.Costalba, J.Kiiski, E. de Groot

12.01 0:03 +1.60 1...Bf2 2.Qg2 Qxg2+ 3.Kxg2 Bd4 4.Nd3 Kd5
5.Kf1 Kc4 6.Ke2 Kb3 7.Nxf4 Kxb2
8.Nxh5 (2.277.259) 759

13.01 0:03 +2.50 1...Bf2 2.Qg2 Qxg2+ 3.Kxg2 Bd4 4.Nd3 Kd5
5.Kf1 Kc4 6.Ke2 Kb3 7.Kd2 Bxb2
8.Nxb2 Kxb2 9.Kd3 (2.792.141) 780

14.01 0:05 +2.60 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qc6+ Kg5
5.Qd5+ Kh4 6.Qd8+ Kxh3 7.Qd7+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Be3
11.Kg2 Kf5 12.Nc3 (4.069.046) 808

15.01 0:06 +2.33 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qc6+ Kg5
5.Qd5+ Kh4 6.Qd8+ Kxh3 7.Qd7+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Kf5
11.Kg2 Bc5 12.Nc3 a3 13.bxa3 Bxa3 (5.242.040) 830

16.01 0:09 +2.33 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qc6+ Kg5
5.Qd5+ Kh4 6.Qd8+ Kxh3 7.Qd7+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Kf5
11.Kg2 Bc5 12.Nc3 a3 13.bxa3 Bxa3 (7.923.409) 862

17.01 0:13 +2.25 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qc6+ Kg5
5.Qd5+ Kh4 6.Qd8+ Kxh3 7.Qd7+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Kf5
11.Kg2 Be1 12.Nd4+ Ke5 13.Nc6+ Ke6
14.Kh3 Bf2 (12.074.653) 878

18.01 0:24 +2.33 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qc6+ Kg5
5.Qd5+ Kh4 6.Qd8+ Kxh3 7.Qd7+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Be1
11.Kg2 Kf5 12.Nd4+ Ke5 13.Nc6+ Kd5
14.Ne7+ Kd6 (21.846.841) 896

19.01 0:43 +2.37 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qc6+ Kg5
5.Qd5+ Kh4 6.Qd8+ Kxh3 7.Qd7+ Kh4
8.Qd8+ Qg5 9.Qd1 Qf5 10.Kg2 Be3
11.Qd8+ Qg5+ 12.Qxg5+ Kxg5 13.Ne2 Bd2
14.Kh3 h4 (38.582.044) 891

20.01 1:11 +2.56 1...Bf2 2.Qg2 (63.787.628) 886

21.01 1:30 +2.27 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qa6+ Kg5
5.Qa5+ Kh4 6.Qd8+ Kxh3 7.Qd7+ Kh4
8.Qd8+ Qg5 9.Qd1 Qf5 10.Kg2 Be3
11.Qe1+ Kg5 12.Qc3 Qg6 13.Ne2 Kh4+
14.Kf1 Qb1+ (80.215.197) 884

21.29 5:58 +2.27 1...Qf2 (299.639.722) 836

22.01 6:53 +3.64 1...Qf2 (344.372.890) 832

23.01 8:15 +3.68 1...Qf2 2.Qxf2 Bxf2 3.Kg2 Bd4 4.Nd3 Ke7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kd6 8.Na2 Kd5
9.Kd3 Bd4 10.Nc3+ Kc5 11.Na2 h4
12.Nc1 Kd5 13.Na2 Be5 14.Nc3+ Kc5 (410.099.102) 827

24.01 14:34 +3.72 1...Qf2 2.Qxf2 Bxf2 3.Ne2 Kc5 4.Nxf4 Bh4
5.Nd3+ Kd4 6.Nc1 Ke3 7.b4 Be7 8.Na2 Kd2
9.b5 Bc5 10.f4 Kc2 11.Kg2 Kb2 12.b6 Bxb6
13.Nb4 a3 14.Kf3 Kb3 (716.381.221) 819

25.01 29:43 +3.94 1...Qf2 2.Qxf2 Bxf2 3.Nd3 Bd4 4.Kg2 Kc7
5.Kf1 Bxb2 6.Nb4 a3 7.Ke2 Kb6 8.Kd3 Kb5
9.Na2 Kc5 10.Kc2 Kd4 11.Kd2 Kc4
12.Kc2 Bd4 13.Nc1 h4 14.Na2 (1.433.196.861) 803

26.01 78:31 +3.84 1...Qf2 2.Qxf2 Bxf2 3.Kg2 Bd4 4.Nd3 Kc7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kb6 8.Kd3 Kb5
9.Na2 Kc5 10.Kc2 Kc4 11.Nc1 Bd4
12.Kb1 Be3 13.Kc2 Bc5 14.Na2 Bd4 (3.716.772.961) 788

27.01 128:12 +3.94 1...Qf2 2.Qxf2 Bxf2 3.Nd3 Bd4 4.Kg2 Kc7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kd6 8.Kd3 Kc5
9.Na2 Bf6 10.Nc1 Kb4 11.Kc2 Kc4
12.Na2 Be5 13.Nc1 Bd4 14.Na2 h4 (6.108.421.391) 794

Eelco

[Eelco de Groot]

It was a long wait to complete the analysis for 29 plies but the final eval I think was worth some waiting, possibly this is actually 30/31 plies so instead of 320 minutes Ancalagon did another ply and that cost another doubling of time. Weak Squares is slightly altered but it remains to be seen if it is worth anything and what would be a good tuning of the weight, the version with Weak Squares set to zero did also very well here. So considering the bitboard operations needed for calculating weak squares, I'm not sure it is really paying off in terms of strength.

A question one may ask; in order to make progress in positional play, it is possible that there are timecontrols where it is not possible to add new positional rules without losing some elo? It depends on the timecontrol you are aiming at possibly? if that is true there is something to be said for doing a lazy eval that depends on the timecontrol. But maybe this is nonsense. Or too complicated to be practical

Anyway, the new build managed to find Qf2 one ply earlier and is now within 5 minutes !

8/8/3k4/7p/p2b1p2/5PqP/1P6/2N2Q1K b - -

Engine: Ancalagon 1.3 Weak Squares 150 Build 20 (Athlon 2009 MHz, 256 MB)
by T. Romstad, M.Costalba, J. Kiiski, E. de Groot

12.01 0:02 +1.41 1...Bxb2 (2.095.438) 775

13.01 0:03 +1.11 1...Bxb2 2.Qa6+ Kd5 3.Qb7+ Ke6
4.Qc6+ Kf7 5.Qc4+ Kf6 6.Qc6+ Kg5
7.Qc5+ Kg6 8.Qc2+ Kh6 9.Qxb2 Qxh3+
10.Kg1 Qg3+ 11.Kf1 Qxf3+ 12.Ke1 Qg3+
13.Kf1 (2.529.360) 793

13.05 0:05 +1.12 1...Bf2 (4.660.262) 810

14.01 0:07 +2.49 1...Bf2 2.Qg2 Qxg2+ 3.Kxg2 Bd4 4.Nd3 Kd5
5.Kf1 Kc4 6.Ke2 Kb3 7.Kf1 Bxb2
8.Nc5+ Kb4 9.Nd3+ Kc3 10.Nxf4 (6.331.692) 825

15.01 0:09 +2.33 1...Bf2 2.Qd3+ Ke6 3.Qc4+ Kf6 4.Qc3+ Kg5
5.Qe5+ Kh4 6.Qf6+ Kxh3 7.Qf5+ Kh4
8.Qf6+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Kf5
11.Kg2 Bc5 12.Nc3 a3 13.bxa3 Bxa3 (8.351.383) 837

16.01 0:12 +2.33 1...Bf2 2.Qd3+ Ke6 3.Qc4+ Kf6 4.Qc3+ Kg5
5.Qe5+ Kh4 6.Qf6+ Kxh3 7.Qf5+ Kh4
8.Qf6+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Kf5
11.Kg2 Bc5 12.Nc3 a3 13.bxa3 Bxa3 (10.969.186) 852

17.01 0:17 +2.25 1...Bf2 2.Qd3+ Ke6 3.Qc4+ Kf6 4.Qc3+ Kg5
5.Qe5+ Kh4 6.Qf6+ Kxh3 7.Qf5+ Kh4
8.Qf6+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Kf5
11.Kg2 Be1 12.Nd4+ Ke5 13.Nc6+ Kd5
14.Ne7+ Ke6 (14.959.590) 867

18.01 0:27 +2.33 1...Bf2 2.Qd3+ Ke6 3.Qc4+ Kf6 4.Qc3+ Kg5
5.Qe5+ Kh4 6.Qf6+ Kxh3 7.Qf5+ Kh4
8.Qf6+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Be1
11.Kg2 Kf5 12.Nd4+ Ke5 13.Nc6+ Kd5
14.Ne7+ Kd6 (24.250.709) 884

19.01 0:44 +2.29 1...Bf2 2.Qd3+ Ke6 3.Qc4+ Kf6 4.Qc3+ Kg5
5.Qe5+ Kh4 6.Qe7+ Kxh3 7.Qd7+ Kh4
8.Qd8+ Qg5 9.Qd1 Be3 10.Ne2 Qf5
11.Qd8+ Kh3 12.Qd1 Qb5 13.Qc2 Qg5
14.Qc8+ Kh4 (39.717.188) 883

20.01 1:21 +2.31 1...Bf2 2.Qd3+ Ke6 3.Qc4+ Kf6 4.Qc3+ Kg5
5.Qe5+ Kh4 6.Qe7+ Kxh3 7.Qd7+ Kh4
8.Qd8+ Qg5 9.Qd1 Qf5 10.Kg2 Be3
11.Qd8+ Qg5+ 12.Qxg5+ Kxg5 13.Ne2 Bd2
14.Kh3 Be1 (72.202.078) 885

20.27 3:49 +2.31 1...Qf2 (194.674.086) 849

21.01 4:34 +2.51 1...Qf2 (231.876.943) 843

22.01 5:58 +3.32 1...Qf2 (300.050.471) 835

23.01 8:06 +3.82 1...Qf2 2.Qxf2 Bxf2 3.Kg2 Bd4 4.Nd3 Kc7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kb6 8.Kd3 Kb5
9.Na2 Kc5 10.Kc2 Kc4 11.Kb1 Be5
12.Kc1 Kb3 13.Kb1 (403.574.888) 830

24.01 17:19 +3.86 1...Qf2 2.Qxf2 Bxf2 3.Nd3 Bd4 4.Kg2 Kc7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kb6 8.Na2 Kb5
9.Kd3 Kc5 10.Kc2 Kc4 11.Kb1 Kd4
12.Nb4 Ke3 13.Nd5+ Kxf3 14.Ka2 (848.847.125) 816

25.01 29:16 +4.05 1...Qf2 2.Qxf2 Bxf2 3.Nd3 Bd4 4.Kg2 Kc7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kb6 8.Na2 Kc5
9.Kd3 Kb5 10.Kc2 Kc4 11.Kb1 Kd4
12.Nb4 Ke3 13.Ka2 Kxf3 14.Nd3 (1.424.265.520) 810

26.01 61:37 +3.84 1...Qf2 2.Qxf2 Bxf2 3.Kg2 Bd4 4.Nd3 Kc7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kb6 8.Na2 Kc5
9.Kd3 Bf6 10.Kc2 Kc4 11.Kb1 Bd4
12.Nc1 Be3 13.Na2 Bc5 14.Kc2 Bd4 (2.976.015.881) 804

27.01 129:21 +4.47 1...Qf2 2.Qxf2 Bxf2 3.Nd3 Bd4 4.Kg2 Kc7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kb6 8.Kd1 Kb5
9.Na2 Kc4 10.Kc2 Kd4 11.Kb3 Ke3
12.Nb4 Kxf3 13.Nc2 Ke4 14.Nxa3 (6.065.396.761) 781

28.01 160:26 +4.96 1...Qf2 2.Qxf2 Bxf2 3.Kg2 Bd4 4.Nd3 Kc7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kb6 8.Na2 Kc6
9.Kd2 Kd5 10.Kd3 Kc5 11.h4 Bf6
12.Nc1 Kd5 13.Kc2 Kc4 14.Kb1 Bxh4 (7.466.507.814) 775

29.01 592:31 +6.49 1...Qf2 2.Qxf2 Bxf2 3.Kg2 Bd4 4.b3 Be3
5.bxa4 Bxc1 6.a5 Be3 7.a6 Kc6 8.Kf1 Kb5
9.Ke2 Kxa6 10.Kd3 Kb5 11.Ke4 Kc4
12.h4 Kc3 13.Kf5 Kd3 14.Kg5 Ke2 (23.955.023.268) 673


With Weak Squares calculated but not weighted, 6 minutes:

8/8/3k4/7p/p2b1p2/5PqP/1P6/2N2Q1K b - -

Engine: Ancalagon 1.3 Weak Squares 0 Build 19 (256 MB)
by T. Romstad, M.Costalba, J. Kiiski, E. de Groot

12.01 0:01 +1.52 1...Bxb2 2.Qa6+ (903.163) 704

13.01 0:02 +1.15 1...Bxb2 2.Qa6+ Kd5 3.Qb5+ Ke6
4.Qc6+ Kf5 5.Qc2+ Kg5 6.Qc5+ Kg6
7.Qc2+ Kh6 8.Qxb2 Qxf3+ 9.Qg2 Qd1+
10.Qg1 Qxg1+ 11.Kxg1 Kg5 12.Kg2 Kf5 (2.252.004) 787

13.09 0:04 +1.16 1...Bf2 (3.825.703) 800

14.01 0:06 +1.89 1...Bf2 (5.159.650) 817

15.01 0:06 +2.33 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qa6+ Kg5
5.Qa5+ Kh4 6.Qd8+ Kxh3 7.Qc8+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Kf5
11.Kg2 Bc5 12.Nc3 a3 13.bxa3 Bxa3 (5.628.961) 822

16.01 0:10 +2.33 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qa6+ Kg5
5.Qa5+ Kh4 6.Qd8+ Kxh3 7.Qc8+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Kf5
11.Kg2 Bc5 12.Nc3 a3 13.bxa3 Bxa3 (8.457.352) 843

17.01 0:15 +2.35 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qa6+ Kg5
5.Qa5+ Kh4 6.Qd8+ Kxh3 7.Qc8+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Be3
11.Kh2 Bd2 12.Kh3 h4 13.Nd4 Be3
14.Ne6+ Kf5 (12.901.011) 859

18.01 0:24 +2.33 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qa6+ Kg5
5.Qa5+ Kh4 6.Qd8+ Kxh3 7.Qc8+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Be1
11.Kg2 Kf5 12.Nd4+ Ke5 13.Nc6+ Kd5
14.Ne7+ Kd6 (21.507.655) 875

19.01 0:41 +2.41 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qa6+ Kg5
5.Qa5+ Kh4 6.Qd8+ Kxh3 7.Qc8+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Be1
11.Kg2 Kf5 12.Nd4+ Ke5 13.Nc6+ Kd5
14.Ne7+ Kd6 (36.686.661) 887

20.01 1:10 +2.29 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qa6+ Kg5
5.Qa5+ Kh4 6.Qd8+ Kxh3 7.Qc8+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Be1
11.Kg2 Kf5 12.Kh2 Bb4 13.Kh3 Bd6
14.Kh4 Be5 (62.661.734) 886

21.01 2:17 +2.31 1...Bf2 2.Qa6+ Ke7 3.Qb7+ Kf6 4.Qa6+ Kg5
5.Qa5+ Kh4 6.Qd8+ Kxh3 7.Qd7+ Kh4
8.Qd8+ Qg5 9.Qxg5+ Kxg5 10.Ne2 Be1
11.Kh2 Kf5 12.Kg2 h4 13.Kh3 Kg5
14.Nd4 Bf2 (121.862.277) 884

21.29 6:00 +2.31 1...Qf2 (301.836.056) 837

22.01 6:18 +2.51 1...Qf2 (317.530.598) 838

23.01 8:12 +3.32 1...Qf2 (408.618.568) 829

24.01 12:15 +3.84 1...Qf2 2.Qxf2 Bxf2 3.Kg2 Bd4 4.Nd3 Kc7
5.Kf1 Bxb2 6.Nb4 a3 7.Ke2 Kd6 8.Kd2 Kc5
9.Na2 Kc4 10.Kc2 Bd4 11.Kb1 Bf6
12.Kc2 Be5 13.Kb1 Bd4 (604.369.452) 822

25.01 32:42 +3.90 1...Qf2 2.Qxf2 Bxf2 3.Nd3 Bd4 4.Kg2 Kc7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kb6 8.Kd3 Kc5
9.Na2 Kb5 10.Kc2 Kc4 11.Nc1 Bd4
12.Na2 h4 13.Kb1 Kb3 14.Nc1+ Ka4 (1.579.938.952) 805

26.01 65:51 +4.30 1...Qf2 2.Qxf2 Bxf2 3.Nd3 Bd4 4.Kg2 Kc7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kb6 8.Kd3 Kc5
9.Na2 h4 10.Kc2 Kc4 11.Nc1 Bh8
12.Na2 Bd4 (3.190.462.640) 807

27.01 90:34 +4.58 1...Qf2 2.Qxf2 Bxf2 3.Nd3 Bd4 4.Kg2 Kc7
5.Kf1 Bxb2 6.Ke2 a3 7.Nb4 Kb6 8.Kd3 Kc5
9.Na2 Kb5 10.Kc2 Kc4 11.Kb1 Kd4
12.Nb4 Ke3 13.Ka2 Kxf3 14.Nc2 Ke4 (4.357.973.901) 801

28.01 225:01 +5.65 1...Qf2 2.Qxf2 Bxf2 3.Kg2 Bd4 4.b3 Be3
5.bxa4 Bxc1 6.a5 Kc6 7.Kf1 Kb5 8.Ke2 Kxa5
9.Kd3 Kb4 10.Kc2 Be3 11.Kd3 Kc5
12.Ke4 Kc4 13.Ke5 Kd3 14.Kd5 Bd4 (10.752.470.792) 796

best move: Qg3-f2 time: 267:40.110 min n/s: 795.211 nodes: 12.771.180.121


I am not sure this is not at least as good as with Weak Squares at 150, if you consider that not doing the calculation makes the engine faster.

Eelco

[MattieShoes]

A question one may ask; in order to make progress in positional play, it is possible that there are timecontrols where it is not possible to add new positional rules without losing some elo? It depends on the timecontrol you are aiming at possibly? if that is true there is something to be said for doing a lazy eval that depends on the timecontrol. But maybe this is nonsense. Or too complicated to be practical

Engines can certainly be tuned for different time controls, so I guess in theory, they could be tuned with different evals at different time controls... Dr. Hyatt recently posted in the programming forum about how changing piece values in crafty resulted in an increase in strength at very fast time controls but a decrease in strength at slower ones. (I think it was increasing R/Q values) Also, I think tuning the aggressiveness of pruning in search may depend on time control as well.

I don't know if any engines actively change their search/eval based on time control but it's certainly feasable.

One could also change eval based on how far from the root node it is. Of course, this might introduce continuity problems...