Andscacs - New version 0.92

[cdani]

Hello!
I published a new version, 0.92.

http://www.andscacs.com/downloads/andscacs092.zip

Changes:

* Improved various evaluation parameters in general, specially about attacks.
* Speed and move ordering improvements.
* Solved little bugs.
* Improvement of various imbalances.
* Improvements in some endgames.
* Added option to save hash.

Should be clearly stronger:

Code: Select all

Test with 1000 seconds + 1:
 # PLAYER              : RATING  ERROR   POINTS  PLAYED    (%)
 1 Andscacs 0.91295    : 2874.0    7.7    431.0     782   55.1%
 2 Andscacs 0.91       : 2838.0    7.7    351.0     782   44.9%

No BMI2 version this time.
I plan to try to do a Ryzen optimized version in a few days.

I will not publish the detailed changes anytime soon, as I will wait at least that Andscacs is out of TCEC. Sure you will understand
But I can comment that no Stockfish changes worked. Also that I improved speed nicely thanks to templates.
One interesting change that I want to comment, but interested people will have to work the details for themselves, is that it now saves two moves to hash. Not that was a big winner, just another little one.

[Tony P.]

Congrats on the release and the strength improvement!

One interesting change that I want to comment, but interested people will have to work the details for themselves, is that it now saves two moves to hash.

This sounds excellent! Have you ever experimented with saving more than 2 moves per position? If so, did it work well?

[cdani]

Congrats on the release and the strength improvement!

One interesting change that I want to comment, but interested people will have to work the details for themselves, is that it now saves two moves to hash.

This sounds excellent! Have you ever experimented with saving more than 2 moves per position? If so, did it work well?

Thanks. No, I have only tried with 2. I have not enough room for more in the current hash.

[Damir]

Thanks a lot for a much improved version Dani

[Tony P.]

It's a bit strange that Andscacs promoted to a rook in the following game (5'+3'' time control, resignation after -5 eval for 3 moves in a row), though it didn't influence the result at all. (Of course it's far not the only engine that's prone to underpromotion when the new piece is going to be taken on the next ply anyway.)

[pgn][Event "Computer chess game"]
[Date "2017.09.17"]
[White "Fizbo19x64"]
[Black "Andscacs 0.92"]
[Result "1-0"]
[BlackElo "3000"]
[ECO "A00"]
[Opening "Polish (Sokolsky-Orangutan) Opening"]
[Time "22:09:25"]
[Variation "Outflank Variation, 1...c6"]
[WhiteElo "3000"]
[TimeControl "300+3"]
[Termination "normal"]
[PlyCount "94"]
[WhiteType "program"]
[BlackType "program"]

1. b4 c6 2. e3 d5 3. Bb2 Qb6 4. a3 a5 5. c4 axb4 6. c5 Qc7 7. axb4 {(7.axb4
Rxa1 8.Bxa1 Bf5 9.Nf3 Nd7 10.Be2 e5) +0.68/16 9} Rxa1 {(7. ... Rxa1 8.Bxa1
e5 9.Be2 Nf6 10.Nf3 Nbd7 11.d4 e4 12.Nfd2 Be7 13.Nc3 0-0 14.0-0 b6 15.Bb2
Bb7 16.Qc2 Ra8 17.Ra1 Rxa1+ 18.Bxa1 bxc5 19.bxc5 Qa5) -0.27/21 9} 8. Bxa1
{(8.Bxa1) +0.68/1} e5 {(8. ... e5 9.Nf3 Nd7 10.Nc3 Ngf6 11.d4 e4 12.Nd2 b6
13.Bb2 Be7 14.Be2 0-0 15.0-0 Bb7 16.Qc2 Ra8 17.Ra1 Rxa1+ 18.Bxa1 Nf8 19.Bb2
Ne6 20.Qd1) -0.22/21 17} 9. Nf3 {(9.Nf3 Nd7 10.Be2 Be7 11.h3 Nh6 12.0-0 f6)
+0.76/17 8} Nd7 {(9. ... Nd7 10.Nc3 Be7 11.d4 e4 12.Nd2 b6 13.Bb2 bxc5
14.bxc5 Nh6 15.Qb3 0-0 16.Be2 Bb7 17.0-0 Ra8 18.Ra1 Rxa1+ 19.Bxa1 f5 20.Qd1
Nf6 21.Nf1 g6 22.Bb2 Kg7) -0.22/22 10} 10. Be2 {(10.Be2 Be7 11.h3 Nh6
12.Qc2 f6 13.0-0 0-0) +0.72/17 14} Nh6 {(10. ... Nh6 11.0-0 Be7 12.Nc3 0-0
13.e4 d4 14.Na4 f6 15.Bc4+ Nf7 16.Be6 Nb8 17.Bc4 Nd7) -0.01/20 7} 11. Nc3
{(11.Nc3 Be7 12.Na4 f6 13.Bc3 b5 14.cxb6 Nxb6) +0.61/18 14} Be7 {(11. ...
Be7 12.0-0 f5 13.Bb2 0-0 14.Na4 f4 15.Qa1 e4 16.Nd4 Rf6 17.f3 fxe3 18.dxe3
Ne5 19.Nb6 exf3 20.Nxf3 Nxf3+ 21.Bxf3 Ng4 22.Bxg4 Rxf1+ 23.Kxf1 Bxg4)
-0.21/20 12} 12. Na4 {(12.Na4 f6 13.Bc3 b5 14.cxb6 Nxb6 15.h3 Nxa4)
+0.58/17 11} O-O {(12. ... 0-0 13.0-0 e4 14.Nd4 b5 15.cxb6 Nxb6 16.Nxb6
Qxb6 17.b5 cxb5 18.Bxb5 Ng4 19.Qb3 Qd6 20.f4 exf3 21.Nxf3 Bd7 22.Bxd7 Qxd7
23.Bd4 Ra8 24.h3) -0.18/19 8} 13. O-O {(13.0-0 f6 14.Nxe5 Qxe5 15.b5 cxb5
16.Bxb5 Bxc5) +0.68/17 23} e4 {(13. ... e4 14.Nd4 b5 15.cxb6 Nxb6 16.Nxb6
Qxb6 17.Qc2 Bd7 18.Bc3 Ng4 19.h3 Nf6 20.Ra1 Qb7 21.Ra6 Rb8 22.Nxc6 Bxc6
23.Bxf6 Bxf6 24.Qxc6 Qxc6 25.Rxc6) -0.11/20 21} 14. Nd4 {(14.Nd4 Ne5 15.Nb6
Bg4 16.Nb5 Bxe2 17.Nxc7 Bxd1) +0.74/18 13} b5 {(14. ... b5 15.Nc3 Ne5
16.Ndxb5 Qb8 17.Nd4 Qxb4 18.Na4 Qb8 19.h3 Qa7 20.Bc3 Bxc5 21.Nxc5 Qxc5
22.Qc2 Re8 23.Ra1 Qd6 24.Ra8 Bd7 25.Rxe8+ Bxe8 26.f4 exf3 27.Nxf3 f6
28.Nd4) -0.17/20 7} 15. cxb6 {(15.cxb6 Nxb6 16.Qc2 Nxa4 17.Qxa4 Bd7 18.Rc1
c5) +0.88/18 14} Nxb6 {(15. ... Nxb6 16.Nxb6 Qxb6 17.Qb3 Nf5 18.b5 Nxd4
19.Bxd4 c5 20.Be5 Bb7 21.Ra1 f6 22.Bc3 Kh8 23.Qb2 Qe6 24.d4 c4 25.Bb4 Rc8
26.Bxe7 Qxe7 27.Ra7 c3) -0.25/21 22} 16. Bc3 {(16.Bc3 Bd6 17.Nxb6 Qxb6
18.d3 Bxb4 19.Bxb4 Qxb4) +1.06/19 6} Nf5 {(16. ... Nf5 17.d3 Nxa4 18.Qxa4
Bd6 19.dxe4 Nxd4 20.Bxd4 Bxh2+ 21.Kh1 dxe4 22.Qa1 f6 23.g3 Bxg3 24.fxg3
Qxg3 25.Rf4 Qh3+ 26.Kg1 Qg3+ 27.Kh1) -0.01/20 13} 17. Nxb6 {(17.Nxb6 Qxb6
18.Qa1 Nxd4 19.Bxd4 Qxb4 20.Bxg7 Re8) +1.02/17 22} Qxb6 {(17. ... Qxb6
18.Qa1 f6 19.Qa5 Qb7 20.b5 Nxd4 21.Bxd4 cxb5 22.Qxb5 Qxb5 23.Bxb5 Be6
24.Ra1 Bb4 25.d3 Rb8 26.Ba6 Bd6 27.h3 Kf8 28.Kf1 Kf7 29.Ke2 Kg6 30.Kd2 Kg5)
+0.13/22 7} 18. Nxf5 {(18.Nxf5 Bxf5 19.Qa1 g6 20.Rb1 Be6 21.h3 Qc7)
+0.87/21 22} Bxf5 {(18. ... Bxf5 19.Qa4 Be6 20.b5 cxb5 21.Qxb5 Qxb5 22.Bxb5
Rb8 23.Bc6 Rb6 24.Ba8 Bf6 25.Bxf6 gxf6 26.f3 f5 27.Kf2 Kg7 28.f4 Rb2 29.Ke2
Kf6 30.Bc6 Rc2) +0.02/20 8} 19. Qa1 {(19.Qa1 g6 20.Rb1 Be6 21.h3 Qc7 22.Bd4
Rb8) +0.87/21 12} f6 {(19. ... f6) +0.36/19 7} 20. Rb1 {(20.Rb1 Qb7 21.Ba6
Qc7 22.b5 Rb8 23.bxc6 Rxb1+) +0.91/20 12} Rb8 {(20. ... Rb8 21.b5 c5 22.Qa5
Qb7 23.b6 Ra8 24.Qb5 Bd7 25.Qb2 Bf8 26.f4 exf3 27.Bxf3 Be6 28.Qb3 Bd6
29.Kh1 Bf7 30.Kg1 Kh8 31.Ra1) +0.23/20 14} 21. b5 {(21.b5 Bc5 22.Rb2 Qc7
23.Qa2 Bd7 24.h3 Qa7) +0.99/16 11} c5 {(21. ... c5 22.Qa2 Qb7 23.b6 Ra8
24.Qc2 Bd6 25.f4 Be6 26.Qb2 Bc8 27.Qb5 Bd7 28.Qb2 Bf8 29.Qc2 Be6 30.Qd1 Bf7
31.Bg4 Bd6) +0.18/21 10} 22. Qa2 {(22.Qa2 Be6 23.f3 exf3 24.Bxf3 Qd6 25.Qa7
Bd8) +0.88/17 12} Qb7 {(22. ... Qb7 23.b6 Ra8 24.Qc2 Bd8 25.Qb3 Rb8 26.Ba5
Be7 27.Bc3 Bf8 28.Qb5 Ra8 29.Ra1 Bd7 30.Qb1 Rxa1 31.Bxa1 Bd6) +0.18/20 9}
23. b6 {(23.b6 Ra8 24.Qc2 Be6 25.h3 Bd8 26.Qb2 Rb8) +1.25/17 6} Ra8 {(23.
... Ra8 24.Qb3 h6 25.f4 Be6 26.Qd1 Bd8 27.Qb3 Bd7 28.Qb2 Be8 29.Qb3 Ba4
30.Qb2 Bc6 31.Qb3 Kh8 32.h3 Rb8 33.Ba5 Be7 34.Qc2) +0.15/22 8} 24. Qb2
{(24.Qb2 Bd7 25.Bb5 Bd8 26.h3 Rb8 27.Ba5 h6) +1.10/18 14} Bg6 {(24. ... Bg6
25.Qb3 Bd6 26.Ra1 Rb8 27.Rb1 Be8 28.Qc2 Qc6 29.Qa2 Ra8 30.Qb2 Rb8 31.Qa2)
+0.01/21 8} 25. Bg4 {(25.Bg4 h5 26.Be6+ Bf7 27.Bxf7+ Kxf7 28.Qb5 g6)
+1.27/18 9} Kh8 {(25. ... Kh8 26.Be2 Be8 27.d3 Bg6 28.dxe4 Bxe4 29.Rc1 Kg8
30.Qb5 Rb8 31.Ba5 Bd6 32.Rxc5 Bxc5 33.Qxc5 Rc8 34.Qa3 Bg6 35.Kf1 Be4
36.Kg1) +0.01/23 10} 26. Be2 {(26.Be2 h6 27.h3 Bd8 28.Bf1 Rb8 29.Ba5 c4)
+1.24/16 12} h6 {(26. ... h6 27.Qc2 Bd8 28.Qb3 Be8 29.f4 Ba4 30.Qb2 Bc6
31.Bf1 Rb8 32.Ba5 Bd7 33.Be2 Ra8 34.Bc3 Kh7 35.Qb3 Rb8 36.Ba5 Ra8 37.Bc3)
+0.01/22 9} 27. Bf1 {(27.Bf1 Be8 28.h3 Bd8 29.Bb5 Bh5 30.g4 Bg6) +0.72/19
20} Be8 {(27. ... Be8 28.d3 Bg6 29.Qb5 Ra3 30.Qb2 Ra8) +0.01/23 8} 28. Be2
{(28.Be2 Bd8 29.Bb5 Bh5 30.Bf1 Be8 31.h3 Kh7) +0.57/21 19} Bc6 {(28. ...
Bc6 29.Qc2 Ba4 30.Qb2 Bc6) +0.01/21 11} 29. d3 {(29.d3 d4 30.exd4 exd3
31.Bxd3 Bxg2 32.Re1 Bc6) +0.64/17 13} exd3 {(29. ... exd3 30.Bxd3 Bd8 31.h4
Be8 32.Qc2 Ra3 33.Qb2 Ra8) -0.01/19 8} 30. Bxd3 {(30.Bxd3 Bd8 31.h3 Be7
32.Qd2 Bd8 33.Bc2 Bd7) +0.72/19 8} Bd8 {(30. ... Bd8 31.Qc2 Be8 32.h3 Bh5
33.Qb2 Be8 34.h4 Bc6 35.h5 Be8 36.Be2 Bd7 37.g4 Bc6 38.Qc2 Bxb6 39.Qg6 c4
40.Bd4 Ra6 41.Bf3) -0.26/22 20} 31. h3 {(31.h3) +0.72/17 8} Be8 {(31. ...
Be8 32.Qc2 Ra3 33.Qb2 Ra8) -0.01/23 15} 32. Qd2 {(32.Qd2 Bc6 33.Bc2 Ba4
34.Bxa4 Rxa4 35.Ba5 Ra3) +0.84/16 9} Ra3 {(32. ... Ra3 33.Qb2 Ra8) -0.01/23
15} 33. Ba5 {(33.Ba5 Be7 34.h4 Bc6 35.Bc3 Bd8 36.h5 Kg8) +1.00/17 4} Bf7
{(33. ... Bf7 34.Bf5 Be7 35.Bc3 Ra6 36.Ba5 g6 37.Bg4 Kg7 38.Bf3 Be6 39.Rd1
Qb8 40.Rc1 Qb7 41.Rd1) -0.01/21 9} 34. Bf5 {(34.Bf5 Kg8 35.Bc8 Qb8 36.Bg4
Kh8 37.Qe1 Qb7) +1.46/17 5} g6 {(34. ... g6 35.Bd3 Be7 36.Bc3 Kg7 37.Be2
Ra4 38.Qc2 Ra8 39.Qb2 Rb8 40.Bf3 Be6 41.Bg4 Bf7 42.Bd1 Ra8 43.Bf3 Be6
44.Rc1) -0.30/20 6} 35. Bc8 {(35.Bc8 Qb8 36.e4 g5 37.exd5 Bg6 38.Re1 Qxc8)
+1.89/18 7} Qb8 {(35. ... Qb8 36.Bg4 Kg7 37.Bc3 Qb7 38.Be2 Be7 39.Qb2 Ra8
40.h4 h5 41.f3 Bd8 42.Bd3 Rb8 43.Kf2 Qc6 44.b7 Bc7 45.Ke2 c4 46.Bc2 Bg3)
-0.15/21 5} 36. e4 {(36.e4 g5 37.exd5 Bg6 38.Re1 Kg7 39.Bg4 Bxb6) +2.48/18
6} Kg7 {(36. ... Kg7 37.exd5 Rxa5 38.Qxa5 Qxc8 39.Qa2 Qb7 40.Qa7 Qxd5 41.b7
Bc7 42.b8Q Bxb8 43.Qxb8 c4 44.Qc7 g5 45.Rb6 Qd1+ 46.Kh2 Qd5 47.Rb7 Kg6
48.Ra7 Kg7 49.Qe7 Kg6 50.Rd7 Qe6) -0.98/21 12} 37. exd5 {(37.exd5 g5 38.Bf5
Qe5 39.Bc2 Bxd5 40.Re1 Qd4) +3.09/17 4} g5 {(37. ... g5 38.Bf5 Qe5 39.Bc2
Qd6 40.b7 Bxa5 41.Qc1 Bb4 42.Qxa3 Bxa3 43.b8Q Qxb8 44.Rxb8 Bxd5 45.Rd8 Be6
46.Kf1 c4 47.Ke1 Bb2 48.Ke2 c3) -1.18/19 10} 38. Qe1 {(38.Qe1 Rxa5 39.Qxa5
Qxc8 40.Qb5 Qb7 41.Qc6 Bxd5) +2.69/18 6} Rxa5 {(38. ... Rxa5 39.Qxa5 Qxc8
40.Qb5 Qb7 41.Qc6 Bxd5 42.Qxb7+ Bxb7 43.Rd1 Bxb6 44.Rd7+ Kg6 45.Rxb7 Ba5
46.Rb5 Bb4 47.Kf1 Kf5 48.Ke2 Ke4 49.g4 Kd5 50.Kd3 Be1 51.Rb6 Bxf2 52.Rxf6)
-1.48/22 9} 39. Qxa5 {(39.Qxa5) +2.69/1} Qxc8 {(39. ... Qxc8 40.Qb5 Qf5
41.d6 Bc4 42.Qb2 Qd3 43.b7 Qxd6 44.b8Q Bc7 45.Qxc7+ Qxc7 46.Qb7 Qxb7
47.Rxb7+ Bf7 48.f3 Kg6 49.Kf2 h5 50.Rc7 c4 51.Ke3 Bg8 52.Rc5 Bf7 53.Kd4 g4
54.fxg4 hxg4 55.h4 f5 56.g3 Be6) -1.87/25 7} 40. Qb5 {(40.Qb5 Qf5 41.Rd1 c4
42.b7 Bc7 43.Qc6 Bb8) +2.65/19 4} Qf5 {(40. ... Qf5 41.d6 Bc4 42.Qb2 Qd5
43.b7 Qxd6 44.b8Q Bc7 45.Qxc7+ Qxc7 46.Qb7 Qxb7 47.Rxb7+ Bf7 48.f3 h5
49.Kf2 Kg6 50.Rc7 c4 51.Ke3 Bg8 52.Kd4 g4 53.fxg4 hxg4 54.h4 g3 55.Rc5 Be6
56.Rc6 Bf7) -1.84/27 7} 41. Rd1 {(41.Rd1 Qc2 42.Qb1 Qxb1 43.Rxb1 Bxd5
44.Rd1 Bxb6) +3.19/22 8} c4 {(41. ... c4 42.b7 Bc7 43.b8Q Bxb8 44.Qxb8 Bxd5
45.Qb5 Be6 46.Qxf5 Bxf5 47.Kf1 c3 48.Ke2 Kf7 49.Ra1 h5 50.Ra3 c2 51.Kd2 Ke6
52.Ra6+ Ke7 53.Kc1 Bd3 54.Ra3 Bf5 55.Ra5 Bd3 56.f3 h4) -1.31/22 6} 42. b7
{(42.b7 Qc2 43.b8Q Qxd1+ 44.Kh2 Qd4 45.Qd7 Ba5) +5.83/20 4} Bc7 {(42. ...
Bc7 43.b8R Bxb8 44.Qxb8 Bxd5 45.Qb5 Be6 46.Qxf5 Bxf5 47.Kf1 c3 48.Ke2 Kf7
49.Rc1 c2 50.Kd2 Ke6 51.Re1+ Kd6 52.Ra1 h5 53.f3 g4 54.fxg4 hxg4 55.h4 Be4
56.g3 Ke5 57.Ra5+ Ke6) -1.32/23 3} 43. Qc6 {(43.Qc6 Bb8 44.d6 Qe5 45.Qc8
Qe2 46.Qxb8 Qxd1+) +7.58/19 4} Bb8 {(43. ... Bb8 44.d6 Qe6 45.Qc8 Qe2
46.Qg4 Qe8 47.d7 Qd8 48.Qf3 Bc7 49.Qc6 h5 50.Re1 Kh6 51.g3 g4 52.hxg4 c3
53.Qxc3 hxg4 54.Re8 Bxe8) -3.14/17 6} 44. d6 {(44.d6 Qe5 45.Qc8 Qe2 46.Qxb8
Qxd1+ 47.Kh2 g4) +7.71/19 7} c3 {(44. ... c3 45.d7 c2 46.d8Q cxd1R+ 47.Qxd1
Qe5 48.g3 h5 49.Qg2 Bc7 50.Qd3 Bg6 51.Qe3 Qxe3 52.fxe3 g4 53.hxg4 Bb8
54.gxh5 Bxh5 55.Qc6 Bxg3) -4.55/20 6} 45. d7 {(45.d7 c2 46.d8Q cxd1Q+ h5+
Qf4+) +9.04/18 7} c2 {(45. ... c2 46.d8Q cxd1R+ 47.Qxd1 Qe5 48.g3 h5
49.Qcd7 h4 50.Qh5 Qe1+ 51.Kg2 Qe4+ 52.Qf3 Qxf3+ 53.Kxf3 hxg3 54.fxg3 Kg6
55.Qd8 Be5 56.b8Q Bxb8 57.Qxb8 Bd5+ 58.Kf2 Bf7 59.g4 Bd5 60.Qc8 Bf7 61.Qf5+
Kg7 62.Ke3 Bg6 63.Qd7+ Bf7 64.Qe7) -5.34/21 6} 46. d8=Q {(46.d8Q cxd1Q+ h5+
Qe5+ 48.Qe4 Qea1+) +9.76/18 6} cxd1=R+ {(46. ... cxd1R+ 47.Qxd1 Qe5 48.g3
h5 49.Qcd7 h4 50.Qh5 Qe1+ 51.Kg2 Qe4+ 52.Qf3 Qxf3+ 53.Kxf3 hxg3 54.fxg3 Kg6
55.Qd8 Be5 56.b8Q Bxb8 57.Qxb8 Bd5+ 58.Ke3 Kf7 59.Qc8 Kg6 60.Qd8 Bf7 61.Qd7
Bc4 62.Qe8+ Bf7 63.Qe4+ f5 64.Qc6+ Kg7 65.Qd6) -5.35/23 6} 47. Qxd1
{(47.Qxd1) +9.76/1} Qe5 {(47. ... Qe5 48.g3 h5 49.Qcd7 h4 50.Qh5 Qe1+
51.Kg2 Qe4+ 52.Qf3 Qxf3+ 53.Kxf3 hxg3 54.fxg3 Kg6 55.Qd8 Be5 56.b8Q Bxb8
57.Qxb8 Bd5+ 58.Ke3 Kf7 59.Qc7+ Kg6 60.Qd7 Bc4 61.Kf3 Bf7 62.Qd3+ f5)
-5.43/21 4 Black resigns} 1-0[/pgn]

[cdani]

Thanks a lot for a much improved version Dani

Thanks! We will see if this holds on rating lists...

It's a bit strange that Andscacs promoted to a rook...

Yes, it happens often when the position is lost. Someone gave a probable explanation here at Talkchess not long ago, if I remember well. Seems a typical search artifact of most engines.

[Hugo]

Thank you very much !

Kind regards, Clemens Keck

[Graham Banks]

Hello!
I published a new version, 0.92.

http://www.andscacs.com/downloads/andscacs092.zip

Thanks Daniel.

[schack]

I would be interested in what compiler flags you try in producing a Ryzen optimized executable.

[cdani]

I would be interested in what compiler flags you try in producing a Ryzen optimized executable.

I use Visual Studio. The idea is to try to compile Andscacs with the same flags but in the Ryzen computer, hoping that the pgo build obtain optimized results.