Most Positional/Strategic Chess Engines

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

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Albert Silver
Posts: 2831
Joined: Wed Mar 08, 2006 8:57 pm
Location: Rio de Janeiro, Brazil

Re: Most Positional/Strategic Chess Engines

Post by Albert Silver » Thu Dec 28, 2017 6:26 am

BrendanJNorman wrote:Hey guys,

Now that I am up to my neck in aggressive, tactical chess engines, I'm looking in the direction of more positional and strategic oriented engines.

I'd like your thoughts on which play the most human-like, seemingly strategic chess with a focus on static themes and very little king attacks or tactical play.

Any ideas?

Thanks!

EDIT: These can be ANY strength, just need to be playing in a POSITIONAL style.
Komodo.
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."

BrendanJNorman
Posts: 1400
Joined: Sun Feb 07, 2016 11:43 pm
Full name: Brendan J Norman

Re: Most Positional/Strategic Chess Engines

Post by BrendanJNorman » Thu Dec 28, 2017 7:07 am

Thanks guys,

I'm now testing all of the suggested guys on the black side of the Old Indian (1.d4 Nf6 2.c4 d6 3.Nc3 Nbd7 4.e4 e5) to see how they handle the position against Alfil 11 (human-like and strong, but not too strong to smother the opponent's playing style).

Results should be interesting.

User avatar
Ovyron
Posts: 2495
Joined: Tue Jul 03, 2007 2:30 am

Re: Most Positional/Strategic Chess Engines

Post by Ovyron » Thu Dec 28, 2017 9:00 am

matejst wrote:Zappa Mexico 2 XII with dr Wael Deeb settings. Very positional, very human-like, IMHO plays better than the vanilla version.
Can you give more details about this? I remember using some settings for Firebird with success, I didn't know he made ones for Zappa.

BrendanJNorman
Posts: 1400
Joined: Sun Feb 07, 2016 11:43 pm
Full name: Brendan J Norman

Re: Most Positional/Strategic Chess Engines

Post by BrendanJNorman » Thu Dec 28, 2017 10:41 am

Some interesting results...

Wasp 1.25's positional play didn't stand up to Alfil's dynamic kingside attacking.

[pgn][Event "Alfil vs Positional Guys in Old Indian"]
[Site "BRENDANNORMD8A2"]
[Date "2017.12.28"]
[Round "1"]
[White "Alfil 11"]
[Black "Wasp 1.25"]
[Result "1-0"]
[ECO "A55"]
[WhiteElo "2556"]
[BlackElo "2200"]
[PlyCount "64"]
[EventDate "2017.??.??"]
[TimeControl "60+1"]

1. d4 Nf6 2. c4 d6 3. Nc3 Nbd7 4. e4 e5 5. Nf3 {-0.03/12 2 (5.Nf3 exd4 6.Qxd4
Be7 7.Be Nb8 8.Qd3 0-0 9.0-0 Re8 10.Bd2 Nc6) 2} Be7 {-0.45/12 2 (5. ... Be7 6.
Be 0-0 7.0-0 c5 8.d5 Qa5 9.h3 Nb6 10.Be3 Bd7 11.Qc2 h6) 2} 6. Be2 {-0.01/13 3
(6.Be2 0-0 7.0-0 exd4 8.Qxd4 Nc5 9.Bd2 Ne6 10.Qd Bd7 11.Nd5 Nxd5 12.exd5) 3}
O-O {-0.54/13 2 (6. ... 0-0 7.0-0 Kh8 8.Be3 exd4 9.Nxd4 Ne5 10.f4 Neg4 11.Bd
c5 12.Nf5 Bxf5 13.exf5) 2} 7. dxe5 {0.01/13 3 (7.dxe5 Nxe5 8.0-0 Nxf+ 9.Bxf3
Re8 10.Qe2 Nd7 11.Nd5 Ne5 12.Bh5 Bf6 13.Nxf6+) 3} Nxe5 {-0.31/13 2 (7. ...
Nxe5 8.0-0 c6 9.Bf4 Nxf3+ 10.Bxf3 Be6 11.Qe Qa5 12.Rad1 Rad8 13.h3 Rfe8 14.Qd3
Nd7) 2} 8. O-O {-0.06/12 1 (8.0-0 Nxf3+ 9.Bxf3 Re8 0.Nd5 c6 11.Nxe7+ Qxe7 12.
Re1 d5 13.cxd5 cxd5) 1} Be6 {-0.19/13 2 (8. ... Be6 9.Nxe5 dxe5 10.Qb3 Qd6 11.
Be3 Rfd8 1.Nd5 Nxe4 13.Rfd1 Nc5 14.Bxc5 Qxc5 15.Nxe7+ Qxe7 16.Qxb7) 2} 9. Nxe5
{0.09/11 2 (9.Nxe5 dxe5 10.Qxd8 Rfxd8 11.Nd5 Nxd5 1.cxd5 Bd7 13.Be3 Bh4 14.
Rad1) 2} dxe5 {-0.26/13 2 (9. ... dxe5 10.Qb3 Qd6 11.Be3 Rfd8 1.Nd5 Nxe4 13.
Rad1 Nc5 14.Bxc5 Qxc5 15.Qxb7 Bd6 16.Rfe1) 2} 10. Qb3 {0.06/12 3 (10.Qb Qc8
11.Nd5 Bd8 12.Nxf6+ Bxf6 13.Rd1 Re8 14.Rd3 Qb8 15.Be3 c6) 3} Qd6 {-0.15/13 2
(10. ... Qd6 11.Be3 c6 1.f4 exf4 13.Rad1 Qc7 14.Bxf4 Bc5+ 15.Kh1 Qb6 16.e5
Qxb3 17.axb3 Ng4) 2} 11. Nd5 {0.21/11 2 (11.Nd5 Nxd5 1.cxd5 Bc8 13.Be3 f5 14.
f3 Bh4 15.Rad1 fxe4 16.fxe4) 2} Nxe4 {-0.04/14 2 (11. ... Nxe4 1.Nxe7+ Qxe7
13.Qxb7 Bf5 14.Qc6 Rab8 15.Qa4 c6 16.Rd1 Rfc8 17.Qa5 Qc5 18.Qe1 Rd8 19.Be3) 2}
12. Nxe7+ {0.45/11 2 (1.Nxe7+ Qxe7 13.Qxb7 Bf5 14.Re1 Qd6 15.Rd1 Qb6 16.Qxb6
axb6 17.Be3) 2} Qxe7 {-0.19/14 2 (1. ... Qxe7 13.Qxb7 Bf5 14.Qc6 Rab8 15.Qa6
c5 16.Qa3 Rfd8 17.Be3 Qb7 18.Rfd1 Be6 19.Rxd8+ Rxd8 20.Re1) 2} 13. Qxb7 {
0.45/12 2 (13.Qxb7 Bf5 14.Re1 Rfb8 15.Qc6 Qc5 16.Qxc5 Nxc5 17.b3 e4 18.Be3 Nd3)
+0.45/1 2} Bf5 {-0.09/14 2 (13. ... Bf5 14.Qc6 Rab8 15.Qa6 c5 16.Qa3 Ng5 17.
Bxg5 Qxg5 18.Rfd1 Rfc8 19.Rac1 Rc6 0.Bf3 e4) 2} 14. Rd1 {0.42/11 4 (1.Rd1 Nf6
15.Qc6 Bc2 16.Rd2 Be4 17.Qa6 Qc5 18.Rd1 Rab8 19.Be3) 4} Nf6 {-0.36/11 2 (14. .
.. Nf6 15.b3 Rfb8 16.Qa6 Ne4 17.Bf3 Rb6 18.Qa5 c5 19.Be3 Rc8 0.Be2) 2} 15. Bg5
{0.48/11 2 (15.Bg5 Rfb8 16.Qc6 h6 17.Bc1 Bc 18.Re1 Be4 19.Qa6 Rb6 20.Qa5) 2}
Rab8 {-0.26/14 2 (15. ... Rab8 16.Qxa7 Rxb 17.Rd2 Rxd2 18.Bxd2 Rd8 19.Ba5 Ne8
20.Rd1 Rxd1+ 21.Bxd1 Qd7 22.Be2 Qa4 23.Qc5) 2} 16. Qxa7 {0.34/11 1 (6.Qxa7
Rxb2 17.Re1 Qd6 18.Be3 Qc6 19.Qc5 Qxc5 20.Bxc5 Rd8 21.a4) 1} Rxb2 {-0.04/14 2
(16. ... Rxb 17.Rd2 Qb4 18.Rxb2 Qxb2 19.Re1 Qb4 20.Kf1 Ng4 21.Bxg4 Bxg4 22.
Qxc7 Rc8 23.Qxe5 Qxc4+ 24.Kg1 Qxa2) 2} 17. Bf3 {0.50/11 3 (17.Bf e4 18.Bc1
Rb4 19.Ba3 exf3 20.Qa5 Rfb8 21.Bxb4 Rxb4 22.Qxf5) 3} Qe6 {-0.20/12 2 (17. ...
Qe6 18.Re1 c6 19.Bc1 Rbb8 0.Qc5 e4 21.Bf4 Rb2 22.Re2 Rxe2 23.Bxe2 Rd8) 2} 18.
Qxc7 {0.46/10 3 (18.Qxc7 e4 19.Rd6 Qe5 20.Bxf6 gxf6 21.Bh5 Rc2 22.Re1 Rxa2)} e4
{-0.05/13 2 (18. ... e4 19.Re1 Rc8 0.Qa5 h6 21.Bc1 Rbb8 22.Be2 Bg4 23.Bxg4
Nxg4 24.Bf4 Rb2 25.f3) 2} 19. Re1 {0.60/10 1 (9.Re1 Rc8 20.Qa5 Qxc4 21.Rac1
Qd4 22.Be3 Qb4 23.Rxc8+ Bxc8) 1} Rc8 {-0.28/12 2 (19. ... Rc8 0.Qa5 h6 21.Qc3
Qxc4 22.Bxf6 Qxc3 23.Bxc3 Rxc3 24.Bxe4 Bxe4 25.Rxe4 Rcc2) 2} 20. Qa5 {0.68/10
1 (20.Qa5 Rxc4 2.Rad1 Rc8 22.a4 Ne8 23.Rd4 Rcc2 24.Bxe4 Bxe4) 1} h6 {-0.41/12
2 (0. ... h6 21.Bc1 Rbb8 22.Be2 Bg4 23.f3 exf3 24.gxf3 Ra8 25.Qc3 Bf5 26.Bd3
Qd7) 2} 21. Bc1 {0.92/9 1 (2.Bc1 Qb6 22.Qxb6 Rxb6 23.Be2 Be6 24.Be3 Ra6 25.c5)
1} Rbb8 {-0.51/13 2 (1. ... Rbb8 22.Be2 Bg4 23.Bf1 Qc6 24.Qc3 Bf5 25.Bd2 Qd6
26.Be2 Rd8 27.Rad1 Ng4 28.g3) 2} 22. Be2 {1.31/11 2 (2.Be2 Bg4 23.Bf1 Ra8 24.
Qd2 Rc5 25.Rb1 Qc6 26.a3 Ne8 27.Qd4) 2} Bg4 {-0.61/12 2 (2. ... Bg4 23.Bf1
Qc6 24.Qc3 Bd7 25.a4 Bf5 26.Bd2 Qd6 27.a5 Ng4 28.g3 Qc5) 2} 23. Bf1 {1.26/11
2 (3.Bf1 Ra8 24.Qd2 Ra4 25.Bb2 Rca8 26.Bc3 Bf5 27.Bd4 Qc6 28.c5) 2} Qc6 {
-0.76/12 2 (3. ... Qc6 24.Qc3 Ra8 25.Bb2 Bf5 26.Qe5 Rcb8 27.a4 Be6 28.c5 Bd5
29.Bd4 Be6) 2} 24. Qe5 {1.56/10 2 (4.Qe5 Ra8 25.a4 Ra6 26.Qb5 Be6 27.c5 Raa8
28.Qxc6 Rxc6) 2} Re8 {-0.74/11 2 (4. ... Re8 25.Qc3 Bf5 26.Ba3 Qc7 27.Rad1
Red8 28.Bb2 Qa7 29.Qe3 Qe7 30.Rxd8+ Rxd8) 2} 25. Qg3 {1.57/10 1 (25.Qg3 Ra8
26.Bb2 Bf5 27.a4 Ra5 28.Red Nh5 29.Qh4 Nf6) 1} Kh7 {-0.91/12 2 (5. ... Kh7 26.
Be3 Red8 27.c5 Be6 28.Reb1 Rxb1 29.Rxb1 Qa8 30.a3 Ba2 31.Rb6 Qxa3) 2} 26. a4 {
1.65/9 2 (6.a4 Be6 27.Be2 Bxc4 28.Bxh6 gxh6 29.Rec1 Ng4 30.Bxc4) 2} Be6 {
-1.17/11 2 (6. ... Be6 27.Qh4 Bc8 28.a5 Ba6 29.Bg5 Kg6 30.Bxf6 Qxf6 31.Qg4+
Kh7 32.Rab1 Bb7) 2} 27. Qh4 {1.65/9 2 (7.Qh4 Bxc4 28.Bxh6 gxh6 29.Bxc4 Qxc4
30.Qxf6 Rb4 31.a5) 2} Bc8 {-1.28/12 1 (27. ... Bc8 28.a5 Ba6 29.Bf4 Rbd8 30.
Rec e3 31.fxe3 Rd2 32.Qe1 Red8 33.e4 Rb2) 1} 28. Ra3 {1.93/10 1 (28.Ra3 Kg8
29.Rg3 h5 30.a5 Bg4 3.Ba3 Qa4 32.Qg5) 1} Bb7 {-1.17/10 2 (8. ... Bb7 29.a5
Ba6 30.Rh3 Bc8 31.Bxh6 Bxh3 32.Bf4+ Kg8 33.Bxb8 Bxg2 34.Bxg2 Rxb8 35.Bxe4 Qxc4)
2} 29. Rg3 {2.50/9 2 (9.Rg3 e3 30.Bd3+ Kh8 31.Bxe3 Ne4 32.Rxg7 Kxg7 33.Bxe4) 2
} Ra8 {-1.02/10 1 (29. ... Ra8 30.Be3 Rxa4 3.Rb1 Nd7 32.Qh5 Re7 33.Qf5+ Kg8 34.
Bxh6 Qxh6 35.Rxb7) 1} 30. Rxg7+ {6.60/12 1 (30.Rxg7+ Kxg7 3.Qxh6+ Kg8 32.Bb2
e3 33.Qg5+ Kf8 34.Ba3+ Re7 35.Rxe3 Ne4) 1} Kxg7 {-5.72/13 1 (30. ... Kxg7 3.
Qxh6+ Kg8 32.Bb2 e3 33.Qg5+ Kf8 34.Ba3+ Re7 35.Rxe3 Ne4 36.Qxe7+ Kg8 37.Bd3
Re8 38.Bxe4 Rxe7 39.Rg3+ Kf8 40.Bxc6 Bxc6 41.Bxe7+ Kxe7) 1} 31. Qxh6+ {
6.84/12 1 (3.Qxh6+ Kg8 32.Bb2 e3 33.Qg5+ Kf8 34.Ba3+ Re7 35.Rxe3 Ne4 36.Qxe7+
Kg8) 1} Kg8 {-8.12/15 4 (31. ... Kg8 32.Bb2 e3 33.Bxf6 exf2+ 3.Kxf2 Qxf6+ 35.
Qxf6 Rxe1 36.Kxe1 Rxa4 37.Qg5+ Kf8 38.Qd8+ Kg7 39.Qd7 Rb4 40.Qd4+ Kg8 41.Qd8+
Kg7 42.Qg5+ Kh8) 4} 32. Bb2 {6.84/12 1 (32.Bb2 e3 33.Qg5+ Kf8 34.Ba3+ Re7 35.
Rxe3 Ne4 36.Qxe7+ Kg8 37.Bd3 Re8) +6.84/2 1} e3 {adjudication} 1-0[/pgn]

While Romi made Alfil look like an amateur with his mature positional play!

Look at that good knight vs bad bishop scenario - very human-like, the way Romi seems to be angling for this outcome.

[pgn][Event "Alfil vs Positional Guys in Old Indian"]
[Site "BRENDANNORMD8A2"]
[Date "2017.12.28"]
[Round "1"]
[White "Alfil 11"]
[Black "RomiChess P3K"]
[Result "0-1"]
[ECO "A55"]
[WhiteElo "2556"]
[BlackElo "2406"]
[PlyCount "154"]
[EventDate "2017.??.??"]
[TimeControl "60+1"]

1. d4 Nf6 2. c4 d6 3. Nc3 Nbd7 4. e4 e5 5. Nf3 {-0.03/12 2 (5.Nf3 exd4 6.Qxd4
Be7 7.Be Nb8 8.Qd3 0-0 9.0-0 Re8 10.Bd2 Nc6) 2} Be7 {-0.14/13 4 (6.Re2 exd 7.
Nxd4 Nc5 8.f3 Bd7 9.Bf4 0-0 10.0-0 c6) 4} 6. Be2 {-0.01/13 3 (6.Be2 0-0 7.0-0
exd4 8.Qxd4 Nc5 9.Bd2 Ne6 10.Qd Bd7 11.Nd5 Nxd5 12.exd5) 3} b6 {-0.16/12 2 (6.
... b6 7.Qc Bb7 8.0-0 0-0 9.Rd1 Qe8 10.Bg5 c6 11.Qd3 Rd8 12.dxe5) 2} 7. O-O {
0.18/11 2 (7.0-0 0-0 8.Qc exd4 9.Nxd4 Bb7 10.Nf5 Nc5 11.Nxe7+ Qxe7 12.Nd5) 2}
Bb7 {-0.04/14 4 (8.Rc2 0-0) -0.0/14 4} 8. dxe5 {0.23/11 2 (8.dxe5 Nxe5 9.Nxe5
dxe5 10.Qa4+ Qd7 11.Qxd7+ Kxd7 1.Rd1+ Bd6 13.Bf3) 2} dxe5 {-0.20/14 3 (9.Ra4
c6)} 9. Qc2 {0.32/11 1 (9.Qc2 0-0 0.Rd1 Qe8 11.Nd5 Rc8 12.Bg5 Nxd5 13.cxd5
Bxg5 14.Nxg5) 1} Bd6 {-0.38/12 2 (9. ... Bd6 10.Bg5 h6 11.Bxf6 Qxf6 1.Nd5
Bxd5 13.cxd5 0-0 14.Bc4 Nc5) 2} 10. Bg5 {0.28/11 3 (10.Bg5 0-0 11.Rfd1 Nc5 12.
Nxe5 Ncxe4 1.Bf3 Nxc3 14.Qxc3 Bxf3 15.Qxf3) 3} h6 {-0.11/14 3 (11.Rxf6 Qxf6
12.Nd5 Qd8 1.Rd1 0-0) 3} 11. Bh4 {0.21/12 4 (11.Bh 0-0 12.Rad1 Qe8 13.Nd2 c6
14.Nb3 Qe6 15.Qd2 Bb4 16.Bd3 Rad8) 4} g5 {-0.19/14 3 (12.Rg 0-0 Re8 Kh7 14.h3
14.g4 Nxg4) 3} 12. Bg3 {0.46/12 2 (1.Bg3 Qe7 13.Nd5 Qe6 14.c5 bxc5 15.Nxf6+
Nxf6 16.Bb5+ c6 17.Ba4) 2} O-O {-0.28/13 3 (1.Rad1 Qe7 14.Nd5 Bxd5 15.exd5
Nc5 16.Qf5 16.Bd3 Rd8 17.b4 Nxd3) 3} 13. Rfe1 {0.50/11 7 (13.Rfe1 Qe8 14.Nd5
a5 15.h3 Qe6 16.Rad1 c6 1.Nxf6+ Nxf6 18.Rxd6) 7} Qe8 {-0.25/13 3 (14.Rd5 Nc5
15.Rxf6 15.Nd2 Ne6 16.Bh5 Raxd8 17.Nf a5 18.Bg4 Nc5 19.Bf5) 3} 14. h4 {
0.48/11 4 (1.h4 g4 15.Nh2 Qe6 16.Qc1 Kh7 17.Nb5 Bxe4 18.Bxg4 Nxg4 19.Rxe4) 4}
g4 {-0.37/13 2 (14. ... g4 15.Nd Bb4 16.Nb5 Qd8 17.Rad1 a5 18.Nc3 Re8 19.Qc1
c6 20.a4) 2} 15. Nh2 {0.45/11 1 (5.Nh2 Qe6 16.Qd1 Nxe4 17.Nxe4 Bxe4 18.Bxg4
Bf5 19.Qf3 Bxg4 20.Nxg4) 1} h5 {-0.07/12 2 (15. ... h5 16.Qd Kh7 17.Qg5 Qe6
18.Nf1 Nc5 19.Nd2 Rg8 20.Qe3 a5 21.Rad1 Rad8 22.Kh2) 2} 16. Qd2 {0.54/11 2
(16.Qd Bb4 17.Qg5+ Kh8 18.Bd3 Bxc3 19.bxc3 Kh7 20.Bf4 exf4) 2} Kh7 {-0.14/12
3 (17.Raf1 Bb4 18.f Rxg8 Nc5) 3} 17. Rad1 {0.42/10 3 (17.Rad1 Bb4 18.Bd Rd8
19.Re2 Rg8 20.a3 Ba5 21.Bc2 Bxc3) 3} Rd8 {0.02/12 2 (17. ... Rd8 18.f3 Rg8 19.
Nf1 Nc5 0.Kh2 gxf3 21.gxf3 Be7 22.Qe3) 2} 18. Nd5 {0.42/9 2 (18.Nd5 Nxe4 19.
Bd3 f5 0.Rxe4 Bxd5 21.cxd5 fxe4 22.Bxe4+) 2} Nxd5 {-0.08/13 2 (18. ... Qe7 19.
a3 a5 0.Nf1 Nc5 21.Nfe3 Bc8 22.Bd3 Bd7) 2} 19. cxd5 {0.51/11 2 (19.cxd5 Qe7 0.
Qc2 Kg8 21.Qc1 Kg7 22.Nf1 f5 23.exf5 Rxf5 24.Bd3) 2} Qe7 {-0.34/12 2 (19. ...
Qe7 0.Qc2 a5 21.Bb5 f6 22.Nf1 Bb4 23.Nd2 Nc5 24.Re3 Qd6 25.Nc4) 2} 20. Qc2 {
0.57/11 2 (0.Qc2 Bc5 21.b4 Bd6 22.a3 Nf6 23.Bd3 a5 24.bxa5 bxa5 25.Rc1) 2} Rc8
{-0.42/12 1 (20. ... Rc8 2.Nf1 c6 22.dxc6 Rxc6 23.Qd3 Nc5 24.Qd5 Qe6 25.Ne3
Qxd5) 1} 21. Bb5 {0.48/11 3 (21.Bb5 c6 22.dxc6 Bxc6 2.Bxc6 Nb8 24.Qd3 Rxc6 25.
Rc1 Rxc1 26.Rxc1) 3} Nb8 {-0.29/13 2 (1. ... c6 22.Bc4 cxd5 23.exd5+ f5 Qf7
24.Qd3 25.Qc3 Kg8 26.a4 a6 27.Rc1) 2} 22. Qd3 {0.56/10 1 (22.Qd3 Rfd8 23.Qe3
c6 24.Bd3 Nd7 25.Rc Nc5 26.Bc2 cxd5) 1} Bb4 {0.13/13 2 (2. ... c6 23.a3 23.b4
24.dxc6 Nxc6 Rfd8 25.Qc4 Rc7) 2} 23. Re2 {0.51/12 2 (3.Re2 c6 24.Bc4 Rfd8 25.
Nf1 b5 26.Bb3 Na6 27.Bc2 Kg8 28.Ne3 cxd5) 2} Rfd8 {-0.19/13 2 (3. ... Bc5 24.
Nf1 c6 25.Bc4 25.b4 Bd4 26.Ne3 Bxe3 27.Qxe3 Nd7 28.Bb3 cxd5) 2} 24. Nf1 {
0.62/10 1 (24.Nf c6 25.Ne3 cxb5 26.a3 Nc6 27.Nf5 Qc5 28.axb4 Nxb4) 1} c6 {
-0.42/13 2 (4. ... Bc5 25.Bc4 Rd6 26.b4 Bd4 27.Ne3 Bxe3 28.Qxe3 Nd7 29.Bd3 Rd8
30.dxc6 Bxc6 31.Rc2) 2} 25. Bc4 {0.57/12 1 (25.Bc4 b5 26.Bb3 Na6 27.Qe3 Bd6
28.Nd2 Ba8 29.Rde Nb4 30.dxc6 Bxc6) 1} b5 {-0.27/13 1 (25. ... b5 26.Bb3 Na6
27.Ne3 Nc5 28.Qc2 Qf6 29.Nf5 cxd5 30.Bxd5 Bxd5 3.Rxd5 Ne6 32.Bxe5 Rxc2) 1} 26.
Bb3 {0.53/12 1 (26.Bb3 Na6 27.Qe3 Bd6 28.Nd2 Nb4 29.Qxa7 cxd5 30.Qb6 Nd3 3.
exd5 Nxb2) 1} Na6 {-0.30/13 2 (6. ... Bc5 27.Qc3 f6 28.Ne3 Bxe3 29.Qxe3 29.d6
Rxd6 30.Rxd6 Qxd6 31.Qxa7) 2} 27. Qe3 {0.53/11 1 (27.Qe3 Bd6 28.Nd2 Ba8 29.Nb
cxd5 30.exd5 Nc5 31.Bxe5 Nxb3) 1} f6 {-0.08/13 2 (7. ... cxd5 Bxc8 28.Rxd5
Rxd5 29.Bxd5 Nc7 30.Bc6 a5 31.a3 Bc5) 2} 28. a3 {0.43/11 2 (8.a3 Bc5 29.Qd2
Qg7 30.Bh2 Qh6 31.Ne3 Bd4 32.Nf5 Qxd2 33.Rexd2) 2} Bc5 {-0.27/12 1 (28. ...
Bc5 29.Qd2 Nc7 30.Ne3 cxd5 3.exd5 Bxe3 32.Qxe3 Bxd5 33.Bc2+ Kg8 34.Qxa7 Qd7) 1
} 29. Qd2 {0.50/11 1 (29.Qd2 b4 30.Ne3 bxa3 3.Nf5 Qf8 32.bxa3 Bxa3 33.d6 Nc5
34.Bc2) 1} Nc7 {-0.31/12 1 (29. ... Qf8 30.Qc2 a5 3.dxc6 Rxd1 32.Qxd1 Bxc6 b4
33.Qd3 bxa3 34.bxa3) 1} 30. Ne3 {0.65/11 1 (30.Ne3 Qf8 3.Qc2 Ne8 32.dxc6
Rxd1+ 33.Qxd1 Bxc6 34.Qc1 Bxe3) 1} Ne8 {-0.14/12 1 (30. ... Ne8 3.Nf5 Qf8 32.
Qa5 Nd6 33.dxc6 Nxf5 34.Rxd8 Rxd8 35.Qc7+ Qg7 36.Qxd8 Nxg3 37.Rc2) 1} 31. Rc1
{0.64/9 1 (3.Rc1 Ng7 32.Qd3 Qd6 33.Rec2 Bd4 34.Kf1 Kh8 35.Qxd4) 1} Bxe3 {
0.01/13 2 (31. ... cxd5 3.Rxc8 Bxc8 33.Bxd5 Be6 34.Qc2 Bxd5 35.exd5+ Kg8 36.
Qd3 Qd6 37.Qg6+ Ng7 38.Nc3) 2} 32. Rxe3 {0.56/11 1 (32.Rxe3 Nd6 33.Bh2 Kg7 34.
Re2 Ba8 35.Rc2 Qf7 36.Qd3 cxd5 37.exd5) +0.56/1 1} cxd5 {0.00/13 1 (32. ...
cxd5 33.Rxc8 Bxc8 34.exd5 Nd6 35.Rc3 Bf5 36.Rc6 Be4 37.Qc3 Rd7 38.Kh2 Rd8) 0.
00/3 1} 33. Rxc8 {0.59/11 1 (33.Rxc8 Bxc8 34.exd5 Nd6 35.Re Bf5 36.Qb4 Kg7 37.
Bh2 Kg8 38.Re2) 1} Rxc8 {0.03/14 1 (33. ... Nd6 34.Rec3 Kg7) +0.03/4 1} 34.
exd5 {0.56/11 1 (34.exd5 Nd6 35.Re2 Qc7 36.Re Qd7 37.Bc2+ Kg8 38.Qd3 Qf7 39.
Qg6+) 1} Nd6 {0.15/13 1 (34. ... Kg7 Qd8 Qxc8) +0.5/13 1} 35. Re1 {0.50/10 1
(35.Re Qd7 36.Bc2+ Kg7 37.Bb1 Rc5 38.Qd3 e4 39.Qe3 Rxd5) 1} Qc7 {0.39/12 1
(35. ... a5 36.Ba2 Kg7 37.Bb 37.Qxa5 Bxd5 38.a4 Rc5) 1} 36. Ba2 {0.76/10 1
(36.Ba2 Qc5 37.Bb+ Kg7 38.Rd1 a6 39.Qd3 e4 40.Qb3 f5) 1} Qg7 {0.31/12 1 (36. .
.. Qg7 37.Bb+ Kg8 38.b3 Rc5 39.Qe3 Qc7 40.Bg6 Bxd5 41.Bxh5 Nf5 42.Qd3 Nxg3) 1}
37. Bb1+ {0.73/11 1 (37.Bb+ Kg8 38.Qb4 Qd7 39.Bg6 Nc4 40.Rd1 a5 41.Qb3 Bxd5) 1
} Kg8 {0.17/13 1 (37. ... Kg8 38.Qb4 Qd7 39.Qd2 Kg7 40.Qd3 Qf7 4.b4 Kf8 42.Rd1
Ke8 43.Bc2 Kd8 44.Qb3) 1} 38. Rd1 {0.64/10 1 (38.Rd a6 39.Qb4 Qd7 40.Bg6 Nf5
41.Bxh5 Nxg3 42.fxg3 Bxd5) 1} f5 {0.57/13 1 (38. ... f5 39.Qg5 f4 40.Bh2 Qxg5
4.hxg5 h4 42.Kh1 g3 43.Bg1 a5 44.g6 e4 45.Rd2) 1} 39. Qb4 {0.56/11 1 (39.Qb4
Qf6 40.Bh2 Bxd5 4.Qe1 Be4 42.Ba2+ Kg7 43.Qd2 Nf7 44.Bxf7) 1} Qf6 {0.37/14 1
(39. ... gxf3 40.gxf3 Kh7 4.b3 Kh8 42.Qe1 Rg8 43.Kh2 Re8 44.Qc3) 1} 40. Ba2 {
0.42/11 1 (40.Ba2 a6 4.Re1 f4 42.Bh2 e4 43.Bb3 Kg7 44.Rd1 Kg8 45.Ba2) 1} f4 {
0.81/12 1 (40. ... f4 4.Bh2 a6 42.Re1 Rc7 43.Bb3 Kh8 44.Ba2 Rc8 45.Bb3) 1} 41.
Bh2 {0.42/9 0 (41.Bh2 Rc2 42.Bb1 Rc8 43.b3 Rd8 44.g3 fxg3 45.fxg3) +.42/9 0} a6
{0.57/12 1 (4. ... Qg6 42.Qe1 Qf6 43.Qb4 Qg6 44.Qe1 Qf6 45.Qb4) 1} 42. Bb3 {
0.34/10 2 (4.Bb3 e4 43.Re1 g3 44.fxg3 a5 45.Qxa5 Qd4+ 46.Kh1 Qxb2) 2} Ra8 {
0.56/12 1 (42. ... Ra8 43.a4 a5 44.Qc3 b4 45.Qc7 f3 46.Bc2 Rc8 47.Qh7+ Kf8 48.
Bg6) +0.56/2 1} 43. a4 {0.57/10 1 (43.a4 Rd8 44.axb5 Nxb5 45.Ba2 Kg7 46.Qe Re8
47.g3 fxg3) 1} Rc8 {0.60/12 1 (43. ... Rc8 44.Qe Re8 45.axb5 axb5 46.Qb4 f3
47.Rc1 fxg2 48.Kxg2 Rd8) 1} 44. axb5 {0.50/11 1 (44.axb5 axb5 45.Rf Kg7 46.
Re1 e4 47.Rf1 Kg6 48.Qd2 g3 49.fxg3) 1} axb5 {0.33/12 1 (44. ... axb5 45.Ra
Rc7 46.Re1 Rc8 47.Qd2 Kh7 48.Re2 Kg8 49.Qb4 e4) 1} 45. g3 {0.59/11 2 (45.g3
Rf8 46.gxf4 exf4 47.Rd4 Nf5 48.Rxf4 Qxb 49.d6+ Kh8 50.Bd1) 2} f3 {0.52/13 1
(45. ... Kh7 46.Re e4 47.Ra1 Qh6 Qg7 Qh6 Qg7 Qh6) 1} 46. Qd2 {0.54/12 2 (46.
Qd Kg7 47.Ra1 e4 48.Kf1 Qh6 49.Qd1 Kf7 50.Qe1 Kg6 51.Bg1 e3) 2} Kh7 {0.49/13
1 (46. ... Qh6) +0.49/3 1} 47. Kf1 {0.53/11 2 (47.Kf1 Kg7 48.Qd3 e4 49.Qe3 Qe5
50.Qd e3 51.Qxe3 Qxe3 52.fxe3) 2} Qh6 {0.75/12 1 (47. ... Qh6 48.Qe e4 49.Ra1
Rc5 50.Rd1 Rc7 51.Kg1 Rc8) 1} 48. Qb4 {0.54/12 1 (48.Qb4 Qf6 49.Kg e4 50.Qd2
Kg7 51.Kf1 Rc7 52.Qb4 Rc8 53.Qa3) 1} Rc1 {0.77/13 1 (48. ... Rc 49.Qe1 Rxd1
50.Qxd1 Kg8 51.Bc2 b4) 1} 49. Qe1 {0.54/13 0 (49.Qe1 Rxd1 5.Qxd1 e4 51.Qc2
Qf6 52.Kg1 Kg6 53.Qc7 Qe5 54.Qc3 Kf5 55.Qd2) 0} Rxd1 {0.77/15 1 (49. ... Kg8
50.Kg Bc8) 1} 50. Qxd1 {0.50/13 1 (50.Qxd Kg7 51.Kg1 Kf6 52.Ba2 e4 53.Qc2 Kf5
54.Qc3 b4 55.Qc2 Ke5) 1} Kg8 {0.81/13 1 (50. ... Kg8 5.Kg1 Bc8 52.Qe1 Bf5 53.
Qc3 Qf6 54.Qe3 Kh7 55.Qb6 Bd3) 1} 51. Ba2 {0.48/12 1 (5.Ba2 e4 52.Bg1 b4 53.
Ke1 Kf7 54.Bb3 Qg7 55.Qd2 Qf6 56.Qc1 Ke8) 1} Qg6 {1.04/13 1 (5. ... Kf7) 1}
52. Kg1 {0.32/11 0 (52.Kg1 Kf7 53.Kh1 Ne4 54.Qe1 Nxf2+ 55.Qxf2 Qa6 56.b4 e4 57.
Kg1) +.32/11 0} Qe4 {0.86/13 1 (53.Bb3 53.Qc2 54.Qd2 Kf7 55.Qg5) +0.86/3 1} 53.
Kh1 {0.00/11 1 (53.Kh Qe2 54.Qc1 Ne4 55.d6+ Kh7 56.Qc7+ Kh6 57.Qc1+ Kh7) 1}
Kf7 {0.85/13 1 (53. ... Bxd5 54.Bxd5+ Qxd5 Ne4 h7 h6 f8 Qd6 fxd6 58.Kg) 1} 54.
Bg1 {0.26/11 1 (54.Bg Qa4 55.Bb3 Qb4 56.Qc2 e4 57.Ba2 Qa5 58.Qb1 Ke7 59.Kh2) 1
} Qf5 {0.69/12 0 (54. ... Qf5 55.Qc1 Qf6 56.Bb3 Kg6) +.69/12 0} 55. Qc1 {
0.42/12 0 (55.Qc1 Qf6 56.Qc7+ Qe7 57.Qc3 Qf6 58.Bb3 e4 59.Qc1 Qe5 6.Qg5 Qxg5)
0} Qf6 {0.64/13 1 (55. ... Bc8) +0.64/3 1} 56. Qc7+ {0.43/12 0 (56.Qc7+ Qe7 57.
Qc3 Qf6 58.Kh2 e4 59.Qc7+ Kg8 6.Qb6 Qe5 61.Qc5 e3) 0} Qe7 {0.75/14 1 (56. ...
e4 57.Bb3+ Ke8 58.Qc3) +0.75/4 1} 57. Qc3 {0.34/12 1 (57.Qc3 Qf6 58.Kh2 e4 59.
Qc7+ Ke8 60.Qc Qe5 61.Qg5 Qxg5 62.hxg5 e3) 1} Kg6 {0.77/12 1 (57. ... Kg6 58.
Qc2+ e4 59.Kh2 Bc8 60.Bb3 Bf5 6.Qc3 Qf6 62.Bc2 Qxc3 63.bxc3 Kf6 64.Bb3 Ke5 65.
Kh1) 1} 58. b4 {0.45/10 1 (58.b4 Qf6 59.Qc7 e4 60.Qb8 Bc8 6.Bb1 Qe7 62.Kh2
Kf5) 1} Bc8 {0.91/13 1 (58. ... Kf5 59.Qd2 Qf7 60.Qh6 Qg6 6.Qe3 Kf6) 1} 59.
Kh2 {0.45/12 1 (59.Kh2 Bf5 60.Qe3 Kg7 6.Qb6 Bd3 62.Qc5 Kg6 63.Qe3 Bc4 64.Bb1+
e4) 1} Bf5 {0.82/13 1 (59. ... Bf5 60.Bb3 Qg7 6.Qc6 Qd7 62.Qc3) 1} 60. Bb3 {
0.48/13 1 (60.Bb3 Kf6 6.Qe3 Kg7 62.Qb6 Bd3 63.Qc5 Bf5 64.Ba2 Kf6 65.Qc1 Qg7) 1
} Kg7 {0.81/14 1 (60. ... Qf8) +0.8/14 1} 61. Qc6 {0.43/12 1 (6.Qc6 Kg6 62.Qc5
e4 63.Qe3 Bc8 64.Kh1 Bb7 65.Qd4 e3 66.Bc2+ Kf7) 1} Bb1 {0.81/12 1 (6. ... Bb1
62.Qc5 Bg6 63.Qc6 Bf5 64.Qb6 Bg6 65.Qc5 Kh7 66.Qc6) 1} 62. Bd1 {0.48/11 0 (62.
Bd1 Bf5 63.Qc5 Kf6 64.Bb3 Kg7 65.Qe3 Ne4 66.Qe1 Qd6 67.Qa1) +.48/11 0} Qf6 {
0.87/12 1 (62. ... Qf6 63.Bb3 Bd3 64.Qc7+ Kg6 65.Qc Nf5 66.Qd2 e4 67.Bc2 Bxc2
68.Qxc2 Qe5) 1} 63. Qc1 {0.12/12 1 (63.Qc Bd3 64.Qc7+ Kg6 65.Qb8 e4 66.Qg8+
Kh6 67.Qe6 Kg6 68.Bb3 Bc4) 1} Bd3 {0.84/13 1 (63. ... Kg6 Nf5 Kf7 Qe7) +0.84/
3 1} 64. Qc7+ {-0.06/13 1 (64.Qc7+ Kg6 65.Qb8 Bc4 66.Bc2+ e4 67.Bxe4+ Nxe4 68.
Qe8+ Kg7 69.Qxe4 Qd6 70.Qd4+) -0.06/3 1} Kg6 {0.86/13 1 (64. ... Kg6 65.Qb8
Kh6 66.Qc7 e4 67.Qb6 Qe7 68.Qe3+ Kg6) +0.86/3 1} 65. Bb3 {0.00/13 0 (65.Bb3 e4
66.Qc5 Kf5 67.Qc1 Bc4 68.Bxc4 bxc4 69.b5 Ke5 7.Qc3+ Kf5 71.Qxf6+) 0} Bc4 {
1.02/15 1 (65. ... Nc4 66.Kh Nd6 67.Qa7 e4 68.Qa8 Qe5 69.Qg8+ Qg7 70.Qe6+) 1}
66. Bxc4 {0.00/14 1 (66.Bxc4 bxc4 67.b5 e4 68.Qc5 Kf5 69.Qe3 Ke5 70.Qc3+ Kf5 7.
Qe3) 1} Nxc4 {1.10/16 1 (66. ... e4 67.Qc8 67.Bxd6 68.Qb8 Kh6 69.Qa7 Qf5 70.
Qa2 Qe5) +.10/16 1} 67. Qc8 {-0.09/14 0 (67.Qc8 Nd6 68.Qg8+ Kf5 69.Qh7+ Qg6 7.
Qd7+ Ke4 71.Qc6 Qf6 72.Qc5 Qe7 73.Qc6 Qf8) 0} Nd6 {1.13/15 0 (67. ... Nd6 68.
Qb8 Kf5 69.Qa7 Ke4 7.Qc5 Qf8 71.Kh1 Qd8 72.Bh2 Qf6 73.Qc6 Qh6 74.Bg1 Kd3 75.
Bh2 Qg6 76.Qc5 Qh6 77.Kg1) 0} 68. Qg8+ {-0.45/13 0 (68.Qg8+ Kh6 69.Qa8 e4 7.
Qc6 Qe5 71.Qc5 Kg6 72.Kh1 e3 73.Qxe3 Qxe3 74.fxe3) 0} Kh6 {1.33/16 1 (68. ...
Kh6 69.Qa8 e4 70.Qc6 Qe5 7.Kh1 e3 72.Qc1 Kg6 73.Qxe3) 1} 69. Qa8 {-0.73/14 0
(69.Qa8 e4 7.Qa7 Qe5 71.Qc5 Kg6 72.Kh1 Kf6 73.Kh2 e3 74.Qxe3 Qxe3 75.fxe3 Ke5)
0} e4 {1.23/16 1 (69. ... e4 70.Qa7 Qe5 7.Qc5 Kg7 72.Qa7+ Kg6 73.Qa2 Qd4 74.
Qa8 Kf7 75.Kh1 Kg7 76.Qc6 Qe5 77.Qc7+ Kg6 78.Qc5) 1} 70. Qa7 {-0.75/13 0 (7.
Qa7 Qe5 71.Qa2 e3 72.fxe3 Ne4 73.Qa6+ Kg7 74.Qb7+ Kg6 75.Qa6+ Nf6 76.Bf2) 0}
Qe5 {1.53/16 1 (70. ... Qe5 7.Qc5) 1} 71. Qc5 {-0.79/13 1 (7.Qc5 e3 72.Qxe3+
Qxe3 73.fxe3 Ne4 74.Bf2 Kg6 75.Kg1 Kf6 76.Kf1 Ke5 77.d6) 1} Kg6 {1.85/17 1 (7.
... Kg6 72.Kh1 e3 73.Qxe3) 1} 72. Qe3 {-1.42/13 0 (72.Qe3 Qxd5 73.Kh1 Nc4 74.
Qc5 Qxc5 75.bxc5 e3 76.fxe3 Kf5 77.Bf2 Ke4 78.c6)} Qxd5 {1.94/14 0 (72. ...
Qxd5 73.Qf4 Kg7 74.Kh1 Nc4 75.Qc7+ Kg6 76.Qc8 Kf7 77.Qc7+ Kf6 78.Qf4+)} 73. Kh1
{-1.65/14 1 (73.Kh Nc4 74.Qc5 Qe5 75.Qc6+ Nd6 76.Qc5 Qxc5 77.bxc5 Ne8 78.Kh2
e3 79.fxe3) 1} Nc4 {2.67/16 1 (73. ... Nc4 74.Qf4 e3 75.fxe3 f2+) +2.67/6 1}
74. Qf4 {-3.48/13 0 (74.Qf4 e3 75.fxe3 f2+ 76.e4 fxg1Q+ 77.Kxg1 Qd4+ 78.Kh2
Qd2+ 79.Qxd2 Nxd2 8.Kh1) 0} e3 {2.76/15 1 (74. ... e3 75.fxe3 f2+ 76.e4 fxgQ+
77.Kxg1 Qd4+ 78.Kh1 Qf6 79.Kg2 Ne5 80.Qd2 Nc6 81.Kh2 Qe7 82.Qe3 Nxb4 83.Qb6+
Kg7 84.Qxb5 Qxe4 85.Qxh5) 1} 75. fxe3 {-3.50/13 0 (75.fxe3 f2+ 76.e4 fxg1Q+
77.Kxg1 Qd4+ 78.Kh2 Qd2+ 79.Qxd2 Nxd2 8.Kh1 Nxe4 81.Kg2) 0} f2+ {3.06/15 1
(75. ... f2+) +3.06/5 1} 76. e4 {-3.50/13 1 (76.e4 fxgQ+ 77.Kxg1 Qd4+ 78.Kh2
Qd2+ 79.Qxd2 Nxd2 80.Kh1 Nxe4 81.Kg2 Kf5) 1} fxg1=Q+ {3.45/14 1 (76. ...
fxgQ+) 1} 77. Kxg1 {-3.29/5 0 (77.Kxg1 Qd4+ 78.Qf2 Ne3 79.Kh2)} Qd4+ {
adjudication} 0-1[/pgn]

Will be a full blog post on this "positional styles" test later...

matejst
Posts: 153
Joined: Mon May 14, 2007 6:20 pm

Re: Most Positional/Strategic Chess Engines

Post by matejst » Thu Dec 28, 2017 11:32 pm

Dear Brendan,

somehow, I don't think you have chosen the optimal type of positions to test these engines: why not try the QID, or the QG?

I tested most of the engines I wrote about at fixed depth (although "fixed depth" doesn't always means the same, but I tried atdifferent depths), with Karpov's repertoire. Modern engines prune too much sometimes, and I believe that many evaluation terms are removed to obtain a faster... bean counter?

In these positions, where there were less tactical motives, Wasp 1.25 emerged as a clear winner. In openings like the Queen Indian, Zarkov and Wasp showed the most "human-like" evaluation, while they had problems in the QID and the Tchigorin Ruy Lopez, openings they basically don't understand. I also used these engines to analyze my own repertoire: the Zarkov/Wasp franchise is basically sound, with an eval very similar to Komodo, often better.

To analyze the KID, or test an engine in this kind of positions, I would rather use Gandalf 7b or, if you really want, the Baron, but with less pruning and with learning on. Gandalf 7b remains an exceptional engine, blending tactics and strategy. With a modern search (and a 32 bit version for my linux system), it would have a special place among my favourite engines. I didn't test the Baron with all the options it has, although I believe there's a lot to be found.

I analysed recently quite a few games from the TCEC: TBs acted as a surrogate to positional knowledge very often. I am certain Wasp would have won at least two games without them, since the engines it was playing against just looked lost before being able to reach TBs positions in their calculations.

There is an interesting debate about learning in another thread and, imho, every author should implement it in a way or another -- it doesn't bring Elo right away, but in the long run, it's probably the way to go. Ed Schroder pointed to new paths, but the dominant paradigm -- greater depths and better rating -- is difficult to change.

Twenty years ago, chess playing programs were products: an opening book, a GUI, learning abilities very often and they were worth buying. Today, when everybody focuses on "engines" and "speed", it became meaningless: I need time to analyse positions myself: to let an engine analyse a few minutes more and reach a good depth is natural.

Lucas Monk, Ed Schroder, Mark Uniacke, are probably the one that got it right: we don't really need bases with millions of games, engines that play meaningless if they don't achieve depth 30 and there are no tactics involved, and TBs should be an option. There has to be a middle way. The Rybka-Zappa match demonstrated it clearly.

And since I feel buying again a product, I'll buy something more pleasant, comfortable, and meaningful than the combination of ChessX and an old, free engine. Something like Genius 5, Nimzo 3.5, or Rebel Decade. Something that makes you play, or, when you don't have enough time, like me, analyse.

BrendanJNorman
Posts: 1400
Joined: Sun Feb 07, 2016 11:43 pm
Full name: Brendan J Norman

Re: Most Positional/Strategic Chess Engines

Post by BrendanJNorman » Fri Dec 29, 2017 3:17 am

matejst wrote:Dear Brendan,

somehow, I don't think you have chosen the optimal type of positions to test these engines: why not try the QID, or the QG?
Hey Buddy,

Here's my explanation...

QID has sharp Topalov-like lines if white wants, QG has Bf4 with 0-0-0 or Exchange with f3/e4, or Minority Attack stuff.

I wanted a certain type of position where black would not be defending per sey, but making his own ideas on how to proceed - the same for white.

The Old Indian which I used, was used by guys like Capablanca and Petrosian decades ago as a "solid" response to 1.d4 and later by Bronstein as a safe way to get into the King's Indian without allowing the Samisch Attack (f3/Be3/Nge2/0-0-0 etc).

I think it's a good choice for testing, and it must be admitted, that Alfil had to use great creativity to earn that tactical win against Wasp 1.25.
matejst wrote:I tested most of the engines I wrote about at fixed depth (although "fixed depth" doesn't always means the same, but I tried at different depths), with Karpov's repertoire. Modern engines prune too much sometimes, and I believe that many evaluation terms are removed to obtain a faster... bean counter?

In these positions, where there were less tactical motives, Wasp 1.25 emerged as a clear winner. In openings like the Queen Indian, Zarkov and Wasp showed the most "human-like" evaluation, while they had problems in the QID and the Tchigorin Ruy Lopez, openings they basically don't understand. I also used these engines to analyze my own repertoire: the Zarkov/Wasp franchise is basically sound, with an eval very similar to Komodo, often better.
Here are some openings I think are good to test for "human" positional play.

1.d4 Nf6 2.c4 d6 3. Nc3 Nbd7 4.e4 e5 (maneuver in cramped, but solid position with black)

1.e4 c5 2.c3 d5 3. exd5 Qxd5 4.d4 Nf6 5.Nf3 e6 6.Bd3 cxd4 7.cxd4 (play vs IQP with black)

1. e4 e6 2.d4 d5 3.Nc3 Bb4 4.e5 b6 5.Qg4 Bf8 (another Petrosian fav - find the best method of development+piece squares for black)

matejst wrote:To analyze the KID, or test an engine in this kind of positions, I would rather use Gandalf 7b or, if you really want, the Baron, but with less pruning and with learning on. Gandalf 7b remains an exceptional engine, blending tactics and strategy. With a modern search (and a 32 bit version for my linux system), it would have a special place among my favourite engines. I didn't test the Baron with all the options it has, although I believe there's a lot to be found.
Have you tried Frenzee 3.5.19? It it very similar to Gandalf in this way and around the same strength - definitely worth a try.
matejst wrote:I analysed recently quite a few games from the TCEC: TBs acted as a surrogate to positional knowledge very often. I am certain Wasp would have won at least two games without them, since the engines it was playing against just looked lost before being able to reach TBs positions in their calculations.

There is an interesting debate about learning in another thread and, imho, every author should implement it in a way or another -- it doesn't bring Elo right away, but in the long run, it's probably the way to go. Ed Schroder pointed to new paths, but the dominant paradigm -- greater depths and better rating -- is difficult to change.

Twenty years ago, chess playing programs were products: an opening book, a GUI, learning abilities very often and they were worth buying. Today, when everybody focuses on "engines" and "speed", it became meaningless: I need time to analyse positions myself: to let an engine analyse a few minutes more and reach a good depth is natural.
Interesting approach, I often fear limiting depth with the current engines will indirectly affect playing style in some way, but maybe it won't.
matejst wrote:Lucas Monk, Ed Schroder, Mark Uniacke, are probably the one that got it right: we don't really need bases with millions of games, engines that play meaningless if they don't achieve depth 30 and there are no tactics involved, and TBs should be an option. There has to be a middle way. The Rybka-Zappa match demonstrated it clearly.

And since I feel buying again a product, I'll buy something more pleasant, comfortable, and meaningful than the combination of ChessX and an old, free engine. Something like Genius 5, Nimzo 3.5, or Rebel Decade. Something that makes you play, or, when you don't have enough time, like me, analyse.
Yep all agreed.

I'm pretty content with engines at the moment, but I'd really love a new GUI with features for testing and running tournaments (swiss tournaments, team matches - so positional vs attacking team matches, or programmer countries!) and stuff like this with nice Retina graphics for my laptop - that'd be awesome.

At some point, I might just create the product myself (via freelancing programmers) as a commercial project.

matejst
Posts: 153
Joined: Mon May 14, 2007 6:20 pm

Re: Most Positional/Strategic Chess Engines

Post by matejst » Fri Dec 29, 2017 9:51 am

Dear Brendan,

Written words can be very treacherous, allowing lots of misunderstandings.
I often fear limiting depth with the current engines will indirectly affect playing style in some way, but maybe it won't.
I just meant that the engines on the top of the rating lists are the one with the fastest search. It was something some authors tried to hide in the late '90 and at the beginning of this millennium.
Have you tried Frenzee 3.5.19?
You know my philosophy: I stick to a few engines only. It's the same with books: nowadays, I prefer reading the same novels again and again, to find what I have missed before and how much I have changed myself. I tried Rodent, but its game in simple positions -- and I like to check this kind of positions because I find them extremely difficult to play -- was not what I expected, and my tweaking didn't achieve anything. So, for the time been, I'll try the Baron with the option "prune 3", then "prune 2" -- it seems to be better in sharp positions, detecting threats at lower depths, while still playing well sensitive middlegame positions.

About a GUI: I read the thread about a new, modern interface. But it was overwhelming -- everybody wanted something different. If you decide to do something about, first, you have to know that you'll end loosing money; then, there are projects that could be continued, like Jose. Nobody here mentions it, because it's not tailored for engine matches, but it could be a good interface for chess players.

User avatar
Ovyron
Posts: 2495
Joined: Tue Jul 03, 2007 2:30 am

Re: Most Positional/Strategic Chess Engines

Post by Ovyron » Fri Dec 29, 2017 10:01 am

Jose? What was Jose good for? I downloaded all the Database GUIs I could find and Jose was much inferior to Scid...

Scid is open source, I wonder if one could Fork Scid into an engine playing version...

Also, what about Lucas Chess? It's already up there, I have managed to play engine matches without problems, and it is also Open Source, I think.

If Lucas Chess was picked, it urgently needs Winboard engines support...

PK
Posts: 820
Joined: Mon Jan 15, 2007 10:23 am
Location: Warsza
Contact:

Re: Most Positional/Strategic Chess Engines

Post by PK » Fri Dec 29, 2017 10:42 am

I admit I have problems with making Rodent play positional chess - plus a meta-problem with describing what's wrong, but I will try, hoping for some exchange of ideas.

Rodent seems to have decent understanding of outposts, and some special case code that I'm proud of, that allows it to handle pawn chains in King's Indian type of positions (not that it's perfect, but similar code for French Defense pawn chain works much worse). Recent versions tend to like fianchetto, to the point that when I force 1.e4 c5 with no book, I get either accelerated Dragon or Rossolimo as an attempt to avoid it. Again, there is some special case code to discourage fianchetto that attacks defended enemy pawn.

This on the plus side. On the minus side, I still have problems with telling my engine to keep its pawn mass compact enough (though I try to fix it with each major release). For some reason, Rodent is unwilling to keep tension and exchanges central pawns too soon. And there is a major obstacle for achieving good positional style, namely relatively weak endgame play.

Beside that I have an impression that modern, variable depth search with lots of pruning is really not that good in comparing quiet positions. It seems that it even requires certain amount of exaggeration to work well. If you force Stockfish to play at shallow depths, its king safety scores look ridiculous. On the other hand, if you allow it to play at its normal depth, allowing it to see that quick and simple attacks tend to peter out, it becomes much more positional.

Positional engine might require much lower evaluation weights than are common nowadays, but if you do that, you lose all the benefits of much-needed score exaggeration. You increase the risk of aimless shuffling as well.

So I wonder if a positional chess engine should have search somewhat similar to good old Fruit 2.1, with its shallow late move reduction, further constrained by history scores. But it is not good for strength.

BrendanJNorman
Posts: 1400
Joined: Sun Feb 07, 2016 11:43 pm
Full name: Brendan J Norman

Re: Most Positional/Strategic Chess Engines

Post by BrendanJNorman » Fri Dec 29, 2017 10:53 am

PK wrote:I admit I have problems with making Rodent play positional chess - plus a meta-problem with describing what's wrong, but I will try, hoping for some exchange of ideas.

Rodent seems to have decent understanding of outposts, and some special case code that I'm proud of, that allows it to handle pawn chains in King's Indian type of positions (not that it's perfect, but similar code for French Defense pawn chain works much worse). Recent versions tend to like fianchetto, to the point that when I force 1.e4 c5 with no book, I get either accelerated Dragon or Rossolimo as an attempt to avoid it. Again, there is some special case code to discourage fianchetto that attacks defended enemy pawn.

This on the plus side. On the minus side, I still have problems with telling my engine to keep its pawn mass compact enough (though I try to fix it with each major release). For some reason, Rodent is unwilling to keep tension and exchanges central pawns too soon. And there is a major obstacle for achieving good positional style, namely relatively weak endgame play.

Beside that I have an impression that modern, variable depth search with lots of pruning is really not that good in comparing quiet positions. It seems that it even requires certain amount of exaggeration to work well. If you force Stockfish to play at shallow depths, its king safety scores look ridiculous. On the other hand, if you allow it to play at its normal depth, allowing it to see that quick and simple attacks tend to peter out, it becomes much more positional.

Positional engine might require much lower evaluation weights than are common nowadays, but if you do that, you lose all the benefits of much-needed score exaggeration. You increase the risk of aimless shuffling as well.

So I wonder if a positional chess engine should have search somewhat similar to good old Fruit 2.1, with its shallow late move reduction, further constrained by history scores. But it is not good for strength.
I have made a few more attempts at positional personalities for Rodent, with interesting results.

One is just 2200 level (16000 NPS limit), but against humans is very convincing (plays much like the strong 2200-2300 master you might meet in an open tournament), another is his "Uncle" who is similar, but no NPS limit.

They're to be called "Positional Vlad" and "Positional Sergei" :lol:

I also made one called "Dynamic Pawns" which handles pawns in a very interesting dynamically positional way, but is not tactically aggressive.

Would have sent them (as well as sample games) already, but I know you're busy preparing the new Rodent release - I'll send them to you when you're done?

Post Reply