Stockfish 4 PA_GTB

[Jeremy Bernstein]

Stockfish 4 PA_GTB v001 is released!

No changes from the previous PA_GTB build based on Stockfish 3 beyond updating to the tip of the Stockfish upstream trunk.

Jeremy

[Luigi335]

Thank you, Jeremy!

[User923005]

This thing is really great for analysis.
I have a few million analyzed positions, and as the count grows, the analysis speed continues to accelerate.
I typically get to depth 30 (+/-2ply) in a minute.

Here is a sample analysis:
This is row 765 from one of my current analysis files:
r1bq1rk1/ppp2pp1/2np1n1p/4p3/1b2P3/2NP1NP1/PPP2PBP/R1BQ1RK1 w - -

Code: Select all

765                     
    Avoid move: 
    Best move (Stockfish-4-64-PA_GTB-001): Nc3-d5
    Not found in: 01:00
     2/2	00:07	         365	51	-0.06	Bd2 Bg4
     3/3	00:07	         714	100	-0.09	Bd2 Bg4 h3
     4/4	00:08	       2,090	287	-0.25	Be3 Be6 Qe1 d5
     5/5	00:08	       2,721	373	-0.20	Be3 Be6 Qe1 Bxc3 Qxc3
     6/6+	00:08	       6,414	876	-0.12	Bd2 Be6 a3 Bc5 Na4
     6/6	00:08	       6,919	944	-0.20	Bd2 Be6 a3 Bc5 b4 Bb6
     7/7	00:08	       8,177	1,054	-0.19	Bd2 Be6 a3 Bc5 b4 Bb6 Kh1
     8/8+	00:08	       9,145	1,174	-0.11	Bd2 Be6 a3 Bc5 b4 Bb6 b5
     8/8	00:09	      16,117	2,066	-0.09	Bd2 Be6 a3 Bc5 Na4 Bd4 c3 Bb6 Nxb6 axb6
     9/9	00:09	      23,862	2,863	-0.13	Na4 Ba5 c3 Bg4 b4 Bb6 Bd2 Re8 Nxb6 axb6
    10/11+	00:09	      33,212	3,966	-0.05	Na4 Ba5 c3 Bg4 b4 Bb6 h3 Be6 Bd2
    10/11	00:09	      41,226	4,918	-0.02	Na4 Ba5 c3 Bg4 b4 Bb6 h3 Be6 Bd2 Bd7 Nxb6 axb6
    11/13-	00:09	      78,679	9,314	-0.10	Na4 d5 exd5 Nxd5 a3 Bd6 h3 Be6
    11/13	00:09	     147,176	16,506	-0.16	a3 Ba5 Bd2 Be6 b4 Bb6 Na4 d5 exd5 Nxd5 c3
    12/16+	00:09	     218,233	24,253	-0.08	a3 Ba5 Nd5 Nxd5 exd5 Ne7 d4 Nxd5 dxe5 Be6 exd6 cxd6 Nd4
    12/16	00:10	     310,145	34,258	-0.18	a3 Ba5 Nd5 Be6 c4 Bb6 Be3 Bd7 b4 Bxe3 fxe3 a6 Nxf6+ Qxf6
    13/16	00:10	     358,261	38,011	-0.18	a3 Ba5 Nd5 Be6 c4 Bb6 Be3 Bd7 b4 Bxe3 fxe3 a6 Nxf6+ Qxf6
    14/19+	00:10	     568,381	59,404	-0.10	Nd5 Bc5 c3 Bb6 Nxb6 axb6 b4 Bg4 h3 Bd7 Bd2
    14/19	00:10	     629,620	65,503	-0.09	Nd5 Bc5 c3 Bb6 Nxb6 axb6 b4 Bg4 Re1 d5 exd5 Nxd5 Qb3 Re8 Bd2
    15/19	00:11	     710,579	70,994	-0.05	Nd5 Bc5 c3 Bb6 Nxb6 axb6 b4 Bg4 b5 Ne7 Bd2 Ng6 Re1 Qc8 a4 Qe6 c4
    16/23+	00:11	   1,016,115	99,805	+0.03	Nd5 Bc5 c3 Bb6 a4 Nxd5 exd5 Ne7 d4
    16/23+	00:11	   1,591,802	153,278	+0.11	Nd5 Bc5 c3 Bb6 a4 Nxd5 exd5 Ne7 d4 e4 Nh4 Nxd5 a5
    16/23	00:11	   1,956,808	185,796	+0.15	Nd5 Bc5 c3 a5 Be3 Bxe3 fxe3 Bd7 Nd2 Nxd5 exd5 Ne7 e4 Ng6 Nc4 Kh7 Qf3
    17/23-	00:11	   2,268,936	210,848	+0.07	Nd5 Bc5 c3 a5 Be3 Bxe3 fxe3 Bd7 Nd2 Nxd5 exd5 Ne7 e4 f5 Qh5 f4
    17/23	00:12	   2,452,557	225,792	+0.04	Nd5 Bc5 c3 a5 Be3 Bxe3 Nxe3 Bd7 Nd2 a4 Re1 Re8 Bf3 Be6 Nd5 a3 b4 Bxd5 exd5
    18/23	00:12	   3,062,642	274,553	 0.00	Nd5 Bc5 c3 a5 Be3 Bxe3 Nxe3 Bd7 Nd2 a4 Re1 Re8 Bf3 Be6 Bg2 Bd7 Bf3 Be6 Bg2 Bd7
    19/23+	00:12	   3,382,843	292,481	+0.08	Nd5 Bc5 c3 a5 Be3 Bxe3 fxe3 Bd7 Nd2 Nxd5 exd5 Ne7 d4 Bb5 Rf2 exd4 cxd4
    19/23+	00:12	   3,816,527	325,004	+0.16	Nd5 Nxd5 exd5 Ne7 d4 Nxd5 dxe5 Be6 Nd4 dxe5 Nxe6 fxe6 c4
    19/24	00:12	   4,785,401	396,536	+0.13	Nd5 Nxd5 exd5 Ne7 d4 f6 c3 Ba5 Qb3 Kh8 Bd2 Bb6 Rfe1 Bd7 Nh4 g5 Nf3 Bf5 Be3 c6 dxc6 bxc6
    20/27	00:14	  13,070,721	963,420	+0.07	Nd5 Bc5 c3 a5 Be3 Bxe3 fxe3 Nxd5 exd5 Ne7 Nd2 c6 dxc6 bxc6 Nc4 Be6 Qf3 Bxc4 dxc4 Rb8 Rab1 f5
    21/27	00:14	  14,605,590	1,054,479	 0.00	Nd5 Bc5 c3 a5 Be3 Bxe3 Nxe3 Bd7 Nd2 a4 Re1 Re8 Bf3 Be6 Bg2 Bd7 Bf3 Be6 Bg2 Bd7
    22/27	00:14	  16,231,882	1,146,400	 0.00	Nd5 Bc5 c3 a5 Be3 Bxe3 Nxe3 Bd7 Nd2 a4 Re1 Re8 Bf3 Be6 Bg2 Bd7 Bf3 Be6 Bg2 Bd7
    23/27	00:15	  20,008,529	1,357,246	+0.02	Nd5 Bc5 c3 a5 Be3 Bxe3 Nxe3 Bd7 Nd2 a4 Re1 Re8 Bf3 Be6 Qc2 Bd7 b4 b5 a3 Rb8 Nf5 Qc8 Qa2 Bxf5 exf5
    24/34	00:16	  33,024,893	2,020,983	 0.00	Nd5 Bc5 c3 a5 Be3 Bxe3 Nxe3 Re8 Re1 a4 Nh4 Bd7 Bf3 Ne7 Qc1 Nc6 Nhf5 Ne7 Nh4 Nc6 Nhf5 Ne7 Nh4
    25/34	00:17	  39,139,055	2,291,647	 0.00	Nd5 Bc5 c3 a5 Be3 Bxe3 Nxe3 Re8 Re1 a4 Nh4 Bd7 Bf3 Ne7 Qc1 Qc8 Qd1 Qd8 Qc1 Qc8 Qd1 Qd8
    26/34	00:19	  59,409,953	3,057,483	+0.01	Nd5 Bc5 c3 a5 Be3 Bxe3 Nxe3 Re8 Re1 a4 Nh4 Bd7 Qe2 Be6 Nhf5 a3 b3 Ne7 Nxe7+ Qxe7 Rab1 Bd7 Bf3 Qe6 Nf5 d5 Rbd1 Qb6
    27/34	00:21	  71,120,396	3,423,694	+0.01	Nd5 Bc5 c3 a5 Be3 Bxe3 Nxe3 Re8 Re1 a4 Nh4 Bd7 Qe2 Be6 Nhf5 a3 b3 Ne7 Nxe7+ Qxe7 Rab1 Bd7 Bf3 Qe6 Nf5 d5 Rbd1 Qb6
    28/37	00:24	 104,567,984	4,287,681	+0.04	Nd5 Bc5 c3 a5 Be3 Bxe3 Nxe3 Ng4 Nd5 Re8 Re1 Ne7 Ne3 Nxe3 Rxe3 Bg4 d4 Ng6 d5 Qd7 c4 f5 Qb3 fxe4 Nd2 a4 Qxb7 Reb8
    29/39	00:30	 155,277,007	5,193,037	+0.06	Nd5 Bc5 c3 a5 Re1 Re8 Be3 Bxe3 Nxe3 a4 Nh4 Bd7 Qe2 Qc8 b4 a3 Bf3 Ne7 Rab1 Bb5 Bg2 Bd7 d4 exd4 cxd4 Nc6 d5 Ne5 Nhf5 Ng6 Qc2 Bxf5 exf5 Ne5
    30/39	00:38	 223,430,732	5,885,641	+0.10	Nd5 Bc5 c3 a5 Re1 Re8 Be3 Bxe3 Nxe3 a4 Nh4 Bd7 Qe2 a3 b3 b5 Nhf5 b4 cxb4 Rb8 Rac1 Nxb4 f4 Bxf5 Nxf5 c5 fxe5 dxe5 Red1
    31/41	00:51	 349,689,766	6,768,145	+0.12	Nd5 Bc5 c3 a5 Re1 Re8 Be3 Bxe3 Nxe3 a4 Nh4 Bd7 Qe2 a3 b3 b5 Nhf5 b4 cxb4 Rb8 Rac1 Nxb4 f4 Nc6 Qc2 Nb4 Qd2 exf4 gxf4 Bxf5 Nxf5 c5
   10/16/2013 2:39:31 AM, Time for this analysis: 00:01:00, Rated time: 12:45:00

Without the permanent hash feature, analysis could not be accelerated in this way.
I thank you sincerely for this very interesting improvement.

[Thebossss]

Hi Jeremy Bernstein;

Thanks for the file..

Can u pls provide the polyglot adaptor with the settings(soft probe depth etc used in PA_GTB), without engine so that i will use polyglot.ini file to link to the latest built engines.
Because i see u used 10/10/2013 built engine but there are strong builts later but the settings u've used like soft probe depth, hard probe, persistent hash adds weightage to the engine.
So in short i expect polyglot adaptar without engine, so i use polyglot.ini to link engine file (Latest builts)..
I know u r not benifited for sending me this.. but i have no options..(I tried to compile day and night for 2 days but ended with errors)

Thanks hoping that u would send me, waiting for ur reply
You can PM me also..

Regards:
Thebossss

[petevine]

Hi,
Is it currently possible for 2 (or more) instances of stockfish to access and write to the same PH file?
Thx

[Jeremy Bernstein]

Hi,
Is it currently possible for 2 (or more) instances of stockfish to access and write to the same PH file?
Thx

Not a good idea, but I haven't tested it.

jb

[petevine]

There's one thing left that would make the app complete as an analysis tool - being able to manually insert position evals (from TB7 for example). This feature might ilustrate why a locking mechanism is probably required.
Anyways, thanks for a great piece of code!

[Jeremy Bernstein]

There's one thing left that would make the app complete as an analysis tool - being able to manually insert position evals (from TB7 for example). This feature might ilustrate why a locking mechanism is probably required.
Anyways, thanks for a great piece of code!

You can do this via an EPD file. Check out the "importepd" command in the readme.

jb

[petevine]

My bad
Is it safe then to use this option on the console while another instance is analysing?

[Jeremy Bernstein]

My bad
Is it safe then to use this option on the console while another instance is analysing?

Nope. Well, it's untested. There are no protections in my code to deal with simultaneous write access to the PH database. Whether the Komodo Cabinet backend is robust enough to handle it, no idea. But this wasn't an imagined use case for me.

jb