Opening Book Project: v01 after Houdini analysis available!

[Frank Quisinsky]

Hi there,

now the first engine:
Houdini 5.01 Pro x64 with 1 minute per position on i7 4.3 Ghz with 6 cores + 4x hyperthreading analyzed 26.146 positions and found 3.20% "bad lines" with 0.65 / - 0.40 settings.

Note: Stockfish and Komodo analyzed the same positions before with 30 seconds per position on i7 4.0 Ghz with 4 Cores + 4x hyperthreading (for find bigger blunders only with the half of time and lesser cores).

As next Shredder 13 x64 will analyze the same "beta-v03_endpos-n.epd" file.

More information to my opening book project can be found on my webpage under:

Frank's Chess Page
http://www.amateurschach.de

Best
Frank

A nice example to see that more as one engine should analyze positions used for an opening book.

[Frank Quisinsky]

Code: Select all

r2qkb1r/1b1n1ppp/pp1ppn2/8/Q1PNP3/2N1BP2/PP4PP/R3KB1R w KQkq - id "2290"; ce 105; acd 24; acs 60; acn 997787083; pv g4 ;
r4rk1/1bq1bpp1/p2ppn1p/1pn5/3NPPPB/P1N2Q2/1PP1B2P/2KR2R1 w - - id "12033"; ce 101; acd 21; acs 60; acn 1067212334; pv g5 hxg5 ;
r1b1r1k1/pp2qpbp/2pp1np1/2n5/2PNP3/1PN1BPP1/P5BP/R2Q1RK1 w - - id "24794"; ce 97; acd 22; acs 60; acn 1008755763; pv Lf2 ;
r3kb1r/1bq2ppp/p3pn2/1pP5/1n6/2NBPN2/PP2QPPP/R1BR2K1 w kq - id "17962"; ce 95; acd 24; acs 60; acn 1039549517; pv Sxb5 ;
r1b2rk1/2qnbppp/p2ppn2/1p4B1/3NPPP1/P1N5/1PP1Q2P/2KR1B1R w - - id "11899"; ce 94; acd 21; acs 60; acn 1006995720; pv Lh4 ;
r1bq1rk1/pp1nb1pp/4p3/n2pPp2/5P2/PNN1B2Q/1PP3PP/R3KB1R w KQ - id "9263"; ce 93; acd 24; acs 60; acn 995618806; pv Sd4 ;
2r3k1/p2bppbp/3p1np1/q6P/1prNP1P1/4BP2/PPPQN3/1K1R3R w - - id "10827"; ce 93; acd 22; acs 60; acn 1059907224; pv Tc1 Se8 hxg6 fxg6 Sb3 Db5 Sf4 De5 Dd5+ e6 Dxe5 Lxe5 Sd3 Lf6 g5 Ld8 Sd2 T4c7 Sxb4 Tb7 c3 a5 Sd3 e5 f4 exf4 Lxf4 ;
r1b1k2r/pp1p1ppp/4pn2/q1n5/2P5/P1b2P2/1PQBP1PP/R3KBNR w KQkq - id "23296"; ce 93; acd 26; acs 60; acn 1098127485; pv Lxc3 ;
rn1q1rk1/pppb1pbp/3p1np1/3Pp1N1/2P1P3/2N1B1P1/PP2BP1P/R2QK2R w KQ - id "25649"; ce 93; acd 23; acs 60; acn 1001669102; pv h4 ;
r1b1r1k1/pp2qpbp/2pp1np1/4n3/2PNP3/2N3PP/PP2QPB1/R1BR2K1 w - - id "24836"; ce 92; acd 24; acs 60; acn 1040951634; pv Le3 ;

Examples:
For Stockfish and Komodo are the TOP-10 of "bad positions" OK ...
Lesser as 0.50 ...

10 engines will analyze the same database!
Now let us look what Shredder can find.
Need around 20 days again!

The complete files to my project can be found on my webpage ...

Best
Frank

[Frank Quisinsky]

Hi there,

if you like to help ...
Note: With 1 minute per position your comp is still running around 20 days!

Please give your computer the full power for the 26.146 positions. The important "beta-v03_endpos-n.epd" file can be found in my download file. I hope I can add later engine analyzes by others.

Please use the engine you like.

With more engine analysis we can create a perfect opening book with balanced start positions. In the beta-3 database are positions to all of the 500 ECO codes, three moves after ECO code was formed. Working many years on it and thinking that I have collected most important start positions. Many new start positions can be never find in GM databases, comes by engines itself.

With the new tool by Ferdinand all is very easy!
Most important work for this project is done, engine analyzes are missed and need a while.

All stats we can create later will be very interesting.
I hope that others will help a bit.

Best
Frank

[leavenfish]

Any 'book' that can't include my favorite Bent Larsen contribution: 1. g3 e5, 2. Bg2 d5, 3. Nf3 (often aiming for a reversed Alekhine...tempo up) doesn't get the time of day from me!

But seriously (?) very interesting project.

[Frank Quisinsky]

Hi Brian,

after 3. Nf3 Nc6, 4. d3
rarely but I have the position with move transposition if you are looking a bit longer.

In my Shredder *.bkt book ...
Modul 1 (Basic modul) ... your move is included.

after 3. Nf3 ... with e4 or again Nc6.

a00-e99_samarian-barnes-quisinsky.bkt
But will be deactivated if ready because not a balanced position after e4. So, all is right that I have the position only with move transpostion after Nc6 4.d3.

Important here, that all first possible moves are included if users have interest to see that, length 4 moves only. So users can activate decativated lines.

Modul 2 (database modul) can be copy later into that book if ready, means after all the engine analyzes.

I will start working on "Modul 1" in detail with priorities next month.

Best
Frank

[Frank Quisinsky]

Note:
Later the holes in a balaced book have a reason (reject by engines or in other cases ... to danger for 3-fold repetition). And this is right for a balanced book.

Never such a book are available in the past.
And a better book for engine-engine testing we can't created!

... the main interest ... to have the playable balanced positions to all of the 500 ECO codes in one opening book ... easy to changed in the future (database to that book is available) ... with all important move transpositions for it.

Best
Frank

[Tdunbug]

This may be a dumb question. But what program do I use for the opening book analysis? I am thinking of donating my 24 core to the project.

[Frank Quisinsky]

Hi Terence,

I am using Shredder 12 GUI with a small problem. After 5181 positions GUI stopped the still running analyzes.

I delete then the 5181 positions from EPD file and start the analyzes again with position 5182 and so one. Of course I saved the results by Shredder in a special directory before.

24 cores is a lot!
12 cores with 3Ghz and one minute per position is more as enough system power in my opinion, your decision!

It would be great if I have later *.epd files by others with the complete 26.146 analyses of positions. I need alone around 7 months for 10 of such engine files.

To added all the results in one Excel table will be great for looking and further stats we can do with it.

Important is the file:
beta-v03_endpos-n.epd

Can be found in the beta directory of the 30Mb download file (have a look on my new opening book site on my webpage) or in my download selection, included in game extractor 1.3.01 download file by Ferdinand Mosca.

Thanks for your interest and sorry for my bad English again.

Best
Frank

PS:
With Shredder 13 GUI I have grafic problems on all of my computers. The reason that I am using Shredder 12 GUI (bugfree).

[Tdunbug]

I am also having small graphical issues with Shredder 13 but the analysis is going smoothly. When done I will make available the completed .epd

[Dann Corbit]

From your book, these positions have a score in absolute value of 1/2 pawn or more:
{see file frank-high.epd in the following link}:
http://rybkaforum.net/cgi-bin/rybkaforu ... pid=572461

The file frank-zero.epd contains positions with a computer evaluation of zero. These will result in nearly all draws, and some substantial percentage of them will be forced draws.

talkchess can't hold much in the text box and can't accept attachments.