Release: FEOBOS v20 Contempt 1-5 without move transp. avail.

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

Moderators: hgm, Rebel, chrisw

Frank Quisinsky
Posts: 6808
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

Release: FEOBOS v20 Contempt 1-5 without move transp. avail.

Post by Frank Quisinsky »

Hi there,

can be found under:

Frank's Chess Page
http://www.amateurschach.de/main/_new-opening-book.htm

All what we create in Excel can be seen in our Dia-Show. All the new ideas for opening book developments.

Image

Have a look under "FEOBOS Project Cockpit"

current (with books)
FEOBOS v20
without move
transpositions
42,27Mb

We released the FEOBOS opening book with Contempt 1-5 in "bkt, polyglot bin and Chessbase ctg format.

All project files are available.
*.pgn sorted by ranking (again thanks Ferdinand).
The complete development of the Excel files (again thanks to Klaus).
and all other files, over 400.000 engine analysis ...

Excel with 11 millon formulas.

Not ready:
- FEOBOS v20 with move transpositions, I hope in the next 4-5 days.
- Project documentation (only in German language, over 70 sites, think so). I hope end of February.

The cut for the *.pgn file sorted by FEOBOS ranking:

Image

And here the readme to the Contempt 1-5 books in my bad english:

Code: Select all

FEOBOS v20
Tool by Ferdinand Mosca, Philippines
------------------------------------

Ferdinand Mosca programmed a very useful tool for FEOBOS project!

#1+#########!!##
THANKS FERDINAND
#1+#########!!##

With the tool by Ferdinand Mosca the ranking number of analyzed FEOBOS end positions
can be add and sort directly in *.pgn file. For databases without move transpositions
it's no longer necessary to delete "bad" or "0.00 draw" positions by hand in database
systems like Scid or Chessbase.

With the following additional big deal: We can create any "Test-Set" constellation
directly with the *.pgn file!

-	spf.exe (program)
-	run_spf.bat (batch file)
-	feobos_without-dublicates_41614-after-upd.1.pgn (FEOBOS database)
-	feobos_v20-41614_ranking.txt (urgently needed and created with worksheet
    "Set by hand / FEOBOS")
	
Result after running spf.bat:

-	unsorted_feobos_v20-41614-ranking.pgn
-	feobos_v20-41614_ranking.pgn
-	
Please click under worksheet "Set by hand" on button:  "Create text file for 41.614
positions"! On worksheet "FEOBOS" the required information added!  The complete ranking
can be seen in worksheet "Eval out", here _Open sorting tool_ / Sort by ranking_ and
now copy & paste all to from worksheet FEOBOS to clipboard and add it in a new *.txt
file with an editor like Notepad++.

Notepad++ can be found under: https://notepad-plus-plus.org/

Have a look in the following graphics:
1. _feobos-ranking-system.png (explanation of ranking system)
2. _feobos_v20-41614_ranking__cut.png (cut for contempt 1-5 opening books)
3. _ranking-settings.png (6 of 8 ranking criteria can be set / changed)
4. _automatically-engine-settings.png (engine settings)
5. _eval-out_automatically-setting.png (automatically setting must be select) 

Point 1 is improved with new criteria.
Point 2 is improved, because we closed the holes with upd.1 database.
Point 3 is clearly improved since the latest available FEOBOS v10 version.
Point 4 is new for FEOBOS! In worksheet "Settings" engines calculate here own setting
full automatically! "75% of highest white evaluation / 25% of highest black eval. We
gave as setting 1.8%.

With the result that 1.8% of 41.614 positions will be located as "bad move". With
another new option "rescue", positions located as bad move can find the way back in
our database if 1/10 engines located this move as bad move only and in additional
the evaluation isn't higher as 0.10 / -0.05 from the automatically engine setting.
So, we are working tricky because we haven't interest to lose interesting opening
positions.



But most interesting is:
10 completely different and very strong chess engines optimized in team work an
opening book with any decision by humans "what's bad or good". The source of error
is max. minimized.

Have fun with our FEOBOS project and your engine-engine match constellation, tourneys,
rating systems with a strong and by 10 engines analyzed database of A00-E99 opening
position.

FEOBOS:
- best lines 3 moves after ECO code was formed
- new developed ranking system
- engines optimized in teamwork automatically an opening book
- any conceivable test-set can be created
- all *.epd / *.pgn / excel files and engine analysis are available
- *.pgn file is sorted by number of ranking
- ECO code statistics
- engine comparison
- move statistics
- KECR ratings, generated with "best moves"
- all what's  important can be changed with own settings in FEOBOS excel file.
- *.epd analyzed tool under Excel.
- automatic test-set tool or test-set by random.
- developed Excel files with 11 million formulas (most in macro's)
- and much more ...

Note: With the tool by Ferdinand FEOBOS opening books without move transpositions can
      be created automatically. For the opening book with move transposition many work
	  by hand is necessary.


Klaus Wlotzka & Frank Quisinsky (Germany)
January 28th, 2018 (16:00)
Best
Frank
Frank Quisinsky
Posts: 6808
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

Re: Release: FEOBOS v20.1 available, sorry!

Post by Frank Quisinsky »

Hi there,

1. We found a mistake in A00 calculation in our Excel file.
2. The automatically engine Settings for reject positions are not optimal.
3. Readme files are better, to many spelling mistakes.

So we changed that and released version 20.1 opening books for contempt 1-5 in Chessbase, Shredder and Polyglot format.

Should be better now!

Best
Frank
User avatar
pohl4711
Posts: 2434
Joined: Sat Sep 03, 2011 7:25 am
Location: Berlin, Germany
Full name: Stefan Pohl

Re: Release: FEOBOS v20.1 available, sorry!

Post by pohl4711 »

On my website (SALC openings-section) 1000 games testruns of FEOBOS v20 contempt 5, SALC V5 and the Stockfish Framework 8move openings (for comparsion) can be found - just scroll down a little bit...
Very disappointing result of FEOBOS contempt 5 release-version. But - of course - my bad testing-conditions are the reason for that, says the author of FEOBOS...(some things really never change in computerchess...LOL))


http://www.sp-cc.de/salc-openings.htm
Dann Corbit
Posts: 12538
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: Release: FEOBOS v20.1 available, sorry!

Post by Dann Corbit »

The Febos book has interior nodes (before the book exit points) that are very polar.
Some examples:

Code: Select all

rnbqkb1r/pp2pppp/3p1n2/8/3pP3/2N2N2/PPP2PPP/R1BQKB1R b KQkq - acd 32; acs 180; bm dxc3; ce 525; pm dxc3; pv dxc3 Bc4 cxb2 Bxb2 Qa5+ Qd2 Qxd2+ Nxd2 e5 Bb3 Be6 f3 Nbd7 O-O Rc8 Rfd1 b5 Nf1 a5 Ne3 g6 Bc1 Ke7 Rb1 Rc5 Nf1 a4 Bxe6 Kxe6 g4 Rxc2 g5 Ne8 Ne3 Rxa2 Rxb5 h6; 
rnbqk2r/ppp2ppp/4pn2/3p4/1bPP4/6P1/PP1BPP1P/RN1QKBNR w KQkq - acd 32; acs 180; bm Bxb4; ce 437; pm Bxb4; pv Bxb4 dxc4 Bg2 Na6 Ba3 Rb8 b3 cxb3 axb3 b6 b4 c5 bxc5 bxc5 Nf3 O-O O-O Bb7 Nbd2 Nd7 Qb3 Qb6 Qd3 Bxf3 Bxf3 Nb4 dxc5 Nxc5 Qc3 Na4 Qb3 Nc5 Qxb4 Qxb4; 
r1bqk1nr/pp1pppbp/2n3p1/8/2PNP3/2N5/PP3PPP/R1BQKB1R b KQkq - acd 31; acs 180; bm Bxd4; cce 68; ce 388; pm Nf6 {13}; pv Bxd4 Bd2; white_wins 3; black_wins 5; draws 5; 
r3kb1r/1bqn2pp/p3pn2/1pp5/4P3/1BN2N2/PP3PPP/R1BQ1RK1 w kq - acd 26; acs 15; bm Bxe6; ce 352; pm Bxe6; pv Bxe6 Bd6 Re1 h6 Qe2 O-O-O e5 Rhe8 exd6 Qxd6 Qe5 Qxe5 Rxe5 Kc7 Re2 g5 h4 Nf8 Bf5 b4 Rxe8 Nxe8 Na4 Bxf3 gxf3 Rd5 Bg4 Nf6 Be3 Kc6 Rc1 Nxg4;
r1b1k2r/pp1ppNbp/6p1/8/2P1P3/2N5/PP3PPP/R2nKB1R b KQkq - acd 26; acs 15; bm Nxb2; ce 345; pm Nxb2; pv Nxb2; 
r3kb1r/1bqn1ppp/p3pn2/1ppP4/4P3/1BN2N2/PP3PPP/R1BQ1RK1 w kq - acd 25; acs 15; bm dxe6; ce 330; pm dxe6; pv dxe6 fxe6; 
rnbqkb1r/ppp2ppp/5n2/4p1B1/3Pp3/2N5/PPP2PPP/R2QKBNR w KQkq - acd 25; acs 15; bm dxe5; ce 320; pm dxe5; pv dxe5; 
r3kb1r/1bqn1ppp/p3Pn2/1pp5/4P3/1BN2N2/PP3PPP/R1BQ1RK1 b kq - acd 27; acs 15; bm fxe6; ce -319; pm fxe6; pv fxe6 Bxe6 Bd6 Re1 b4 Nd5 Nxd5 exd5 O-O-O a3 Kb8 axb4 Rhe8 g3 Nf8 Bg4 Rxe1+ Qxe1 Ng6 Qe2 Qb6 Ng5 Be7 Bf3 Bxd5 Bxd5 Rxd5 Rxa6 Rd1+ Qxd1 Qxa6 bxc5 Qc4 Be3 Ne5 Qd4 Qxd4 Bxd4; 
rnb1kbnr/ppp1q1pp/3p4/4Np2/3PP3/8/PPP2PPP/RNBQKB1R w KQkq - acd 25; acs 15; bm Qh5+; ce 285; pm Qh5+; pv Qh5+; 
r5k1/pp2ppbp/n5p1/3q1b2/1P1P4/P3PNB1/2r2PPP/R2QKB1R b KQ - acd 28; acs 15; bm Qc6; ce 275; pm Qc6; pv Qc6;
rnbqk2r/pppp1ppp/5n2/4p3/1b2PP2/2N5/PPPP2PP/R1BQKBNR w KQkq - acd 38; acs 1171; bm fxe5; cce 323; ce 259; pm fxe5 {13} Bc4 {3} Nf3 {3}; pv fxe5; white_wins 9; black_wins 1; draws 1; 
rnbqk2r/pppp1ppp/5n2/4P3/1b2P3/2N5/PPPP2PP/R1BQKBNR b KQkq - acd 34; acs 180; bm Bxc3; cce -389; ce -257; pm Bxc3 {9} Ng8 {4}; pv Bxc3 dxc3; white_wins 7; black_wins 0; draws 1; 
rnbqk2r/pppp1ppp/5n2/4P3/4P3/2b5/PPPP2PP/R1BQKBNR w KQkq - acd 36; acs 180; bm dxc3; cce 333; ce 256; pm dxc3 {9}; pv dxc3; white_wins 3; black_wins 0; draws 1; 
r1b1kb1r/2qn1ppp/p3pn2/1ppP4/4P3/1BN2N2/PP3PPP/R1BQK2R w KQkq - acd 24; acs 15; bm dxe6; ce 254; pm dxe6; pv dxe6;
1rb1k1r1/1pqnbpp1/p2ppn1p/8/3NPP1B/2N3Q1/PPP1B1PP/2KR3R w - - acd 32; acs 180; bm Nxe6; cce 123; ce 253; pm Rhf1 {38} Rhe1 {5} Nf3 {4} Bxf6 {2} e5 {2} f5 {2} Bf3 {1}; pv Nxe6; white_wins 33; black_wins 18; draws 3; 
r1bqkbnr/pppp1ppp/8/n3p3/4P3/1BNP4/PPP2PPP/R1BQK1NR w KQkq - acd 36; acs 600; bm Bxf7+; ce 249; pm Bxf7+; pv Bxf7+ Kxf7 Qh5+ g6 Qxe5 Nf6 Qxa5 d6 Nd5 Nxd5 Qxd5+ Kg7 h4 h6 Be3 Be7 Ne2 Bf6 O-O-O a5 Nf4 a4 Qc4 a3 b3 Rf8 Kb1 Be5 Nd5 c6 Nb6 Ra6 Nxc8 Qxc8 h5 g5 d4 Bf4 Bxf4 Rxf4 f3; 
r4rk1/3bbp1p/pqnppp2/1p3P2/4P3/2N2N2/PPPQ2PP/1K1R1B1R w - - acd 34; acs 180; bm Qh6; cce 286; ce 240; pm Qh6 {22} Bd3 {4} Qf4 {4} Ne2 {1}; pv Qh6 Rfd8; white_wins 25; black_wins 5; draws 1; 
rnbqkb1r/1p2n1pp/p2p4/2pP1p2/P3PB2/2N5/1PP3PP/R2QKBNR w KQkq - acd 23; acs 15; bm Qe2; ce 237; pm Qe2; pv Qe2; 
rnbqkbnr/pp2pppp/2p5/8/2BPN3/8/PPP2PPP/R1BQK1NR w KQkq - acd 25; acs 15; bm Nf3; ce 237; pm Nf3; pv Nf3; 
rnbqkb1r/pp1p1ppp/5n2/2pp4/4P3/2N5/PPP2PPP/R1BQKBNR w KQkq - acd 31; acs 180; bm e5; cce 156; ce 234; pm exd5 {59} e5 {30} Nxd5 {2}; pv e5; white_wins 52; black_wins 21; draws 15; 
r1bqk2r/pppp1ppp/2n2n2/4p1N1/2BbP3/8/PPP2PPP/RNBQK2R w KQkq - acd 34; acs 180; bm Nxf7; cce -222; ce 231; pm c3 {3} Nxf7 {2}; pv Nxf7; white_wins 1; black_wins 3; draws 0; 
r1b1k2r/pp1np1bp/6pn/q1p1N3/2P2P2/2NB3P/PP4P1/R1BQ1RK1 w kq - acd 25; acs 15; bm Qe2; ce 225; pm Qe2; pv Qe2 Nxe5 fxe5 Be6 Bg5 Qb6 Bc2 O-O-O Bxe7 Nf7 Ba4 Rd4 b3 Rhd8 Bxd8 Qxd8 Nb5 Rd2 Qe3 Kb8 Rae1 Qg5 Qxg5 Nxg5 Rf2 Rxf2 Kxf2 Nf7; 
rnb1k2r/pp1p1ppp/4p3/q3P3/1b1Nn1Q1/2N5/PPP2PPP/R1B1KB1R w KQkq - acd 35; acs 180; bm Qxg7; cce 230; ce 224; pm Qxe4 {17} Qxg7 {17} Nb5 {5}; pv Qxg7 Rf8 Bh6 Nxc3 a3 Be7 Bd2 Qd5 Bxc3 Qe4+ Be2 a6 O-O-O Qg6 Qxg6 fxg6 h4 b6 h5 Bb7 hxg6 hxg6 f3 Bd5 Bd3 Rg8 Rh7 Nc6 Nxc6 dxc6 Bd2 b5 Rdh1 Bc4 Bb4 Bxb4 Bxc4 bxc4 axb4 a5 Rc7 axb4 Rhh7 Ra1+ Kd2; white_wins 19; black_wins 4; draws 6; 
rnb1k1nr/pp1ppp1p/6p1/2P5/4P3/2q2N2/P1PQ1PPP/R1B1KB1R b KQkq - acd 26; acs 15; bm Qxa1; cce 190; ce 219; pm Qxa1 {5} Qxc5 {3}; pv Qxa1 c3 Na6 Bc4 Nxc5 O-O Nxe4 Qc2 Qxc3 Qxe4 Nf6 Qe2 O-O Qxe7 d5 Ne5 Nd7 Nxf7 Qf6 Bb2 Qxb2 Bxd5 Qg7 Nd8+ Kh8 Nf7+ Rxf7 Bxf7 Qf8 Re1 Qxe7 Rxe7 Nf6 Bc4; white_wins 2; black_wins 5; draws 0; 
rnbqkb1r/pppp1ppp/5n2/4p3/3PP3/3B4/PPP2PPP/RNBQK1NR w KQkq - acd 31; acs 180; bm dxe5; cce 49; ce 217; pm c3 {9}; pv dxe5 Ng8 f4 Nc6 Nc3 d6 exd6 cxd6 Be3 g6 Qf3 Bg7 Nge2 Nf6 h3 O-O O-O-O a6 g4 h6 Kb1 Bd7 Ng3 Rb8 Be2 Qc7 a3 b5 g5 b4 gxf6 bxc3 fxg7 Rxb2+ Ka1 Rfb8; white_wins 4; black_wins 3; draws 2; 
rnb1kbnr/ppppq1pp/8/4Np2/4P3/8/PPPP1PPP/RNBQKB1R w KQkq - acd 37; acs 460; bm Qh5+; cce 207; ce 213; pm d4 {18} Qh5+ {13}; pv Qh5+ g6 Nxg6 Qxe4+ Be2 Nf6 Qh3 hxg6 Qxh8 Qxg2 Rf1 Kf7 d3 Nc6 Qh4 Bg7 Be3 Ng4 Nc3 d5 O-O-O Qxh2 Qxh2 Nxh2 Rh1 d4 Rxh2 dxe3 fxe3 Be6 Rdh1 Rd8 Bf3 g5 a3 g4 Bg2 Ne5 Ne2 c6 Rh5 Nd7 Nf4 Rg8 Nxe6 Kxe6 Rf1 a6 Rfxf5 Rh8 Rxh8 Bxh8 Rg5 Nf6 e4; white_wins 21; black_wins 7; draws 2; Opening Latvian Gambit: Greco Variation. 1.e4 e5 2.Nf3 f5 3.Nxe5 Qe7; CaxtonID: 1257; ECO: C40; 
r1bqkb1r/1p1n1ppp/p2ppn2/8/P1BNP3/2N5/1PP2PPP/R1BQK2R w KQkq - acd 16; acs 0; bm Bxe6; ce 202; pm Bxe6; pv Bxe6 Be7 O-O O-O Ba2 Ne5 h3 Nc6 Be3 Nxd4 Qxd4 Be6 Bb3 Nd7 Rfd1 Rc8 Nd5 Bxd5 Qxd5 Qc7 Kh1 Ne5 Qd4; white_wins 0; black_wins 0; draws 0; 
r1b1k1nr/1ppp1ppp/1b3q2/pP2p3/P1BnP3/2N2N2/2PP1PPP/R1BQK2R w KQkq - acd 28; acs 15; bm Nd5; ce 199; pm Nd5; pv Nd5;
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
User avatar
pohl4711
Posts: 2434
Joined: Sat Sep 03, 2011 7:25 am
Location: Berlin, Germany
Full name: Stefan Pohl

Re: Release: FEOBOS v20.1 available, sorry!

Post by pohl4711 »

Dann Corbit wrote:The Febos book has interior nodes (before the book exit points) that are very polar.
Thats not a problem. FEOBOS (SALC, too) are not for playing versus other books. They should be used by all engines in a torunament/test and if you do so, only the endpositions of the openinglines will be the startpositions for the engines thinking and playing. The moves and positions towards that endposition are not of interest.
Jeroen
Posts: 501
Joined: Wed Mar 08, 2006 9:49 pm

Re: Release: FEOBOS v20.1 available, sorry!

Post by Jeroen »

Two quotes from Frank in the discussion in the CSS forum:

"Aber mein Buch ist das Beste....."

"Durch das studieren der FEOBOS Excel wird Wissen erzeugt!!"

I took a quick look, but the FEOBOS books either plays:

a) a line that is well known and is the main line in many other books;
b) an exotic line that is not interesting from a human point of view;
c) a line with several dubious moves in it (I remember one that went 1.e4 c6 with black playing c6-c5?!? within the next moves and white d2-d3 and next d3-d4??!).

A lot of lines (f.e. in the King's Indian) that are very well playable, are excluded, as the shallow search of 10 engines will spit out an eval that is 'not acceptable' according to this method.

The best book there is? Hardly.....
User avatar
pohl4711
Posts: 2434
Joined: Sat Sep 03, 2011 7:25 am
Location: Berlin, Germany
Full name: Stefan Pohl

Re: Release: FEOBOS v20.1 available, sorry!

Post by pohl4711 »

Jeroen wrote:Two quotes from Frank in the discussion in the CSS forum:

"Aber mein Buch ist das Beste....."

"Durch das studieren der FEOBOS Excel wird Wissen erzeugt!!"

I took a quick look, but the FEOBOS books either plays:

a) a line that is well known and is the main line in many other books;
b) an exotic line that is not interesting from a human point of view;
c) a line with several dubious moves in it (I remember one that went 1.e4 c6 with black playing c6-c5?!? within the next moves and white d2-d3 and next d3-d4??!).

A lot of lines (f.e. in the King's Indian) that are very well playable, are excluded, as the shallow search of 10 engines will spit out an eval that is 'not acceptable' according to this method.

The best book there is? Hardly.....
Agreed. For point a) nobody needs FEOBOS, you can use any other “normal“ opening book. And for point b) and c) the question is, why anybody should play those lines out of FEOBOS, if there is no other advantage included (like in my SALC book, where you get a much lower draw-rate and a lot more thrilling games, when you accept on the other hand, that the played lines are point b) and c) - in FEOBOS you get nothing like that, as my testruns proved).
So, I can see definitly no reason for using FEOBOS for engine-tournaments or matches or ratinglists, because there is nothing to gain. Sad. But true.
User avatar
pohl4711
Posts: 2434
Joined: Sat Sep 03, 2011 7:25 am
Location: Berlin, Germany
Full name: Stefan Pohl

Re: Release: FEOBOS v20.1 available, sorry!

Post by pohl4711 »

Jeroen wrote:
A lot of lines (f.e. in the King's Indian) that are very well playable, are excluded, as the shallow search of 10 engines will spit out an eval that is 'not acceptable' according to this method.
.
I see another big problem: The spit out evals are not stable. Using another timecontrol and/or another hardware ans/or other engines, then the evals will fluctuate a lot. So, the situation is, that the strict eval-filters, FEOBOS uses, means, that in fact FEOBOS is filtering more less random eval-fluctuations. And you cannot gain knowledge out of that, of course.
Dann Corbit
Posts: 12538
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: Release: FEOBOS v20.1 available, sorry!

Post by Dann Corbit »

pohl4711 wrote:
Jeroen wrote:
A lot of lines (f.e. in the King's Indian) that are very well playable, are excluded, as the shallow search of 10 engines will spit out an eval that is 'not acceptable' according to this method.
.
I see another big problem: The spit out evals are not stable. Using another timecontrol and/or another hardware ans/or other engines, then the evals will fluctuate a lot. So, the situation is, that the strict eval-filters, FEOBOS uses, means, that in fact FEOBOS is filtering more less random eval-fluctuations. And you cannot gain knowledge out of that, of course.
It is still under development.
I guess it will improve over time.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
User avatar
pohl4711
Posts: 2434
Joined: Sat Sep 03, 2011 7:25 am
Location: Berlin, Germany
Full name: Stefan Pohl

Re: Release: FEOBOS v20.1 available, sorry!

Post by pohl4711 »

Dann Corbit wrote:
pohl4711 wrote:
Jeroen wrote:
A lot of lines (f.e. in the King's Indian) that are very well playable, are excluded, as the shallow search of 10 engines will spit out an eval that is 'not acceptable' according to this method.
.
I see another big problem: The spit out evals are not stable. Using another timecontrol and/or another hardware ans/or other engines, then the evals will fluctuate a lot. So, the situation is, that the strict eval-filters, FEOBOS uses, means, that in fact FEOBOS is filtering more less random eval-fluctuations. And you cannot gain knowledge out of that, of course.
It is still under development.
I guess it will improve over time.
I doubt that. The problems, me and J.Noomen pointed out, are coming out of the general concept, FEOBOS is made of.