Modular opening book SF analysed 87417 pos., beta-1

[Ferdy]

Hi Ferdinand,

must drive to work.
I can check that after my work.

But for the moment I made a quick test.

I have here a file with 76 epd houdini found after the first 2923 analysis. 7 of the postions are higher as -0.40. Your program found 2 of the 7 ... to two highest with -0.48. -0.41 - -0.47 not found. In creteria I have ... 0.65 and -0.40.

So it must be an error in the program.
I will send you in the late evening the files for checking ... unfortunately ... no time for the moment.

With other words:
All what is higher as 0.60 is right.
All what is higher as - 0.40 is wrong.

Later ... in the evening ... with examples!

Best
Frank

The other thins in quick test are again genial!

If the window is [-40, +65] then it is fine that -48, -41 and -47 would not be included because
-48 < -40, -41 < -40 and -47 < -40.
The values -48, -41 and -47 are outside the window [-40, +64].

[Frank Quisinsky]

Hi Ferdinand,

shortly, what I have ...

Please download (1.1Mb):
http://www.amateurschach.de/download/fe ... ta-1.1.zip

Step 1: epd of 2.923 analyzed Houdini positions.
Step 2: same epd (Step 1) sorted with epdOrder by Norm
Step 3: reject 76 lines (higher as 0.65 / -0.40, 8 of the lines higher as -0.40) ... epd of 76 found lines only.
Step 4: Output of your tool (not in criteria, found 70).
Step 5: same epd (step 4) sorted with epdOrder by Norm
Step 6: not in criteria ... 6 of the 8 higher as -0.40 missed, included only 2x -0.48 not the others with -0.41 - - 0.44.

So the result from "not in criteria for higher as 0.60 is right.
The result from "not in criteria" for higher as -0.40 is wrong.

Hope you can follow the mistake!
Maybe my example isn't good?

Best
Frank

[Frank Quisinsky]

Code: Select all

start
refepd, 1-2923-test-file.epd
refpgn, beta-v3.pgn
minscorecp, -45
maxscorecp, +65
mincntqueen, 0
maxcntqueen, 2
end

Ah, ... I added -0.45!
Mistake by myself ...

Forget it!
Typical Frank ...

BIG SORRY !!!!!!!!!!!!!!!!!!!!

Let me test it with more time in the evening.
How embarrassing!

[Frank Quisinsky]

If you are now not to be angry with me ...
Sorry for my mistake again ... really embrassing ...

Engines have different evals, good to see if you are looking in the download file with houdini analysis.

It's possible to do this one ...
Your program (seems to be great now ... will test it with more time in the evening) ...

Example:
Houdini analysed 2.923 positions ...
It would be nice to have the average of ce output from all analysed positions. In this case 2.923.

End of my test ... we have the average from Houdini, Stockfish, Komodo and the others from over 26.000 positions. This one will be very very important for ...

Which setting I have to give ...
Or to understand engines better in eval for the beginning of the games.

Or ...
If I set for Komodo and Stockfish 0.50 / -0.30 I know that for Houdini I have to set a higher setting (all without contempt as UCI parameter). But later, all is ready I know it exactly if I have the information with "Average of ce". So maybe your programm can give me a new file ... maybe called, average of ce ... and here the information can be saved.

Best
Frank

Puh, with your tool it's fantastic to reject and to have it in a new database.
HURRA, must not do it by hand in Chessbase database program.

I will added it on my site and download file to the project in the late evening if you give your permission!

[Ferdy]

Which setting I have to give ...
Or to understand engines better in eval for the beginning of the games.

Or ...
If I set for Komodo and Stockfish 0.50 / -0.30 I know that for Houdini I have to set a higher setting (all without contempt as UCI parameter). But later, all is ready I know it exactly if I have the information with "Average of ce". So maybe your programm can give me a new file ... maybe called, average of ce ... and here the information can be saved.

Try the scoring percentage approach as in the following to compare score of engines. A sample comparison between Komodo and iCE. Say you want to see the scores of engines at 45% and 70% scoring percentage of each engine. At 45% iCE has a score of -0.50, for komodo it is -0.80. At 70% iCE has a score of +0.78 while for komodo it is -0.45. Komodo being one of the top engine in your playground, does not need a big score to get a high scoring performance. The data is from positions in the games in your database, from moves 13 to 24 of that particular player.



It is up to you what scoring percentage you want, then proceed extracting games based on score with reference to the scoring percentage for every particular player.

I will plot later the graphs for other engines.

[Frank Quisinsky]

Hi Ferdinand,

this stats are really interesting.
I will try to do such things later in Excel



Game-Extractor 1.1 is now on my website!

- added in project files, beta-v03.1
- added in my download area

I changed the two big analyzed *.epd files I have (Komodo / Stockfish) for project files beta-v3.1. I added here the Ae "Name of engine information".

For the moment no more changes.
Next update if Houdini is ready (v01).

Best
Frank

Your version 1.1 works fine for me, but I will do much more experiments in the next hours and tomorrow.

[Frank Quisinsky]

Hi Ferdinand,

I made now some experients more.
Your program works really fine.

Your tip for Notepad++ is great too.
If Houdini is ready I will add the name of engine with "Ae" in the *.epd file with the analysis.

I added the *.csv file (comes from Shredder GUI, Shredder saved the analysis in *.csv and *.epd format) to LibreOffice 5.3.0.3 and MS Excel 2016.

If you like please download (1.2Mb):
http://www.amateurschach.de/download/00 ... alysis.zip

- *.csv file
- *.epd file
- *.ods file
- *.xlsx file (Excel 2016)

After the first 5.181 of 26.146 analysis:

Average of 5.181 entries!!

Average Ply = 24.31
Average eval = 0.23
Average nps = 17.284.504
Average nodes = 1.037.158.239

Really nice to see that too, means to compare that in an half year if I have the complete analysis from 10 engines.

Best
Frank

We can do so many things with the analysis ... have x no sorted ideas with the material yet. But for all I need at first your new game-extractor. I am so happy to have it. Have many many thanks !!!!!!!!!!!!!!!!!!!!!!!!!!!!

[Ferdy]

Plot with Stockfish, note for Komodo and Stockfish the points in the plot with negative score has less data because these engines have lower number of losses. It would be fine if you extract games based from the score of these engines say -50/+10 for komodo and -50/+10 for stockfish since they have almost the same scoring percentage at a certain score range.

[Frank Quisinsky]

Hi Ferdinand,

your game-Extractor filtered without move transpositions and setting from criteria the *.pgn.

- filtered without move transpositions and criteria, Output *.pgn and *.epd
- unfiltered without move transpositions and criteria, Output *.pgn and *.epd

= 4 files ...

Missed is filtered "with" move transposition ... here *.pgn only is important. Criteria only is important.

So, in realitiy I need 6 output files (the two *.epd you have and two other *.pgn missed).

Give me the *.pgn with move transpositions (important for books) ... missed
Give me the *.pgn with move transpositions (reject lines) ... missed
Give me the *.pgn without move transpostions (for analyzes, test-set)
Give me the *.pgn without move transpositions (reject lines)
and the two *.epd files you have.

Best
Frank

[Ferdy]

Give me the *.pgn with move transpositions (important for books) ... missed

When the ref pgn defined by the user in criteria.txt has no doubles, then this tool cannot give the pgn with transposition. This is the issue that I am trying to decide of whether to add the output,
"in_criteria_with_doubles_<refpgn>"

Perhaps I will just output this file but will not check if the file contains a game or not.

Give me the *.pgn with move transpositions (reject lines) ... missed

Currently the output unfiltered_<refpgn> or rejected can have doubles. I did not remove doubles in this file if ever there is double.



This is my plan for v1.2.

Code: Select all

Create the following output files.
A. for pgn files
1. in_criteria_unique_<ref_pgn_name>.pgn
2. in_criteria_not_unique_<ref_pgn_name>.pgn
3. not_in_criteria_unique_<ref_pgn_name>.pgn
4. not_in_criteria_not_unique_<ref_pgn_name>.pgn

B. For epd files
1. in_criteria_<ref_epd_name>.epd
2. not_in_criteria_<ref_epd_name>.epd