FEOBOS v8
---------
FEOBOS v8 based on the first 21.074 of 26.146 ranking positions from
by 8 engines analyzed FEOBOS basic database.
The following engines analyzed the complete 26.146 basic database with
60 sec. per position, 4.096Mb for hashtables, 6 cores / 10 threads (HT)
on Intel Core i7-5930K, 4.3Ghz hardware with Windows 10 Professional
operating system:
01. Houdini 5.01 Pro x64, 10-Threads
02. Shredder 13 x64, 10-Threads
03. Fire 5 POP x64, 10-Threads
04. Wasp 2.01 BMI2 x64, 10-Threads
05. Chiron 4 x64, 10-Threads
06. Booot 6.2 POP x64, 10-Threads
07. Texel 1.07a9 x64, 10-Threads
08. Andscacs 0.91 BMI2 x64, 10-Threads
--
09. ASMFishM 20170814 BMI2 x64, 10-Threads is still running
10. Komodo 11.22 x64, 10-Threads will be follow
More information to the FEOBOS project can be found under:
http://www.amateurschach.de/main/_new-opening-book.htm
- the complete analyses of engines
- Excel workbook by Klaus Wlotzka (Germany) with a lot of statistics.
The core idea of the project!
- the important basic databases / files
- readme and much more information
To build rankings of opening positions is a complete new idea by
Klaus Wlotzka and myself.
With a tool by Ferdinand Mosca (Philippines) we can sort a *.pgn file
with the FEOBOS ranking number!!
That's given the possibility to create any Test-Set constellation you like.
Best 500 for a test-set or best B01 positions for a test-set.
In download directoy are all important files included!
I hope self explaining!
Have fun with our FEOBOS project and your engine-engine match constellation,
tournaments, rating systems with a strong and by 8 engines analyzed database
of A00-E99 opening positions.
Frank Quisinsky
Trier, Germany
August 18th, 2017 (12:00)
FEOBOS v8
Tool by Ferdinand Mosca, Philippines
------------------------------------
Ferdinand Mosca programmed a very important tool for FEOBOS!
#1+#########!!##
THANKS FERDINAND
#1+#########!!##
Note:
Set rank into *.pgn & sort *.pgn to rank number is possible with the toolbox
by Norm Pollock in combination with Scid database system too.
"tagInsert" (NEW: Old one was renamed to "txtCreate") and "txtZeros"
"tagInsert" takes a file of values and puts them into a tag type chosen by the user.
"txtZeros" is the tool that put the zeros in front so that the pgn could be sorted by SCID.
Tools by Norm Pollock can be found under: http://40h.000webhostapp.com/
Scid database system can be found under: http://scid.sourceforge.net/
With the tool by Ferdinand Mosca the ranking number of analyzed FEOBOS end positions can be
set and sort directly in *.pgn file. It's no longer necessary to delete "bad" or "0.00 draws"
positions by hand in database systems like Scid or Chessbase.
With the following big additional deal:
We can create any Test-Set constellation directly with the *.pgn file very easy!
spf.exe (program)
run_spf.bat (batch file)
feobos_without-dublicates_26146-before-upd.1.pgn (FEOBOS basic database)
feobos_v8-26146_ranking.txt (urgently needed)
_unsorted_feobos_v8-26146-ranking.pgn (for an example)
_feobos_v8-26146_ranking.pgn (for an example)
The feobos_v8-26146_ranking.txt can be since version 92 produced with the Excel workbook
by Klaus Wlotzka.
- click under worksheet "Set by hand" on button ...
"Create text file for 26.146 positions (worksheet FEOBOS)!"
- on worksheet FEOBOS the required information added!
- now copy & paste all 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/
Ranking system criterias:
-------------------------
1st criterion: Bad moves
2nd criterion: Draw 0.00 moves
3rd criterion: Queens on board
4th criterion: Unequal first moves of main variant
5th criterion: High eval moves exp. Engines (for the moment not used).
6th criterion: Average of Score of analyzed engines (8 at the moment).
On the end in ranking are the bad positions. Before bad position the 0,00 / draw positions.
We cut the by 8 engines FEOBOS analyzed basic database after rank 21.084 for FEOBOS v8.
Can be see in a graphic *.png file, included in download directory.
Note:
With worksheed "eval out" and the button "Sort by ranking" can be sort the database
to the current ranking in Excel workbook. Of course, you can cut the database much earlier
as rank 21.084 for a highter contempt for fast 3-fold repetition.
Have fun with our FEOBOS project and your engine-engine match constellation, tournaments,
rating systems with a strong and by 8 engines analyzed database of A00-E99 opening positions.
Frank Quisinsky
Trier, Germany
August 18th, 2017 (12:00)
on Excel worksheet "ECO" you can found a new statistic.
To each of the ECO codes you can find the probality to 0,00 / draw evals.
TOP 5 are: E76, D69, D29, A69, A66
We are working very hard on implementation of ideas for our Excel workbook. In the last weeks we create upd.1 with 15.584 new lines. Will be checked after the still running 26.146 lines. Our Excel table have now a capacity for 41.614 analyses by 14 engines. Some other interesting stats added in the last weeks. Some new worksheets added in the last weeks.
Please have a look in the Dia-Show or download the Excel worksheet. You can learn a lot to ECO codes and engines we added.
with the FEOBOS ranking system in combination with the tool by Ferdinand we have infinite number of Test-Set constellations.
Opening positions sort by a ranking systems is sure a very new and interesting idea.
You can build with FEOBOS *.pgn a Test-Set to the best 500 positions, or to each ECO Code best positions in a few minutes. TOP engines will be playing without fast 3-fold repetitions.
= max. possible contempt!
Example:
If you are using the first 13.800 of 26.146 ranking positions 0/8 analyzed engines produced a 0,00 eval!!
If you are create opening books from the first 13.800 positions you can be sure that max. possible contempt for fast opening draws (3-fold) can be avoid.
Sure, we will have much more fun with eng-eng match, tournaments, rating systems constellations.
We have here a lot of fun with the project here and from version to version all we be stronger and stronger.
undo move 12 we can discuss about thousands of possibilitys. Life isn't long enough.
Let us discuss about the end position!?
Here I have at the moment 8 engine analyses by TOP engines (OK, I have 26.146 x8 engine analyzes by TOP engines). With the automatic calculation of 8 analyzed engine results we are speaking from FEOBOS TOP-Position Number 1 !!!
Why best of 26.146?
7/8 engines produced with the B63 end positions different best moves. And all 7 best moves aren't 0,00 or "Bad moves". The average of 8 evals is in near of 0,00 ... again but no engine gave itself 0,00!
That's are the most interesting opening positions for eng-eng testing. Because with one of such positions we can produced 3-4 moves later hundrets of interesting lines. Fantastic for eng-eng Elo calculations.
At the moment TOP-25 of FEOBOS positions are analyzed from a group of grandmasters. Only one comment I have at the moment ... 23 / 25 are great, two are very complicated and totaly unclear because to many possibilitys. The grandmaster wrote that he need around 1 months for a full analysis for one of the two positions.
Speaking from FEOBOS ranking system, works fine.
The new and in my opinion best way to create opening books, in most of cases 3 moves after ECO code formed.
Why best way?
Humans produced to many blunders ... means if humans do that. Computer chess people aren't strong enough for it. So engines should do that work for 99% blunder free books for engine-engine testing.
Our opening books are full of blunders. Grandmaster databases ar full of blunders, also correspondece databases. So I am search and found a way with FEOBOS for my future of engine testing.
The average of 8 evals is in near of 0,00 ... again but no engine gave itself 0,00! That's are the most interesting opening positions for eng-eng testing.
Well, I respectfully disagree. Positions with a lot of exchanges and a completely open position (like #5) are not very interesting.
Such balanced lines will lead to a lot of draws (and boring games), especially in long time control games. Not many chances for engines to do something wrong there.
Even in my latest King's gambit testset there were 71% draws between Komodo and Stockfish. 50 unbalanced positions, a blitz match and still 71% draws.
Perhaps someone can give your top 5 positions a try in a LTC match between Komodo, SF and/or Houdini on strong hardware.