FEOBOS v8 based on ranking system released ...

[Frank Quisinsky]

Hi Jeroen,

yes, for the moment unlogical first moves (you have written about it) ...
For the moment only ...

And with such moves the strongest and best FEOBOS ranking position was build.

That's computer chess today and the reason that computer chess programs should to the job, should build the opening books for us.

My personal opinion!
The example is great and I hope of such comments!

Very interesting opening line!
I think that the players here try earlier that the opponent is out of book. White do the same black do ... you made me crazy with 2. c5 and the reaction is ... I do the same with 3.d3 ... 4.d4. That's a psychological line, should be available in each good opening book.

End of the day the line is on place 1 of 26.146 in FEOBOS ranking system. I like that.

Best
Frank

[Frank Quisinsky]

Hi Jeroen,

for 0,00 lines we developed an own systems for FEOBOS. Can be see in Excel workbook with a lot of options.

Fast 0,00 draw lines isn't for FEOBOS a topic, sorry!
That's the highligh of the collection we build and it works fantastic with eng-eng matches.

To position 5:
Here you are right to 75-95%.

D69 is on position 2 of 500 ECO codes, means that D69 ECO code produced the second most drawish lines (all that can be found on FEOBOS excel workbook). Believe me it's really an order to found for such difficult ECO codes good lines. I am sure as opening expert you know it. Pos. 5 is great for the drawish D69 ECO code.

So you are right but D69 is one of 500 ECO codes and we have no reason to ignore codes generally. We try to find out to each of the codes the best lines. That's one of the orders for an encyclopaedia of balanced lines.

Best
Frank

[Frank Quisinsky]

Hi Daniel,

not easy to beat such a result by Andscacs.
Stockfish and Komodo will have problems to beat it ... think so!

But most spectacular is the 0,00 quote of Andscacs.
The others are 3 - 12% higher ...

Have a look on 0,00 stats.
http://www.amateurschach.de/common/feob ... 0exp-1.png

So we have a strong number 1 so far!
Andscacs is very strong for opening analyses, nice to have ... and of course for all other things around computer chess.



Best
Frank

[Frank Quisinsky]

Hi there,

the programming from our idea for a ranking system to opening position! Based on 2.5 millions formulas, solved with a macro.

Klaus wrote in CSS Forum a bit about it in german language:
http://forum.computerschach.de/cgi-bin/ ... l?tid=9529

The text included in grafic I try to translate in English:



Grafic added in Excel workbook v93.

Maybe readers here have interest to made it better. We are open for each help all the time FEOBOS is still running.

Best
Frank

[Frank Quisinsky]

Hi Jose,

in KECR we had a big bug but Klaus is working the complete night for a bugfix. Now all should be clean with the result ...

Andscacs is really on place 1!

If you have interest to look in detail, please download the KECR file again or please have a look again on the Dia-Show on ...

FEOBOS / KECR
http://www.amateurschach.de/main/_new-opening-book.htm

Sorry to you and the other programmers we added the engine. But the KECR Ratings based on 6.7 millions formulas and is very complicated.

Best
Frank

[cdani]

Nice! Many thanks. Your project is great

[Frank Quisinsky]

Hi there,

for FEOBOS two databases are still under development. A database with and without move transpositions. The database without move transpositions we can sort with the new tool by Ferdinand by ranking.

If all engines analyses are done we have interest to offer 4 opening books without move transpostions as final results.

1. Contempt 4
No of 10 engines gave 0,00 evals.

2. Contempt 3
Not more as 1 of 10 engines gave 0,00 evals.

3. Contempt 2
Not more as 2 of 10 engines gave 0,00 evals.

4. Contempt 1
Not more as 3 of 10 engines gave 0,00 evals.

The database with move transpositions is more complicated. For three weaks we changed engine settings and the complete database is now overworked. Furthermore, we fixed some bugs.

Can be found on FEOBOS site:
http://www.amateurschach.de/main/_new-opening-book.htm

And with the final FEOBOS result will be opening book 5 available (with move transpositions).

So, the ideas we have ...

I wrote a short readme:

Code: Select all

FEOBOS v8, databases with move transpositions
---------------------------------------------

In download directory can be found the FEOBOS v8 database with move transpositions.
Complicated because I have the reject positions by hand. I rejected the "bad / 0,00 draw
lines" under Chessbase 14 GUI.

Please be informed that a "clean database" can be produced to the end of the project
only.

Note:
Later the final FEOBOS result with move transpositions can be better used for comparing
opening books among themselves.

---

Hardware for 10 "Basic engines":
60 sec. per position, 4.096Mb for hashtables, 6 cores / 10 threads (HT),
Intel Core i7-5930K, 4.3Ghz, Windows 10 Prof.

Step 01, 26.146 positions v01: Houdini 5.01 Pro x64 10-Threads, done
Step 02, 26.146 positions v02: Shredder 13 x64 10-Threads, done
Step 03, 26.146 positions v03: Fire 5 POP x64 10-Threads, done
Step 04, 26.146 positions v04: Wasp 2.01 BMI2 x64 10-Threads, done
Step 05, 26.146 positions v05: Chiron 4 x64 10-Threads, done
Step 06, 26.146 positions v06: Booot 6.2 POP x64 10-Threads, done
Step 07, 26.146 positions v07: Texel 1.07a9 x64 10-Threads, done
Step 08, 26.146 positions v08: Andscacs 0.91 BMI2 x64 10-Threads, done

#######################################
 93.794 positions BASIC database
- 4.336 positions rejected undo Step 08
 89.458 current standings
#######################################

Step 09, 26.146 positions v09: ASMFishM 20170814 BMI2 x64 10-Threads
Step 10, 26.146 positions v10: Komodo 11.22 x64 10-Threads
Step 11, 26.146 positions      Draw positions if not more as 03 of 10 engines evaulate with 0,00
Step 12, 15.468 positions v11: Houdini 5.01 Pro x64 10-Threads
Step 13, 15.468 positions v12: Shredder 13 x64 10-Threads
Step 14, 15.468 positions v13: Fire 5 POP x64 10-Threads
Step 15, 15.468 positions v14: Wasp 2.01 BMI2 x64 10-Threads
Step 16, 15.468 positions v15: Chiron 4 x64 10-Threads
Step 17, 15.468 positions v16: Booot 6.2 POP x64 10-Threads
Step 18, 15.468 positions v17: Texel 1.07a9 x64 10-Threads
Step 19, 15.468 positions v18: Andscacs 0.91 BMI2 x64 10-Threads
Step 20, 15.468 positions v19: ASMFishM 20170814 BMI2 x64 10-Threads
Step 21, 15.468 positions v20: Komodo 11.22 x64 10-Threads
Step 22, 15.468 positions      Draw positions if not more as 03 of 10 engines evaluate with 0,00

Hardware for 3 "Experimental engines":
90 sec. per position, 2.048Mb for hashtables, 4 cores / 8 threads (HT),
Intel Core i7-4770K, 4.0Ghz, Windows 7 Prof. x64

We am looking here for bad moves maybe the 10 "Basic engines" not found.

Step 23, 41.614 positions    : Spark 1.0 x64 8-Threads
Step 24, 41.614 positions    : Junior 13.3.00 x64 8-Threads
Step 25, 41.614 positions    : Hakkapeliitta TCEC v2 x64 8-Threads



Frank Quisinsky
Trier, Germany
August 21st, 2017 (19:00)