Page 86 of 211

Re: SedatChess

Posted: Wed Nov 29, 2023 9:28 pm
by Sedat Canbaz
UPDATE

Note also that 10.000 strength games are played about 20 Elo difference...
And sure I hope to share all strength test results in the next coming days

Meanwhile,
Sun is the latest qualified entry plus as same version, which is constant player too,
And frankly, I'm little bit surprised for 10-15 min work to be superior strong..
Sure all credits are going to SCCT Book Authors, I'm just creator, no more no less...

For Latest Standings:
https://sites.google.com/site/computers ... t-bn-cs-22

Greetings

Re: SedatChess

Posted: Wed Nov 29, 2023 11:48 pm
by OrgZ
Thanks for the update Mr Sedat and i see the draw percentage thus far for loosecanon is close to that or lion and hercules.... may the best book win. hope my next update will be better because at the moment am closing some holes. :difus_19

Re: SedatChess

Posted: Thu Nov 30, 2023 1:35 pm
by Sedat Canbaz
OrgZ wrote: Wed Nov 29, 2023 11:48 pm Thanks for the update Mr Sedat and i see the draw percentage thus far for loosecanon is close to that or lion and hercules.... may the best book win. hope my next update will be better because at the moment am closing some holes. :difus_19
Hello Mr. Tanick,

Not at all...

Oh.. yes low draw ratio by LooseCanon
And its performance is awesome too...

About tuning...
A never ending process )
And as you said, let's best book wins

Greetings

Re: SedatChess

Posted: Thu Nov 30, 2023 1:58 pm
by Sedat Canbaz
Another Update

SCCT Book / NNUE CS XXII - Strength Test

Overall Performance: 21 Elo Difference

Code: Select all

1   SF-POLY 220723    +677/-75/=9248 53.01%  5301.0/10000
2   Constant players  +75/-677/=9248 46.99%  4699.0/10000ö

Draw Ratio (based on 10.000 games): 92% / Constant players = 10 Engines
Individual's Performance

Code: Select all

SF-POLY 220723 - BlueMarlin 15.3a   529.0 - 471.0   +68/-10/=922  52.90%
SF-POLY 220723 - CatroPOLY 060523   533.0 - 467.0   +70/-4/=926   53.30%
SF-POLY 220723 - DeepBlue 180422    532.5 - 467.5   +69/-4/=927   53.25%
SF-POLY 220723 - Raubfisch X48e     536.0 - 464.0   +76/-4/=920   53.60%
SF-POLY 220723 - SF-PB 130222       519.0 - 481.0   +52/-14/=934  51.90%
SF-POLY 220723 - Swordfish 14.8     532.0 - 468.0   +70/-6/=924   53.20%
SF-POLY 220723 - Honey v14.1.01     529.5 - 470.5   +71/-12/=917  52.95%
SF-POLY 220723 - FisherovMonk 1.2   536.0 - 464.0   +78/-6/=916   53.60%
SF-POLY 220723 - Fat Titz 2         528.5 - 471.5   +65/-8/=927   52.85%
SF-POLY 220723 - CFish 010422       525.5 - 474.5   +58/-7/=935   52.55%

Conditions:
2x Epyc 7B12, CuteChess, 1 Core, PonderOFF, 1m+0.6s, 64 Hash, Balsa, 4 MEN
Note also that I tested many more other engines, but shared only the suitable ones!
In short, the target was to pick, which are close to 20 Elo difference/less DRAWISH
--------------------------------------------------------------------------------------

Hash Strength Test:

Code: Select all

Rank Name     Elo    +    - games score oppo. draws 
   1 64 mb   3761    5    5  1002   51%  3757   89% 
   2 16 mb   3760    5    5  1002   50%  3758   88% 
   3 32 mb   3757    5    5  1002   50%  3759   89% 
   4 128 mb  3755    5    5  1002   49%  3759   90% 

The Overall Draw Ratio (based on 2004 games): 89%

Conditions:
2x Epyc 7B12, CuteChess, 1 Core, Ponder OFF, 1m+0.6s, Unique, 4 MEN
SF-Poly 220723 avx2 used (for each player is installed other engine copy)
As final conclusion (for Bullet): 64 MB hash size seems to be as best!
To be sure 100% about yours.., run your own hash strength testings...
Because it depends on your cond.: Time Control, MP, Hardware Speed etc.

GAMES:
https://mega.nz/file/y0BhCbiD#LuAhmKrcM ... I-8xmMTScg

Re: SedatChess

Posted: Thu Nov 30, 2023 3:19 pm
by cubail
CFish 010422 ınterestıng performans thanks test

Re: SedatChess

Posted: Thu Nov 30, 2023 4:23 pm
by Homayoun
Thanks for tests,Mr.Sedat .

Re: SedatChess

Posted: Thu Nov 30, 2023 5:30 pm
by OrgZ
Homayoun wrote: Thu Nov 30, 2023 4:23 pm Thanks for tests,Mr.Sedat .
+1

Re: SedatChess

Posted: Thu Nov 30, 2023 6:48 pm
by Sedat Canbaz
cubail wrote: Thu Nov 30, 2023 3:19 pm CFish 010422 ınterestıng performans thanks test
Not at all...and agreed Cfish has an interesting playing style

But on other hand,
This engine is named as CFish, but it is not as official Cfish version

And meanwhile my 6th strong sense says that,
The author of Deep Blue 180422 is the same person/s,
Who released /created CFish 010422 too...because
The default book, threads etc. settings are quite close...

Greetings

Re: SedatChess

Posted: Thu Nov 30, 2023 6:56 pm
by Anton101
Many thanks for all this info, Mr Sedat.

Best regards;

Re: SedatChess

Posted: Thu Nov 30, 2023 6:58 pm
by Sedat Canbaz
You are welcome Mr. Homayoun and Mr. Anton )