OpenChess

Posted: **Tue Jan 28, 2025 1:01 pm**

Sedat Canbaz wrote: Mon Jan 27, 2025 10:56 pm Breaking News (28.01.2025):
Today I realized to test many more engines which are close to 3800+, where
CorChess is CorChess, managed successfully to define his Champion title!!
But I admit that SF-PRO2 was a serious Champ candidate! but unfortunately
Lost but not due to strength or needed more luck etc. simply lost due not so
Stable..e.g there are 9 games lost on time by SF-PRO2..sad, we are in 2025
But..actually I'm doing my best such as increasing move overhead to 200.. but
APOCALYPSE's engine comes with poor settings e.g no way to increase..and
Just if again SF-PRO eng series will lost on time.I'll will not test them anymore..
Because 'stability' is most important part! I assume for your side too!? Btw,
SF PRO author is informed (in past) about over these instability issues, but
It seems he is not capable to fix the mentioned time management problems...
While I'm mentioning about time management, mostly of chess engines are
Are stable..no many time loses but some (including latest SF dev. series) play
Too fast..there is unbalanced TC issue.. such as half of time control not used..
Strange, right ? I am not a programmer, but according to my experience, min
80%-90% overall time should be used (in game play, in long moves matches)
However, I'm grateful from all authors since all these Top engines are for free!
Besides, no anyone's work is perfect. including mine.., even machines make
Mistake..Ah pardon, we've perfect fools who love to criticize others but theirs
Work is so limited and it's not worth, it will be waste of time to talk more over..
Oh friends that's all for now..see you later and have fun with latest results !)

SCCT LVI's GRAND Champion CorChess - Congratulations to I. Ivec & SF team!

FOR FINAL STANDINGS / GAMES:
https://sites.google.com/site/computers ... t-nn-cs-56

Best,
Sedat

Thank you Sedat for your tournaments!

I'll see what adjustments can be made to the engine.
Keep an eye on the mail

Posted: **Tue Jan 28, 2025 10:36 pm**

Dear Chess Friends,

You are welcome and thanks for your interest )

Greetings

Posted: **Wed Jan 29, 2025 2:41 am**

Great tour Sedat, thank you so much!

Posted: **Wed Jan 29, 2025 11:05 am**

Posted: **Wed Jan 29, 2025 6:26 pm**

Not at all.. it is my pleasure..)

And how nice to see my good chess friends here..

Thank you too )

Posted: **Wed Jan 29, 2025 6:32 pm**

UPDATE

An Unofficial test via newer SF-PRO2 280125 vers.
Normally the CS is over.. but I ran this quick beta test,
Just to help to SF PRO 's author...because previous SF PRO
Series were not so stable ( on my tournament machine...)

And I have good news (in 2000 games per player):
SF-PRO2 280125 has no any game to be lost on time!
My congratulations to SF PRO's author as well !

Note also that,
Current SF-PRO2 280125 has already move overhead option,
So played via 200, and as we see, it seems all stable...

Regarding latest SF-PRO's performance (e.g as 8th rank),
Not sure about you, but I'm satisfied..you may know to be
In Top 10 ranks, not so many engines managed to do that..!

But interesting to point out also that,
Older SF-PRO2 250125 (despite on 9 time loses), scored better!
And I wonder much, how is this possible ?)

Btw, via latest update,
Contact 10 SE is 2nd rank.. great too !

Code: Select all

Rank Name               Elo    +    - games score oppo. draws 
   1 Corchess 190125   3819    3    3  3666   52%  3805   89% 
   2 Contact 10 SE     3818    3    4  3666   52%  3805   89% 
   3 SF-PRO2 250125    3817    5    5  1900   52%  3806   89% 
   4 Killfish 021224   3815    4    4  3400   51%  3806   90% 
   5 Kookaburra 3      3815    4    4  3666   52%  3805   88% 
   6 Stockfish 17      3814    5    5  1900   51%  3806   90% 
   7 SF-PB 221124      3813    4    4  3666   51%  3804   89% 
   8 SF-PRO2 280125    3812    5    5  2000   51%  3806   90% 
   9 Brainlearn 30     3811    4    4  3666   51%  3805   89% 
  10 SpecTral v2       3810    5    5  1900   51%  3806   90% 
  11 Solidor 180924    3810    5    5  1900   51%  3806   89% 
  12 Private 19Sf      3809    5    5  1900   50%  3806   90% 
  13 Sun 15            3809    5    5  2100   51%  3805   88% 
  14 Tactical Titan    3807    5    5  1900   50%  3806   88% 
  15 ShashChess 37     3807    5    5  1900   50%  3806   89% 
  16 Stockfish 190125  3807    5    5  2016   50%  3805   88% 
  17 Stockfish 260125  3806    5    5  1900   50%  3806   89% 
  18 Magic Knight      3806    5    5  1900   50%  3806   89% 
  19 C-Dragon v3       3806    5    5  1900   50%  3806   90% 
  20 SF-POLY 210924a   3806    5    5  1900   50%  3806   91% 
  21 Fidelio 17        3806    5    5  1900   50%  3806   90% 
  22 Artemis 7 TR      3805    5    5  1900   50%  3806   88% 
  23 Steel Fighter 3   3805    5    5  1900   50%  3806   89% 
  24 DarkSeid 5.3      3805    5    5  2216   50%  3805   88% 
  25 Polyfish 140124   3805    5    5  1900   50%  3806   88% 
  26 Incognito 8.2     3805    5    5  1900   50%  3806   90% 
  27 Artemis 6 TR      3804    5    5  2016   50%  3805   88% 
  28 Eman 9.90         3803    5    5  1900   50%  3806   87% 
  29 Deep Ripper       3803    5    5  1900   50%  3806   87% 
  30 Dark SisTer v2    3803    5    5  2216   50%  3805   89% 
  31 Troubadour 1.5    3803    5    5  2100   50%  3805   90% 
  32 SF-PB 280924 SC   3803    5    5  1900   50%  3806   89% 
  33 JigSaw v5.9       3802    5    5  1900   49%  3806   88% 
  34 Aoda 17           3801    5    5  1900   49%  3806   89% 
  35 DeepBlue 0524     3801    5    5  1900   49%  3806   89% 
  36 SF-Ivec 180125    3800    5    5  2100   49%  3805   89% 
  37 RapTora 3.0       3800    5    5  1900   49%  3806   88% 
  38 Beast M 010125    3800    5    5  2216   49%  3805   87% 
  39 Sailfish 4.0      3800    5    5  1900   49%  3806   88% 
  40 Raid 36 TR        3799    5    5  1900   49%  3806   90% 
  41 Marauders 4       3798    5    5  1900   49%  3806   89% 
  42 HypnoS++ 1.01     3796    5    5  2166   49%  3805   88% 
  43 SFNNv6 16.1       3795    5    5  1900   48%  3807   86% 
  44 SF-PRO 311224     3788    5    5  2016   47%  3806   85% 
  45 JudaS++ 1.01      3784    4    4  2216   47%  3807   85% 
  46 Cool Iris 13.97   3782    5    5  1600   46%  3805   85%

GAMES:
https://pixeldrain.com/u/hafeSgrn

Note: Same conditions are used/played as:
https://sites.google.com/site/computers ... t-nn-cs-56

Greetings

Posted: **Wed Jan 29, 2025 8:04 pm**

ZamChess wrote: Tue Jan 28, 2025 8:14 am Thank you very much Sedat !

+1

Posted: **Wed Jan 29, 2025 11:13 pm**

I entered the tour and games to Deepseek and this is what it found, however what surprised me is that it does not know that BrainLearn and SpecTral handle Q-Learning, outstanding!

Key Findings
Extreme competitiveness:

Only 37 Elo points separate 1st (Corchess 190125: 3819) and 46th (Cool Iris 13.97: 3782).

Minimal differences (1-2 points) between leaders reflect technical parity.

Predominant tables:

>85% draws (up to 91% in SF-POLY 210924a), indicating conservative play and risk avoidance.

Stockfish domain:

7 of the top 20 are Stockfish variants (ex: SF-PRO2, SF-PB).

Minor updates (e.g. dev 16 and dev 17 in Stockfish) impact ±1 Elo point.

Reliability by game volume:

Engines with >3000 games have reduced error margins (±3-4 vs. ±5 in <2000 games).

Featured Patterns
Draws not correlated with Elo: Leading engines do not always have the highest rates (ex: Corchess 190125: 89% vs. SF-POLY 210924a: 91%).

Competent alternative engines: Brainlearn 30 (9th) and SpecTral v2 (10th) stand out with innovative approaches.

Conclusions
Microscopic innovation: Differences of 1-2 points require hyper-precise adjustments in algorithms.

Consistency > Risk: Maximizing positional advantages (pawn structure, center control) is key in the face of high draw rates.

Niche Specialization: Engines like Killfish 021224 (4th) can lead in specific areas (ending, tactics).

Best regards, Sedat!

Posted: **Thu Jan 30, 2025 9:31 am**

Anton101 wrote: Wed Jan 29, 2025 11:13 pm I entered the tour and games to Deepseek and this is what it found, however what surprised me is that it does not know that BrainLearn and SpecTral handle Q-Learning, outstanding!

Key Findings
Extreme competitiveness:

Only 37 Elo points separate 1st (Corchess 190125: 3819) and 46th (Cool Iris 13.97: 3782).

Minimal differences (1-2 points) between leaders reflect technical parity.

Predominant tables:

>85% draws (up to 91% in SF-POLY 210924a), indicating conservative play and risk avoidance.

Stockfish domain:

7 of the top 20 are Stockfish variants (ex: SF-PRO2, SF-PB).

Minor updates (e.g. dev 16 and dev 17 in Stockfish) impact ±1 Elo point.

Reliability by game volume:

Engines with >3000 games have reduced error margins (±3-4 vs. ±5 in <2000 games).

Featured Patterns
Draws not correlated with Elo: Leading engines do not always have the highest rates (ex: Corchess 190125: 89% vs. SF-POLY 210924a: 91%).

Competent alternative engines: Brainlearn 30 (9th) and SpecTral v2 (10th) stand out with innovative approaches.

Conclusions
Microscopic innovation: Differences of 1-2 points require hyper-precise adjustments in algorithms.

Consistency > Risk: Maximizing positional advantages (pawn structure, center control) is key in the face of high draw rates.

Niche Specialization: Engines like Killfish 021224 (4th) can lead in specific areas (ending, tactics).

Best regards, Sedat!

Thanks again Mr. Anton,

And as usual, always interesting to read your comments!

Have a great day )

Greetings

Posted: **Thu Jan 30, 2025 9:55 am**

UPDATE 2

Some notes and questions more,
1.Why SF-PRO2 250125 is performed better than other SF-PRO series?
2.Balsa + Unique openings are played any influence about results?
3.Any news about newer Balsa or Unique suite will be ever released?
4.Running 1000 games (per player) via UHO openings are enough..?

Well, let Me start to answer about 1st question:
Why SF-PRO2 250125 is performed better than other SF-PRO series?
I'm not a programmer, but as far as I noticed (in LIVE games too) :
SF-PRO2 250125 spent much more time till mid-game than endgame!
E.g more than 90%-95% of its time is used mainly up to 30-40 moves..
And later, SF-PRO2 250125 is started to play too much quickly, reason
To avoid to not lose on time..yes, in most of its games is managed to
Not lose on time, but just in some (9 games) are ended as time loses!
So I think that as golden key is played here about why SF-PRO2 250125
Performed better..in same time means (at least via this engine ver.):
Spending more time till mid-game is more important than endgame!
But where, the latest tested SF-PRO2 280125 version is used to play
Via more balanced time (per games) due move overhead 200 helped..

About the free SF-PRO 311224 ver, which is ranked as 44 rank:
No comments, exception rather I'd say the available SF-PRO is
Quite new engine too..just it's released about one month ago..
Btw, it seems SF-PRO 311224 is stable, no time loses games!
But why performed not better? here SF PRO's author may know...

About 2nd question,
2.Balsa + Unique openings are played any influence about results?
Of course...here Quality, Strength, Time... are said the last words!
As we noticed, even 1000 games (per player): CorChess was as BEST!
And after 2000 games, again CorChess, after 3000 games: CorChess...
Btw, if you going to ask why via UHO openings, CorChess was not..?
Then I suggest to read 4th section/question...please see below )

On other hand,
I'm not going to prove about my openings knowledge is good or bad..
But in the past, even actually nowadays, my favorite openings are
As one of best... sure I refer for my played tours categories! Btw,
Do you want to prove again via Online, in 2025 too? SORRY, after
25+ years, I'll not do that again and again...but probably many online
Players (in the past) know that my opening books were not so so weak!

About 3rd question,
3.Any news about newer Balsa or Unique suite will be ever released?
Yes..will be a new release.. for example, I am still working over
Newer Unique suite version which is highly based on Balsa openings,
But as main difference will be optimized for 3800+ players as well!
Where it's expecting to see a lot of spectacular games and in same
Time I am working regarding which produce less draw perc. values!
E.g the target is to be about 80%-90% draw ratio in Bullet + in Blitz!
Sure in slower TC, situation is unpredictable..since serious test required
But I guess, the draw numbers will be higher, reason: more time means
More assistance.. more and BIG assşistance (such as via NNUE) means
No much ways to appear blunder moves..so almost all games are ending
Draw! So here we can not blame Top engines or openings..hope helps!

No doubt too that just in case of..the fools, the trols will blame
Again that my openings are Drawish! they will say also that I am a
Cheater, a Liar.. etc. and again and again these 'perfect fools' will forget
To mention that my opening are one of best! As one of most funny thing:
They always forget to say that theirs games (via weak openings) are bad
Quality etc. if still not so clear ?
At least in SCCT Book Tours: all cleared, which is good, which is bad!!
But I'm afraid to say also these perfect fools will never learn about it!
Btw, not always, but sometimes I respect them too! do you know why ?)
Because without the clever fool guys, our hobby would not be boring ?))

About 4th question,
4.Running 1000 games (per player) via UHO openings are enough..?
Definitely no...here it seems at least 5.000 games are required for
Better stats..as reason, here almost all weak, critical openings..maybe
Later I can do more tests via UHO series, but I don't want to promise..
Just in case..who knows ? maybe after 5.000 games: CorChess will take
The lead..but via Balsa or Unique suite series: 1000 games was enough!
Right ? Or maybe I am missing something..?

And as last words, for anyone missed:
Learn also that CorChess is managed many times to be Champion:
https://sites.google.com/site/computers ... eng-champs

After all,
what does it mean? CorChess is so good in fast TC as well, right ?

Ok..that's all for now.) hope helps..and thanks for your reading...

Best,
Sedat

OpenChess

SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess