Chess Forum

Posted: **Sun Feb 25, 2024 2:53 pm**

Breaking News - 25.02.2024:
The current competition is ended...many thanks and congratulations to all!
No doubt that most productive Book Championship ever! exc that I created,
Tested and shared all them, as additional, I'd like to say as good news many
Players gained same, identical points, very interesting, never seen before!
For examples, 2nd rank belongs to two book players, 3rd rank again two..!
As we see here, only OPTIMUS is clear a Winner of Eng Book Standings!
About Human standings, here again we see identical in points, e.g 3 (three)
Players are managed to be Champions: Ding, Dominguez, Karjakin, nice!
Btw, without to not mention this I can not as well, sometimes..X Author may
Not be so active in book tuning...and that sounds perfectly fine to Me...but
Just in case to be not Number One.. it does not mean that he/she can not
Create Champions books..I think, it all depends on our free time, health etc.
Sure I refer about current books by A. Morano plus for H. Can, C. Sarı etc.
But I admit that R. Constantinesco's books are at high level of quality too!
Sure I respect anyone's view, but I suggest to check Champions Rankings
In short, it can be used as very good indicator about who is best of the best..
Be aware of that soon I hope to organize another new book championship,
Where I plan to test Private books as well..right now I'm not sure exactly, but
I may use one of the strongest NNUE engines, rather it will be more close to
Online Game Play...at least it will be under: BULLET, MP, PONDER ON etc.!
And what will be as difference than Online, sure it will be with less players....
But what is as advantage, I mean much better than Online:
All planning books will be on same fast hardware (2x EPYC 128 Cores/256T),
As you know,..many online players suffer / gain due to hardware speed diff.
Besides here... all books will be played as NON - Cherry Picking System !
You know, nowadays...it a TREND )) full of these kind of postings out there )
But via SCCT. I think will be as solution, in short, BEST ones will be ranked
At higher ranks...right? Ok..chess friends that's all for now ...see you soon )

Eng Book Champion: OPTIMUS - Congratulations to Roger Constantinesco!
Human Book Champions: Ding, Dominguez and Karjakin - My Congratulations!

For Final Standings:
https://sites.google.com/site/computers ... book-cs-68

GAMES:
https://mega.nz/file/XspD3TqT#IU-V-Mui6 ... 8g_AcpRDo4

Greetings )

Posted: **Mon Feb 26, 2024 7:46 am**

Hello there )

Just another series of HASH strength testings,
Running in the name of planning Online Like CS!

What is New / Conditions:
2x Epyc 7B12, CuteChess 1.3.1 GUI, 4-MEN Syzygy
All Hash matches are playing at Bullet TC (1m+0.1s)
MP (12 Core/Game), Ponder ON, 5 Concurrent games
A new Unique suite used, played each opening twice
For all Hash players: EMAN 9.90 (ExternalNet) used
For each player, a copy of Eman engine installed
Dual NNUE (Big + Small Evalfiles) are installed too
Learning is disabled (Experience read only enabled)

And here are the 1st played results:

Code: Select all

Rank Name                          Elo     +/-   Games   Score    Draw 
   1 Eman 4096mb                    26      27      66   53.8%   89.4% 
   2 Eman 512mb                      0      14      68   50.0%   97.1% 
   3 Eman 2048mb                    -5      17      68   49.3%   95.6% 
   4 Eman 128mb                     -5      22      68   49.3%   92.6% 
   5 Eman 1024mb                   -16      23      66   47.7%   92.4% 

168 of 1000 games finished..

And I hoıpe the current new hash testing to be useful..

Best,
Sedat

Posted: **Mon Feb 26, 2024 12:57 pm**

UPDATE

After 360 games (per player):
Eman 9.90 MP via 2048 MB Hash is performed as BEST so far!
Where at 2nd place are ranked two players: 4096mb and 512mb

Code: Select all

1 2048mb  44.5 - 45.545.5 - 44.546.5 - 43.547.0 - 43.0  183.5/360
2 4096mb  45.5 - 44.544.5 - 45.545.5 - 44.546.0 - 44.0  181.5/360  -170.75
3 512mb   44.5 - 45.545.5 - 44.545.5 - 44.546.0 - 44.0  181.5/360  -172.75
4 128mb   43.5 - 46.544.5 - 45.544.5 - 45.547.0 - 43.0  179.5/360
5 1024mb  43.0 - 47.044.0 - 46.044.0 - 46.043.0 - 47.0  174.0/360

More Details:
Overall Draw Ratio (Based on 900 games) is high: 93%
Any Player's Move Overhead (ms) is increased to: 200
-----------------------------------------------------

Be aware of that too,
EMAN HASH test is still active, but this time via Balsa,
Because 360 games often may lead to wrong conclusions...
You know, more better...because Unique suite is strong, but
Some Unique's lines can be said as critical (not strongest)
So for less Error Margin, more games will be played out...
In other words, Via Balsa openings is expecting less Error..
But in same time.. it is expecting more draws as well...)

And please stay tuned..very soon more updates coming

Greetings )

Posted: **Mon Feb 26, 2024 4:08 pm**

UPDATE 2

Eman 9.90 MP's HashTable test is completed..!
Overall's Winner is 2048 MB Hash

And once more we noticed that
Giving too much size, it does not mean better, stronger..
Otherwise, 4096 MB Hash should gain more points, right?

But as we see in latest test, via strong openings (e.g via Balsa),
Even small Hash sizes such as 128 MB, 512 MB is still not dead !)

Overall (Balsa + Unique Openings) Ponder ON, 12 Core/Game: 96% DRAWS

Code: Select all

1 2048mb  90.0 - 90.091.0 - 89.090.5 - 89.592.5 - 87.5**     364.0/720
2 512mb   90.0 - 90.090.5 - 89.591.5 - 88.591.0 - 89.0 **    363.0/720
3 128mb   89.0 - 91.089.5 - 90.589.5 - 90.592.0 - 88.0  **   360.0/720
4 4096mb  89.5 - 90.588.5 - 91.590.5 - 89.590.5 - 89.5   **  359.0/720
5 1024mb  87.5 - 92.589.0 - 91.088.0 - 92.089.5 - 90.5    ** 354.0/720

----------------------------------------------------------------------

Balsa Openings - Ponder ON, 12 Core/Game: 99% DRAWS

Code: Select all

1 512mb   45.0 - 45.045.5 - 44.545.0 - 45.046.0 - 44.0**     181.5/360
2 128mb   45.0 - 45.045.5 - 44.545.0 - 45.045.0 - 45.0 **    180.5/360 
3 2048mb  44.5 - 45.544.5 - 45.545.5 - 44.546.0 - 44.0  **   180.5/360  
4 1024mb  45.0 - 45.045.0 - 45.044.5 - 45.545.5 - 44.5   **  180.0/360
5 4096mb  44.0 - 46.045.0 - 45.044.0 - 46.044.5 - 45.5    ** 177.5/360

-----------------------------------------------------------------------

Unique Openings - Ponder ON, 12 Core/Game: 93% DRAWS

Code: Select all

1 2048mb  44.5 - 45.545.5 - 44.546.5 - 43.547.0 - 43.0**     183.5/360
2 4096mb  45.5 - 44.544.5 - 45.545.5 - 44.546.0 - 44.0 **    181.5/360 
3 512mb   44.5 - 45.545.5 - 44.545.5 - 44.546.0 - 44.0  **   181.5/360  
4 128mb   43.5 - 46.544.5 - 45.544.5 - 45.547.0 - 43.0   **  179.5/360
5 1024mb  43.0 - 47.044.0 - 46.044.0 - 46.043.0 - 47.0    ** 174.0/360

Btw, this is not all..Hash test series are still active...for example
As next, all same..exception the matches will be via Ponder OFF
And let's see what will the influences...
And as usual, all the played games will be after the end of testings...

Posted: **Mon Feb 26, 2024 4:12 pm**

UPDATE 2

Eman 9.90 MP+Ponder ON's Hash test is completed..!
The Overall's Winner is 2048 MB Hash

And once more we noticed that
Giving too much size, it does not mean better, stronger..
Otherwise, 4096 MB Hash should gain more points, right?

But as we see in latest test, via strong openings (e.g via Balsa),
Even small Hash sizes such as 128 MB, 512 MB is still not dead !)

Overall (Balsa + Unique Openings) Ponder ON, 12 Core/Game: 96% DRAWS

Code: Select all

1 2048mb  90.0 - 90.091.0 - 89.090.5 - 89.592.5 - 87.5**     364.0/720
2 512mb   90.0 - 90.090.5 - 89.591.5 - 88.591.0 - 89.0 **    363.0/720
3 128mb   89.0 - 91.089.5 - 90.589.5 - 90.592.0 - 88.0  **   360.0/720
4 4096mb  89.5 - 90.588.5 - 91.590.5 - 89.590.5 - 89.5   **  359.0/720
5 1024mb  87.5 - 92.589.0 - 91.088.0 - 92.089.5 - 90.5    ** 354.0/720

----------------------------------------------------------------------

Balsa Openings - Ponder ON, 12 Core/Game: 99% DRAWS

Code: Select all

1 512mb   45.0 - 45.045.5 - 44.545.0 - 45.046.0 - 44.0**     181.5/360
2 128mb   45.0 - 45.045.5 - 44.545.0 - 45.045.0 - 45.0 **    180.5/360 
3 2048mb  44.5 - 45.544.5 - 45.545.5 - 44.546.0 - 44.0  **   180.5/360  
4 1024mb  45.0 - 45.045.0 - 45.044.5 - 45.545.5 - 44.5   **  180.0/360
5 4096mb  44.0 - 46.045.0 - 45.044.0 - 46.044.5 - 45.5    ** 177.5/360

-----------------------------------------------------------------------

Unique Openings - Ponder ON, 12 Core/Game: 93% DRAWS

Code: Select all

1 2048mb  44.5 - 45.545.5 - 44.546.5 - 43.547.0 - 43.0**     183.5/360
2 4096mb  45.5 - 44.544.5 - 45.545.5 - 44.546.0 - 44.0 **    181.5/360 
3 512mb   44.5 - 45.545.5 - 44.545.5 - 44.546.0 - 44.0  **   181.5/360  
4 128mb   43.5 - 46.544.5 - 45.544.5 - 45.547.0 - 43.0   **  179.5/360
5 1024mb  43.0 - 47.044.0 - 46.044.0 - 46.043.0 - 47.0    ** 174.0/360

Btw, this is not all..Hash test series are still active...for example
As next, all same..exception the matches will be via Ponder OFF
And let's see what will the influences...
And as usual, all the played games will be after the end of testings...

Posted: **Mon Feb 26, 2024 11:51 pm**

UPDATE 3

Eman's Hash test via Ponder OFF is ended too,
Where this time, the Winner is 1024mb Hash

Code: Select all

1 1024mb  112.5 - 113.5115.5 - 110.5112.0 - 114.0114.5 - 111.5**     454.5/904
2 128mb   113.5 - 112.5113.0 - 113.0111.5 - 114.5115.0 - 111.0 **    453.0/904 
3 512mb   110.5 - 115.5113.0 - 113.0115.0 - 111.0114.5 - 111.5  **   453.0/904  
4 4096mb  114.0 - 112.0114.5 - 111.5111.0 - 115.0111.5 - 114.5   **  451.0/904
5 2048mb  111.5 - 114.5111.0 - 115.0111.5 - 114.5114.5 - 111.5    ** 448.5/904

But in Overall Standings (Ponder OFF + Ponder ON)
Then 512 MB Hash size's performance is best...!

Code: Select all

Rank Name     Elo    +    - games score oppo. draws 
   1 512mb   3801    3    3  1624   50%  3799   96% 
   2 128mb   3800    3    3  1624   50%  3800   96% 
   3 2048mb  3800    3    3  1624   50%  3800   96% 
   4 4096mb  3799    3    3  1624   50%  3800   96% 
   5 1024mb  3798    3    3  1624   50%  3800   95%

GAMES:
https://mega.nz/file/z8p3mSgK#bcgOsGHew ... U4g4Nd_l7U

Final words and as I mentioned before,
To be sure, run your own HASH strength testings!
In other words, it all depends on our conditions..
As you see, in my tests: almost same in points..
In same time that shows the stability of test cond.!

IMPORTANT NOTE:
If using weak (unbalanced, handicapped..) openings (in your test..),
Then suggested strongly over than 4000 - 5000 games (per player),
Otherwise... just in case of test with small number of games, then
It's expecting to see not so accurate results...I mean especially
If testing Top chess engines, which are close in strength...

But using strong openings, then Elo error margin is much less,
E.g about 1000 games is enough data..to see the real strength!

Hope helps...

Greetings )

Posted: **Tue Feb 27, 2024 12:44 pm**

Breaking News (27.02.2024):
Another new NN CS, where this time almost all are quite close in points!
What does it mean? 1st of all, latest NNUE engines seems to very close,
Almost identical in strength, sure my used conditions played another BIG
Factor too..you know, hardware stability, used openings etc..play really as
Important role over X engine's strength, at least via this tour, all confirms..
Be aware that after checking more closely the results, we will notice that
2 (two) of engines are gained slightly better points than other rest ones e.g
Kookaburra and Corchess are performed with best performance, actually
During almost all of the championship, Kookaburra 2 was leading..where
After several entries, Corchess 170224 managed to be Number One too!
Btw, old, former Champ (AWOL11) tested just for comparing..so in latest
3-4 months it is clear that there is minus plus 10 Elo improvement..nice!
Note that in this competition all engines are played without Eng learning...
As main reasons, simply for more fair conditions..otherwise, the learned
Exp files may play as serious advantage over the next entries...because
Some of the participants are entered later and played in Gauntlet mode..
Note also that 2nd name of this CS can be called Candidates tour, e.g
I hope to use those Top engines, which are capable to use BIN books in
Planning Online-Like CS..but sad news is that unfortunately not all of
The current Top engines have option/future to use BIN books..it's a pity..
However, good news is that there are many strong ones too, which are
Capable to use BIN, even CTG books...and what we can need more ?)
Actually with Polyglot adapter..it's possible but the strength is effecting..
In other words, in the new Online-Like CS, the Book Authors will be free
To use many of current engines, sure which are stable with book settings..
But very soon I will check more seriously, which are best for book tours...
And as final words, My congratulations and many thanks to all authors!

The Champions: CorChess, Kookaburra - Congrats to I. Ivec, Dorzs, SF Team!

For Final Standings:
https://sites.google.com/site/computers ... vs-47-cand

GAMES:
https://mega.nz/file/axwjVDZL#-RxA_R3yr ... 6yN_pZFgAU

Best,
Sedat

Posted: **Thu Feb 29, 2024 7:30 am**

Hello Chess Friends,

Just I'd like to inform you that,
Here is another new Strength test, before FINAL Official start:

To be more clear,
Book Authors will be free to choose one of the current engines!
In other words, in planning tour: Book entries will be played by
Author's choice! like Online..

Be aware of that too, I hope to run/organize
Two separate ratings, firstly with 4 Core/Game, Bullet, Ponder ON
Later... with highest Ranked Books: 8 Core/Game, Blitz, Ponder ON

Btw,
Here are planning engines, which may be used in new Online Like CS...
About current test conditions:
2x Epyc 7B12, CuteChess, 1m+0.5s, Ponder ON, 4 Core/Game, Balsa, 128 hash:

Code: Select all

Rank Name                          Elo     +/-   Games   Score    Draw 
   1 Incognito 5 Pro                13      14      82   51.8%   96.3% 
   2 Polyfish 140124                 4      18      83   50.6%   94.0% 
   3 Brainlearn 27                   4      14      86   50.6%   96.5% 
   4 ShashChess 34.6                 0      11      85   50.0%   97.6% 
   5 SF-PB 080124                    0       0      82   50.0%  100.0% 
   6 Raid v3.4                      -4       8      86   49.4%   98.8% 
   7 CoolIris 11.80                 -4      14      84   49.4%   96.4% 
   8 Eman 9.90                     -12      13      86   48.3%   96.5% 

337 of 2800 games finished.

And meanwhile, sure and just in case of interest,
You can submit your opening book/s at any time...
Because it's time for fun and let's best book win !)

More info very soon...

Greetings )

Posted: **Thu Feb 29, 2024 8:05 am**

UPDATE

Current Leader: Incognito 5 Pro - Congrats to Eduard Nemeth, SF team!

Code: Select all

Rank Name                          Elo     +/-   Games   Score    Draw 
   1 Incognito 5 Pro                 6       5     250   50.8%   98.4% 
   2 CoolIris 11.80                  1       6     249   50.2%   98.0% 
   3 SF-PB 080124                    1       3     250   50.2%   99.6% 
   4 Polyfish 140124                 1       7     251   50.2%   97.2% 
   5 ShashChess 34.6                 0       4     250   50.0%   99.2% 
   6 Brainlearn 27                   0       5     250   50.0%   98.4% 
   7 Raid v3.4                      -4       5     250   49.4%   98.8% 
   8 Eman 9.90                      -6       5     250   49.2%   98.4% 

1000 of 2800 games finished.

Posted: **Thu Feb 29, 2024 8:47 am**

UPDATE 2

Current Leader: Incognito 5 Pro - Congrats to Eduard Nemeth, SF team!

Code: Select all

Rank Name                          Elo     +/-   Games   Score    Draw 
   1 Incognito 5 Pro                 4       5     353   50.6%   98.3% 
   2 Brainlearn 27                   2       5     356   50.3%   98.3% 
   3 Polyfish 140124                 1       5     355   50.1%   98.0% 
   4 CoolIris 11.80                  1       4     355   50.1%   98.6% 
   5 SF-PB 080124                    0       3     355   50.0%   99.4% 
   6 ShashChess 34.6                -2       4     355   49.7%   98.9% 
   7 Raid v3.4                      -3       3     355   49.6%   99.2% 
   8 Eman 9.90                      -3       4     352   49.6%   98.6% 
   1418 of 2800 games finished.

IMPORTANT NOTE:
Please don't forget to mention the chess engine, because as I stated earlier:
In this new tournament, your book/s will be tested according to your wishes...
In other words, you can choose from the currently active strength test...

Chess Forum

Sedat Canbaz

Re: Sedat Canbaz

Re: Sedat Canbaz

Re: Sedat Canbaz

Re: Sedat Canbaz

Re: Sedat Canbaz

Re: Sedat Canbaz

Re: Sedat Canbaz

Re: Sedat Canbaz

Re: Sedat Canbaz

Re: Sedat Canbaz