Sedat Canbaz

[Sedat Canbaz]

Hello Chess Friends,

Another exciting Duel idea is born:
Very soon, probably today I plan to
Organize: FRINGE407 Vs ProteusSFX AI

And if you are going to ask Me why... ?)
For some reasons e.g according to Lichess server,
ProtesusSF is highest rated player (sure in Elo points)
And if I'm not wrong, rated minus plus 3300, right?
Where FRINGE407 as new incomer has lower Elo points,
But managed to defeat him with BIG margin (+228 Elo)
And I wonder, how this can be possible...?)
Anyhow, under SCCT cond.. no cherry picking system!
No hardware speed advantages...and both will be played
By same Top NNUE engine..against any confusions etc.

To be more clear, based on recent Lichess database,
Both Online players are already played each other...
But I assume under not equal hardware speed cond.
So I wonder what will be results under more fair cond.
For this reason, I'm going to create two based books,
Which will be tuned to prefer their recent openings...
Since the planning books will be based on many games,
Especially ProteusSFX will contain thousands of games.
In same time that means, we can not expect identical
Openings.. I mean like previous played theirs 100 games...
However..after all, I think that it's worth giving a try... !)

Btw, My Congratulations to Fringe...really impressive!
And here are their mentioned, previous Lichess results:

Code: Select all

1   FRINGE407      3009  +26/-1/=73 62.50%   62.5/100
2   ProteusSFX-AI  3149  +1/-26/=73 37.50%   37.5/100

In other words,
Let's see who will be Winner under equal hardware cond ?!
And any guess: ProteusSFX or FRINGE407 ?)

Good luck to both..and please stay tuned,
Soon as possible, the Duel will be played...

Have fun )

[Sedat Canbaz]

The DUEL - FRINGE407 Vs ProteusSFX is ended:
The Winner is FRINGE407 - Congartulations!

Code: Select all

Score of FRINGE407 vs ProteusSFX: 10 - 5 - 483 [0.505]
...      FRINGE407 playing White: 10 - 5 - 233  [0.510] 248
...      FRINGE407 playing Black: 0 - 0 - 250  [0.500] 250
...      White vs Black: 10 - 5 - 483  [0.505] 498
Elo difference: 3.5 +/- 5.3, LOS: 90.2 %, DrawRatio: 97.0 %
500 of 500 games finished.

GAMES:
https://pixeldrain.com/u/1Exfjb3b

Conditions:
2x Epyc 7B12, CuteChess, 1 Core, Ponder OFF, 1m+0.6s, 64 MB Hash, 4-MEN
Note: Both participants are played via Raid v3.4 engine (by Tanick Ramz)
And for better performance: for each player is installed a copy of engine..

As final words,
Under equal hardware speed cond. no chances to see
Similar points difference as Online...on other hand,
Sure for a better conclusion more games are required,
But according to latest played results it is very clear
Also that FRINGE407's used openings are more dangerous!

And I hope also that,
You like my current organized Duel...?)

Ok friends... that's all for now.. see you later


Best,
Sedat

[Fringe407]

My dear friend Sedat,

I congratulate you on this marvellous idea and thank you for the test.

I hope that Alex Morales will stop shaking his head behind my back and start working on his engine and his opening book. His debates are doing nothing but taking chess backwards. Beating 2400 elo bots won't make you a champion.

My respects to you Sedat

[Solista]

Thanks for the test. ProteusSF's tables (in his forum) are useless. Because everyone who has a rating of +2900 in Bullet plays in matchmaking mode (which is like Automatic on PlayChess). There is usually a difference of +- 300 Elo max in the formula. Nobody plays against weaker player. The vast majority of the best players have around 3000 Elo in Bullet. Most games for the strongest players are played between 2900 and 3000 Elo. These matches end in a draw 95%. This means that the Elo of these players is always around 3000. Players with <2700 Elo in Bullet have absolutely no chance against me. That's why I usually play against 2900 to 3000. ProteusSF is a coward. He never plays in matchmaking mode. He has no chance there. He plays against Elo max 2700 and manually. Although he already produces a lot of draws against 2700 and loses points. That's why he plays most games (99%) against Elo 2400. That's a joke. I also win all games against Elo 2400. Because that would be a joke, I don't play bullet at all against such weak opponents. My opponents should have at least >2700, otherwise I won't have any fun.

[Sedat Canbaz]

Thanks for the test. ProteusSF's tables (in his forum) are useless. Because everyone who has a rating of +2900 in Bullet plays in matchmaking mode (which is like Automatic on PlayChess). There is usually a difference of +- 300 Elo max in the formula. Nobody plays against weaker player. The vast majority of the best players have around 3000 Elo in Bullet. Most games for the strongest players are played between 2900 and 3000 Elo. These matches end in a draw 95%. This means that the Elo of these players is always around 3000. Players with <2700 Elo in Bullet have absolutely no chance against me. That's why I usually play against 2900 to 3000. ProteusSF is a coward. He never plays in matchmaking mode. He has no chance there. He plays against Elo max 2700 and manually. Although he already produces a lot of draws against 2700 and loses points. That's why he plays most games (99%) against Elo 2400. That's a joke. I also win all games against Elo 2400. Because that would be a joke, I don't play bullet at all against such weak opponents. My opponents should have at least >2700, otherwise I won't have any fun.

Hello Eduard,

Not at all...and thanks for your replay...

You may remember, some time ago,
I was not aware what is going on Online..for this reason
I preferred to not make any comments over.. but nowadays,
After checking latest Lichess database, you have right!

Unfortunately but usually ProteusSF is played vs weaker opponents,
Mainly about 800-1000 Elo weaker.. hehe ) in same time that means
He is played vs weaker engines which are close to GMs level..then
Of course under these easy, funny cond. anyone will get better points!
Just imagine, Stockfish 16.1 Vs Fruit 2.2 match... who will win ?)

In other words (about cherry picking system):
Rather I'd say to deceive oneself... no more no less..

As final words, I think also that
The real problem is in Lichess rules, normally just in case..
Lichess admins should avoid these kind of joke playing system!
Otherwise, I'm afraid to say that
We'll see such users (like ProteusSF) with non-realistic Elo points!

Best,
Sedat

[Sedat Canbaz]

Breaking News (05.04.2024):
Well, here is just another CS based on latest Top engines, in same
Time more original I mean not so closed (in similarity) to each other..
However since NNUE on scene... the originality is almost dead...as
Clear proof, all of them are using NNUE technology ) on other hand,
We have to be satisfied with that.. since is best project to gain better
Points..at least for nowadays...but this is also true, the improvement
Seems to be not so easy...especially with engines which are 3700+
Ok..here are latest played results..and as usually, Stockfish wins..!!

The Champion: Stockfish 16.1 - Congratulations to Stockfish Team!

FOR FINAL STANDINGS:
https://sites.google.com/site/computers ... t-nn-cs-48

GAMES:
https://mega.nz/file/X1xyxALR#VNJTCaI9S ... H5nyOT55q8

Best,
Sedat

[Sedat Canbaz]

Hello Chess Friends,

I am very pleased to announce that
Just another BOOK NNUE CS started..

What is new?
All Book entries will be played by Brainlearn 27 Vs much weaker
Constant engines, which are about 100 Elo weaker than Brainlearn,
Where all weaker engines will use to play via Balsa-Wins book,
In this way, there will be BIG fun, because the overall draws are
Expecting much lower..reason, about 100 Elo Eng NN difference!

Why Balsa (for all constant 10 engines)?,
1st it's one of strongest suites, which prefer various openings!
2nd via Balsa, we will have more chances to see which books are
Better in overall openings..just TREND openings...not enough !!
3rd. Besides, Balsa book includes not so many BIG holes for sure!
Otherwise, let's say we used SPCC, CCRL, TCEC etc. we will have
Many weaker lines, that means easier points.. sorry, this time,
The tour's target is to see stronger lines..that means harder..!

And why Brainlearn 27 (main engine for planing book entries)?
For several reasons, 1st the Strongest, less Drawish, sure
Under these used/played conditions...note also that current
Brainlearn is special ver, compiled by Mr. Tanick Ramz (Orgz),
Futures CTG, BIN, other important future has small size (1.8 MB)
And in same time, compatible to use external NNUE Evalfiles!
You may know, otherwise if using for each book player 76 MB:
It is simply pointless...not logical... right ?

And here are latest Candidate tour results, played at Bullet 1m+06s,
Note that due to no free time, all newer 3750+ are not tested..but
Just..via strong openings: no much improvement...this is a fact!
And I plan to pick 10 engines, which will be about 100 Elo weaker:

Code: Select all

   # PLAYER              :  RATING  POINTS   GAMES   (%)
   1 Brainlearn 27       :  3804.1   651.5    1000    65
   2 RapTora 2.3         :  3797.5   643.0    1000    64
   3 Raid 3.5            :  3796.2   706.5    1100    64
   4 Raid 3.4            :  3788.2   695.0    1100    63
   5 CiChess 040121      :  3710.6   571.0    1100    52
   6 Caissa 1.18         :  3710.3   624.0    1200    52
   7 Cfish 250621        :  3705.9   563.0    1100    51
   8 Eman 6.91           :  3705.0   561.5    1100    51
   9 Crystal 060622      :  3700.4   804.5    1700    47
  10 Stockfish 13        :  3700.0   453.5    1000    45
  11 KDragon 2.5         :  3699.5   852.0    1800    47
  12 KDragon 3.3 MCTS    :  3695.1   840.5    1800    47
  13 CF-EXT 161120       :  3694.4   736.5    1600    46
  14 FatFritz 3          :  3692.0   783.5    1700    46
  15 Obsidian 11         :  3689.3   774.5    1700    46
  16 Berserk 11.1        :  3688.6   775.0    1700    46
  17 Caissa 1.16         :  3688.2   534.0    1200    45
  18 Raubfisch X45       :  3687.6   772.5    1700    45
  19 ShashChess 16       :  3685.5   458.0    1000    46

In other words,
Let's see under these new cond. which books will gain more points?)
DON, OPTIMUS, PYRAMID, TURKEY.. one of them will be next Champ ?
Or any other ? ok...we will see... after all, time will tell...)

And please stay tuned, more info etc. very soon )


Best,
Sedat

[Sedat Canbaz]

UPDATE

DON is entered...

Note: This is the same DON vers, which managed to be
Number One in previous/ recent organized book tours!

Code: Select all

Rank Name                          Elo     +/-   Games   Score    Draw 
   0 DON 010324                     91      13     628   62.7%   74.2% 
   1 Crystal 060622                -66      35      64   40.6%   81.3% 
   2 Caissa 1.18                   -67      39      63   40.5%   77.8% 
   3 FatFritz 3                    -69      36      61   40.2%   80.3% 
   4 Cfish 250621                  -72      36      64   39.8%   79.7% 
   5 KDragon 3.3 MCTS              -74      37      62   39.5%   79.0% 
   6 Berserk 11.1                  -96      41      63   36.5%   73.0% 
   7 CF-EXT 161120                -102      42      63   35.7%   71.4% 
   8 Obsidian 11                  -108      44      63   34.9%   69.8% 
   9 KDragon 2.5                  -116      45      62   33.9%   67.7% 
  10 CiChess 040121               -139      49      63   31.0%   61.9% 

628 of 1000 games finished.

[Sedat Canbaz]

Update 2

This time unfortunately I have not so good news,
The planning:/started new CS is already cancelled!

As reason, for example, in Gauntlet mode:
3 of current 3700 Elo engines are not capable to use Balsa-Wins bin!
Actually normally they are capable to use...but it all depends, e.g
If including a few players (4 books) it's all right..no problem...but
If adding all players (11 books), then Cfish, Cichess and CF-EXT are
Starting to play as without book..very strange...never seen before!
But at least I'm sure that rest 7 engines are used to play via Balsa...

Btw, just I'd like to point out also that
Brainlearn 27 is one best, stable engines if using BIN books..I mean
It does not matter how many books are entered in Gauntlet mode etc.

As final conclusion,
Not sure exactly, very likely the programmers will explain better...
But I think that current strange behavior (about the 3 engines, which
Can't use bin book via 11 players) can be said as Engine bin book bug!
Because if this was as Cutechess GUI bug, then in all other rest engines
Would be appeared this unusual book bug issue...or simply I can say that,
The mentioned 3 engines suffer under these cond...no more.. no less!

Anyhow...soon as possible another new tour coming..
And meantime, I hope to pick more stable engines

Best,
Sedat

[Sedat Canbaz]

Hello there )

Well, this time I have good news,
I start organizing another new CS!

Note that
Current tour will contain two Leagues: Second and Super League,
Where all entries will be played vs Balsa Wins (in 2nd League)
Any Book, which will gain +10 Elo points better (than Balsa..)
Will be qualified Player into planning new Super League tour!

And to be more clear
All books will be played by BrainLearn 27, compiled by Orgz!
For less draws.. Bullet TC (30s+0.6s) used in Second League,
Btw, Balsa Wins is based on wins, played by Balsa series...
Futures small size and its deep opening lines are not tuned
For anyone missed, why Balsa ? to stress the latest books
With more various openings...only TREND is not enough..
Plus in this way, the played results can be used as good
Indicator about which are better in overall openings!

And here are the 1st played results of 2nd League tour,
Interesting... e.g so far Balsa is leading (over DON...)
Sure in my guess, DON book has small chances to qualify..
Anyhow, let's see what will be standings after 1000 games

Code: Select all

Score of DON 210324 vs Balsa Wins: 4 - 7 - 490 [0.497]
...      DON 210324 playing White: 4 - 3 - 241  [0.502] 248
...      DON 210324 playing Black: 0 - 4 - 249  [0.492] 253
...      White vs Black: 8 - 3 - 490  [0.505] 501
Elo difference: -2.1 +/- 4.5, LOS: 18.3 %, DrawRatio: 97.8 %
SPRT: llr 0 (0.0%), lbound -inf, ubound inf
501 of 1000 games finished.

And please stay tuned,
Very soon more info, entries coming!

Good luck to all )


Best,
Sedat