Page 14 of 70

Re: Sedat Canbaz

Posted: Sun Sep 10, 2023 1:03 pm
by Sedat Canbaz
Breaking News (10.09.2023):
Here are new played results of Super League, note that the tour is still active...
And again and again, RapTora's author is managed to be Champion of SL!!
As I mentioned before, Mr. Anton is a real master for sure, for more: Details
https://open-chess.org/viewtopic.php?f= ... &start=320
And it seems A. Ces belongs to the best youngest talents (in programming)
Eduard is Eduard, his engine 2nd rank no BIG surprise that again on Top!
Sure Stockfish 240823 is 2nd too, no doubt another beast, plus same words
Are going for SF-Poly, Raid etc as well, and let's see what will be next results
And soon new updates and please stay tuned.btw, New Champion Rankings
Meanwhile, for previous postings, regarding current League tourneys: News

The Champion: SpecTral 5.0 - Congratulations to Anton Ces, Stockfish team!

Code: Select all

Rank Name                  Elo   +    -  games score oppo. draws 
   1 SpecTral 5.0         3802    9    9  1000   52%  3788   76% 
   2 Stockfish 220823 sl  3798    9    9  1000   51%  3788   77% 
   3 Stockfish 240823     3798    9    9  1000   51%  3788   78% 
   4 SF-Poly 220723       3797    9    9  1000   51%  3789   79% 
   5 Raid v2.73           3794    9    9  1000   51%  3789   78% 
   6 Tactical 220823      3790    9    9  1000   50%  3789   76% 
   7 Eman 9.40            3790    9    9  1000   50%  3789   77% 
   8 Brainlearn 25.2      3785    9    9  1000   49%  3790   73% 
   9 Corchess 240823      3782    9    9  1000   49%  3790   78% 
  10 Cool-Iris 11.10      3776    9    9  1000   48%  3791   80% 
  11 ShashChess 33.2      3770    9    9  1000   47%  3791   76% 
For More Info:
https://sites.google.com/site/computers ... per-league

Have fun,
Sedat

Re: Sedat Canbaz

Posted: Sun Sep 10, 2023 1:16 pm
by Sedat Canbaz
Overall's Super League results so far (1st and 2nd tour):
NOTE: All of SL's games are played with 3mn+1sec

Code: Select all

Rank Name                  Elo    +    - games score oppo. draws 
   1 SpecTral 5.0         3802    9    9  1000   52%  3788   76% 
   2 Stockfish 220823 sl  3798    9    9  1000   51%  3788   77% 
   3 Stockfish 240823     3798    9    9  1000   51%  3788   78% 
   4 SF-Poly 220723       3797    9    9  1000   51%  3789   79% 
   5 Raid v2.73           3794    9    9  1000   51%  3789   78% 
   6 Tactical 220823      3790    9    9  1000   50%  3789   76% 
   7 Eman 9.40            3790    9    9  1000   50%  3789   77% 
   8 Brainlearn 25.2      3785    9    9  1000   49%  3790   73% 
   9 Corchess 240823      3782    9    9  1000   49%  3790   78% 
  10 Cool-Iris 11.10      3776    9    9  1000   48%  3791   80% 
  11 ShashChess 33.2      3770    9    9  1000   47%  3791   76% 
-------------------------------------------------------------------------
1st Tour: started with Bullet learned exp files from 2nd League

Code: Select all

Rank Name                  Elo   +    -  games score oppo. draws 
   1 Stockfish 240823     3812   13   13   500   53%  3791   79% 
   2 SpecTral 5.0         3809   13   13   500   53%  3791   79% 
   3 Raid v2.73           3804   13   13   500   52%  3792   79% 
   4 Stockfish 220823 sl  3799   13   13   500   51%  3792   77% 
   5 Tactical 220823      3798   13   13   500   51%  3793   73% 
   6 SF-Poly 220723       3797   13   13   500   51%  3793   80% 
   7 Eman 9.40            3788   13   13   500   49%  3794   76% 
   8 Corchess 240823      3786   13   13   500   49%  3794   77% 
   9 Brainlearn 25.2      3782   13   13   500   48%  3794   70% 
  10 Cool-Iris 11.10      3779   13   13   500   48%  3794   79% 
  11 ShashChess 33.2      3770   13   13   500   46%  3795   73% 
-------------------------------------------------------------------

2nd Tour: started with learned exp files from Bullet + Blitz tour

Code: Select all

Rank Name                  Elo    +    - games score oppo. draws 
   1 SF-Poly 220723       3798   13   13   500   52%  3784   78% 
   2 Stockfish 220823 sl  3797   13   13   500   52%  3784   77% 
   3 SpecTral 5.0         3795   13   13   500   52%  3784   73% 
   4 Eman 9.40            3791   13   13   500   51%  3785   77% 
   5 Brainlearn 25.2      3789   13   13   500   51%  3785   75% 
   6 Raid v2.73           3785   13   13   500   50%  3785   77% 
   7 Stockfish 240823     3783   13   13   500   50%  3786   78% 
   8 Tactical 220823      3781   13   13   500   49%  3786   79% 
   9 Corchess 240823      3779   13   13   500   49%  3786   79% 
  10 Cool-Iris 11.10      3773   13   13   500   48%  3787   80% 
  11 ShashChess 33.2      3770   13   13   500   48%  3787   78% 
---------------------------------------------------------------------

3rd Tour: currently active, but without learned exp files
----------------------------------------------------------------------

More Details.
And after the end of 3rd tour, any player will contain 1500 games
Plus we will have some chances to compare influences between
Learned: Bullet exp vs Blitz Exp files sure vs NON-learned...too

Best,
Sedat

Re: Sedat Canbaz

Posted: Sun Sep 10, 2023 7:59 pm
by Sedat Canbaz
SCCT NNUE Championship XXXV is ended !

Super League:
https://sites.google.com/site/computers ... per-league

Second League:
https://sites.google.com/site/computers ... eague-3700

Games:
https://mega.nz/file/a1oQTbJL#ZusWDTWcT ... fmWEwhqT-g

Note: Now I'm busy, but 3rd test ranking (no exp) will be published later


Best,
Sedat

Re: Sedat Canbaz

Posted: Mon Sep 11, 2023 10:39 am
by Sedat Canbaz
Hello there )

As I promised..

1st Tour of Super League: Started with learned Bullet EXP files:

Code: Select all

Rank Name                  Elo   +    -  games score oppo. draws 
   1 Stockfish 240823     3812   13   13   500   53%  3791   79% 
   2 SpecTral 5.0         3809   13   13   500   53%  3791   79% 
   3 Raid v2.73           3804   13   13   500   52%  3792   79% 
   4 Stockfish 220823 sl  3799   13   13   500   51%  3792   77% 
   5 Tactical 220823      3798   13   13   500   51%  3793   73% 
   6 SF-Poly 220723       3797   13   13   500   51%  3793   80% 
   7 Eman 9.40            3788   13   13   500   49%  3794   76% 
   8 Corchess 240823      3786   13   13   500   49%  3794   77% 
   9 Brainlearn 25.2      3782   13   13   500   48%  3794   70% 
  10 Cool-Iris 11.10      3779   13   13   500   48%  3794   79% 
  11 ShashChess 33.2      3770   13   13   500   46%  3795   73% 
---------------------------------------Vs---------------------------------------

3rd Tour of Super League: Started without any learned EXP files:

Code: Select all

Rank Name                  Elo    +    - games score oppo. draws 
   1 Stockfish 220823 sl  3812   12   12   500   53%  3789   77% 
   2 Raid v2.73           3812   12   12   500   53%  3790   77% 
   3 SpecTral 5.0         3797   12   12   500   51%  3791   82% 
   4 Eman 9.40            3795   12   12   500   51%  3791   78% 
   5 Stockfish 240823     3794   12   12   500   50%  3791   81% 
   6 Cool-Iris 11.10      3793   12   12   500   50%  3791   80% 
   7 Tactical 220823      3789   12   12   500   50%  3792   82% 
   8 SF-Poly 220723       3787   12   12   500   49%  3792   78% 
   9 Corchess 240823      3782   12   12   500   48%  3793   80% 
  10 Brainlearn 25.2      3778   13   12   500   48%  3793   70% 
  11 ShashChess 33.2      3770   12   12   500   47%  3794   77% 

More Details:
For better conclusion we need more games, but it's clear that
Some of the Top chess engines are performed better, some worse!
However, I think that these stats can't be taken as very serious data,
Because, small test..for better stats, we need much more testings!
Note: 2nd Tour (Bullet + Blitz) is not so good idea for exp stats..
Otherwise, we'll have troubles in recognizing the exp influences!
For this reason, only 1st and 3rd tours published for comparing..


Best,
Sedat

Re: Sedat Canbaz

Posted: Mon Sep 11, 2023 12:44 pm
by Sedat Canbaz
Breaking News - 11.09.2023:
Well, current championship ended, congrats and many thanks to all!
Normally the plan was to be resumed but anyhow, I have good news:
Very soon I hope to organize a new book tour but as next under Fritz,
That means in same time I missed a lot testing CTG opening books )
Note that in planning tour: all books will be allowed (Private + Public)
Plus as usual, all books will be played by same engine: Raubfish SC
And for more fair cond.. any CTG's book learning will be disabled..
As reason, (if Learning ON) BIN books may suffer vs CTG books
Ok..that's all for now..and please be ready for the real 'ARENA' !!

Btw, as I promised, all played games available (in PGN):
https://sites.google.com/site/computers ... t-bn-cs-14

Source:
https://sites.google.com/site/computers ... t-bn-cs-14

Greetings

Re: Sedat Canbaz

Posted: Mon Sep 11, 2023 2:20 pm
by Sedat Canbaz
By the way, the tournament page is created, for more details:
https://sites.google.com/site/computers ... t-bn-cs-15

Greetings )

Re: Sedat Canbaz

Posted: Mon Sep 11, 2023 3:20 pm
by Anton101
Thanks a lot!

Best regards.

Re: Sedat Canbaz

Posted: Tue Sep 12, 2023 6:59 am
by Sedat Canbaz
Breaking News - 12.09.2023:
The current championship is ended too... congrats and thanks to all authors!
Normally the tour was planning to be resumed, but I've lost my interest for two
Main reasons, 1st due to very high draw perc. and as we see 97% draw-ratio!
Really boring for my side, btw, as usual I say over (I mean since no much wins):
What is the point of using an umbrella, If no much 'rain' on horizon ?)
Right? and correct Me please if I am wrong here..?) and as 2nd main reason,
Just in case of new Exp entries: I expect the learning will make not much sense,
E.g (according to my other Gauntlet tests) via concurrent: 64, no BIG influence!
But I admit that e.g as current in Round-Robin tour (via concurrent: 64 games)
Then situation seems to be ok...due to in round-robin tours: too many players!
In short, via concurrent minus plus 10' is ok, but via higher... not so good idea!
Sure some may say such as: run via concurrent 10.. İf so .. will be really sad,
Because current tournament machine is based on 128 Cores / 256 Threads!
Like a huge BUS / PLANE (with many seats) but carries...a few passengers !)
And sorry that the current tour is not resumed ..but I think that will be pointless,
If we will run next Exp in Gauntlet conditions..btw here is another useful saying:
If everything were as it seems, the sea water you hold would be blue

For More Details, Standings, Games:
https://sites.google.com/site/computers ... t-ben-cs-1

Greetings )

Re: Sedat Canbaz

Posted: Tue Sep 12, 2023 7:00 pm
by Sedat Canbaz
Breaking News (12.09.2023):
What is new in this competition: Balsa openings used (instead of Unique..) plus
Main target, testing same qualified engines of Super League vs weaker engines,
Which are about 50-70 Elo, yes.. a new idea, new concept of testing..you know,
Many of us prefer weak openings to reduce the draw ratio...that's all right..and
As usual.if test via weak openings,then thousands of games (per player) required,
And now just opposite..matches vs weaker opponents+strong openings..and after
The tour..more info will be about the overall draws plus we will need thousands of
Games (per player) or not.. My view: 1000 games is enough...after all time will tell..
But there is one true, via strong openings is more safe..plus Elo difference is less..
And for right now we have too many Number ones, if I am not wrong 5 Leaders)
What a democracy under SCCT !)

For Latest Standings:
https://sites.google.com/site/computers ... nnue-cs-36



Best,
Sedat

Re: Sedat Canbaz

Posted: Wed Sep 13, 2023 11:29 pm
by Sedat Canbaz
Breaking News (14.09.2023):
The current championship is ended... congrats and many thanks to all authors!
I realized each player to consist 1350 games, and frankly now it's quite enough,
Regarding to determine any engine strength...yes...depending on our conditions..
We may see different stats..e.g current draw-ratio is not high, just 79% Why less?
If not clear..as I stated before, simply because I used to play weaker engines too
E.g 3650-3700 Elo, but as you may know, if we run matches: 3750 vs 3750 Elo
Then to see a win is became a dream !) exc. via critical weak opening lines etc.
But this is also true (not joke): even via strong openings..we may see more wins!
Actually it's all in our hands.. at least I can say for my side as a tester / director:
I do my best sure just for Computer chess's progress! actually not since today !
What I can say more, the current testing system is ideal..about 1000 games is ok,
E.g not needed 2000-3000 games (per player) plus not boring, (many.wins) and
Via current testing system e.g vs weaker engines such as about 50-70 Elo weaker
I may improve / tune Balsa weak openings.e.g just imagine 3750+ lost to 3675
Sure via current new system.. Engine authors may improve their engines as well..
Not always a defeat is due to book error..sometimes engine/s can be as reason..
Ok... this time: Mr.Tanick Ramz's engine (Raid v2.73) said the last word !!
Actually 2nd place, belongs to Mr. Tanick's engine (SF-Poly 22072), too in short,
Another great talent for sure..plus he is author of opening books as well..nice!
Plus 2nd place belongs to two more engines: BrainLearn and Corchess too!
On other hand, rest Super League's players are so close each other in points

The Champion: Raid 2.73 - My Congratulations to Tanick Ramz, SF team!
Note: Exactly same Raid vers. is managed as 2nd in recent both leagues!
In other words, if not so clear: this is not a luck.. that Raid is Number One!

For Final Rankings:
https://sites.google.com/site/computers ... nnue-cs-36

Games:
https://mega.nz/file/vtx0HZyA#vjNZATm7H ... v4ywwIQVc4

Greetings )