Sedat Canbaz

Post Reply
User avatar
Sedat Canbaz
Posts: 688
Joined: Mon Jun 19, 2023 8:49 am
Has thanked: 730 times
Been thanked: 2359 times

Re: Sedat Canbaz

Post by Sedat Canbaz »

massimilianogoi wrote: Thu Jul 20, 2023 11:12 pm That Raubfisch should be a good candidate. I was just advicing, no time for that here either.
Sedat Canbaz wrote: Thu Jul 20, 2023 7:44 pm
massimilianogoi wrote: Thu Jul 20, 2023 4:58 pm I'm just reading it now. My two cents: you should have used the same engine, it would have turned out a really interesting test.
Well, that's all from Me..

Maybe you can try a newer test..because I am on holiday mode )
Then your work/test would have turned out a really interesting test.

But just do not use such engines, which produce close to 97% (as NON-NNUE 1m+0.5s)
Otherwise..it is too boring....at least for my side.. )
I see.. and thanks for your advice, but anyhow
I don't plan to run new Raubfish SC strength tests..
Reason, I've many... and frankly sometimes I wonder
What I'm missing..really..and this is not as joke )

Sure if you are interested,
Here are some Raubfish SC strength test results:

SCCT BOOK / NNUE CS IV:
https://sites.google.com/site/computers ... -4-details

SCCT BOOK / NNUE CS V:
https://sites.google.com/site/computers ... -cs-5-news

Raubfish SC vs Others:
https://sites.google.com/site/computers ... testings-3

If you say...still not enough.. more data required about
Raubfish SC's draw ratio, then I suggest to check my tours...
Usually they are minus plus with 90% draw-ratio

And frankly so far,
I did not count RaubfishSC games.. but I expect over 300.000+

Btw, it is not so bad idea to compare with others,
I mean with NON-SCCT book results...

Hope helps..

Greetings
User avatar
massimilianogoi
Site Admin
Posts: 394
Joined: Thu Aug 04, 2022 1:42 pm
Has thanked: 603 times
Been thanked: 647 times
Contact:

Re: Sedat Canbaz

Post by massimilianogoi »

That's not the point. You said Raubfisch is a good engine, I trust it. The idea - a real cool one - you had it: testing an engine with many cores without NNUE versus one with a single core but with NNUE. To be completely loyal with the test the same engine should be used, that's what I think, because using another engine means another algorythm with several modifications.
Sedat Canbaz wrote: Fri Jul 21, 2023 2:57 amIf you say...still not enough.. more data required about
Raubfish SC's draw ratio, then I suggest to check my tours...
Usually they are minus plus with 90% draw-ratio

And frankly so far,
I did not count RaubfishSC games.. but I expect over 300.000+

Btw, it is not so bad idea to compare with others,
I mean with NON-SCCT book results...

Hope helps..

Greetings
People who have lost the hope.
User avatar
Sedat Canbaz
Posts: 688
Joined: Mon Jun 19, 2023 8:49 am
Has thanked: 730 times
Been thanked: 2359 times

Re: Sedat Canbaz

Post by Sedat Canbaz »

massimilianogoi wrote: Fri Jul 21, 2023 3:08 am That's not the point. You said Raubfisch is a good engine, I trust it. The idea - a real cool one - you had it: testing an engine with many cores without NNUE versus one with a single core but with NNUE. To be completely loyal with the test the same engine should be used, that's what I think, because using another engine means another algorythm with several modifications.
Well, but it seems you missed..

Your words:
To be completely loyal with the test the same engine should be used,

Then I say:
Almost all my book/nnue tours are played by Raubfish SC vs Raubfish SC
I mean the picture is very clear...regarding to see overall draws...

Sure, if you say (using only Raubfish SC for both players)
E.g Run a test especially NNUE 1cpu vs NON-NNUE 6cpu

Then I hardly doubt that, (even if we use same Raubfish SC)
That we will see much Elo points difference ...
To be more clear, minus plus
The Elo difference is expecting as CF-EXT 291120
I mean, over than 100+ Elo is expecting if we run
Raunfish NNUE 1cpu vs Raunfish NON-NNUE 6cpu

Btw, just do not ask,
Which player will be winner )) if you have some doubts
I strongly believe that the results will be ended in favor
For Raunfish SC 1cpu, and more than 100+ Elo

However,
This is other question, but I think that you missed to read..
I suggest you to re-read on Outskirts chess forum..more details there..
Anyhow, I will try to explain:
In other words, in those days... my target was to see the real power..
I was not sure exactly which engine to use (before starting a tour)
And after running the strength test..I found that it is better to use
Raubfish X48c SC NNUE (instead of using CF-EXT 291120 NON-NNUE)

Ok.. it is time to take a good rest.. see you later )
User avatar
massimilianogoi
Site Admin
Posts: 394
Joined: Thu Aug 04, 2022 1:42 pm
Has thanked: 603 times
Been thanked: 647 times
Contact:

Re: Sedat Canbaz

Post by massimilianogoi »

Ok, thanks for the explanation.
People who have lost the hope.
User avatar
Sedat Canbaz
Posts: 688
Joined: Mon Jun 19, 2023 8:49 am
Has thanked: 730 times
Been thanked: 2359 times

Re: Sedat Canbaz

Post by Sedat Canbaz »

Hello there )

Well, just out of curiosity I've realized to run another new Duel match
For example, almost same test conditions as my previous exp testings,
Main difference, instead of SF -PB, this time Eman 8.91 vs RapTora 2.2

To be more clear,
Both (Eman 8.91 and RapTora 2.2) are played via
Same pre-learned Eman exp (dated on 17.07.2023),
But now, after checking.. both dated as 25.07.2023
So it seems that the engines are used to learn...
Btw, for each player is installed separate exp file..
RapTora 2.2 is the version, which is compiled by
Eduard Nemeth, sure as main author is Anton Ces
And plus all Stockfish developers..

And here are the FINAL results, the Winner:
RapTora 2.2 - Congrats to Anton, Eduard, SF team!

Code: Select all

Score of Eman 8.91 vs RapTora 2.2: 9 - 17 - 74 [0.460]
...      Eman 8.91 playing White: 9 - 1 - 40  [0.580] 50
...      Eman 8.91 playing Black: 0 - 16 - 34  [0.340] 50
...      White vs Black: 25 - 1 - 74  [0.620] 100
Elo difference: -27.9 +/- 34.6, LOS: 5.8 %, DrawRatio: 74.0 %
100 of 100 games finished
GAMES:
https://mega.nz/file/3sBywaoS#mc16D7fGy ... OT_qxhSbPU

Conditions:
Ryzen7 4800H, CuteChess, 1 Core, Ponder OFF, 1m+0.6s, 128 MB Hash, 4-MEN
Note: Unique v50 opening suite is used (played each opening as twice)
CuteChess GUI's concurrent games: 10

And without to not mention this I can't as well,
For example, in my previous recent testings...
We noticed that RapTora 2.2 + Eman exp performed
Clearly better than Eman 8.91 + Eman exp

Ok.. let's say previous test was due to error margin
What about now? probably the Experts will explain
Better than Me... because this is Shocking News!
E.g in my understanding own Exp files have to do
Stronger with own Engines...really interesting..
In other words, I can not remember such record!
Since we are born..we are all learning..and if
You ask Me what I learned today? my replay is:
RapTora 2.2 engine is one of Best of the Best!
And not only in Eng learning..plus in stability!
Same words going for many other engines too!

And once more, I can summarize it:
Success is no accident..It's a hard work..

And as usual, thanks for your interest...


Best,
Sedat
User avatar
massimilianogoi
Site Admin
Posts: 394
Joined: Thu Aug 04, 2022 1:42 pm
Has thanked: 603 times
Been thanked: 647 times
Contact:

Re: Sedat Canbaz

Post by massimilianogoi »

Shocking difference! Could it be that the opening suite influenced the result?

Is this Raptora private? And the experience file?

Definitely intriguing test, I'd like to make it too, but with the same material.
Sedat Canbaz wrote: Tue Jul 25, 2023 8:18 am Hello there )

Well, just out of curiosity I've realized to run another new Duel match
For example, almost same test conditions as my previous exp testings,
Main difference, instead of SF -PB, this time Eman 8.91 vs RapTora 2.2

To be more clear,
Both (Eman 8.91 and RapTora 2.2) are played via
Same pre-learned Eman exp (dated on 17.07.2023),
But now, after checking.. both dated as 25.07.2023
So it seems that the engines are used to learn...
Btw, for each player is installed separate exp file..
RapTora 2.2 is the version, which is compiled by
Eduard Nemeth, sure as main author is Anton Ces
And plus all Stockfish developers..

And here are the FINAL results, the Winner:
RapTora 2.2 - Congrats to Anton, Eduard, SF team!

Code: Select all

Score of Eman 8.91 vs RapTora 2.2: 9 - 17 - 74 [0.460]
...      Eman 8.91 playing White: 9 - 1 - 40  [0.580] 50
...      Eman 8.91 playing Black: 0 - 16 - 34  [0.340] 50
...      White vs Black: 25 - 1 - 74  [0.620] 100
Elo difference: -27.9 +/- 34.6, LOS: 5.8 %, DrawRatio: 74.0 %
100 of 100 games finished
GAMES:
https://mega.nz/file/3sBywaoS#mc16D7fGy ... OT_qxhSbPU

Conditions:
Ryzen7 4800H, CuteChess, 1 Core, Ponder OFF, 1m+0.6s, 128 MB Hash, 4-MEN
Note: Unique v50 opening suite is used (played each opening as twice)
CuteChess GUI's concurrent games: 10

And without to not mention this I can't as well,
For example, in my previous recent testings...
We noticed that RapTora 2.2 + Eman exp performed
Clearly better than Eman 8.91 + Eman exp

Ok.. let's say previous test was due to error margin
What about now? probably the Experts will explain
Better than Me... because this is Shocking News!
E.g in my understanding own Exp files have to do
Stronger with own Engines...really interesting..
In other words, I can not remember such record!
Since we are born..we are all learning..and if
You ask Me what I learned today? my replay is:
RapTora 2.2 engine is one of Best of the Best!
And not only in Eng learning..plus in stability!
Same words going for many other engines too!

And once more, I can summarize it:
Success is no accident..It's a hard work..

And as usual, thanks for your interest...


Best,
Sedat
People who have lost the hope.
User avatar
Sedat Canbaz
Posts: 688
Joined: Mon Jun 19, 2023 8:49 am
Has thanked: 730 times
Been thanked: 2359 times

Re: Sedat Canbaz

Post by Sedat Canbaz »

massimilianogoi wrote: Tue Jul 25, 2023 9:25 pm Shocking difference! Could it be that the opening suite influenced the result?

Is this Raptora private? And the experience file?

Definitely intriguing test, I'd like to make it too, but with the same material.
No, RapTora 2.2 is publicly available...

Right not is too late (midnight) here..
But soon I will try to inform you with more details..

Greetings
Anton101
Posts: 73
Joined: Fri Jul 14, 2023 4:36 pm
Has thanked: 160 times
Been thanked: 206 times

Re: Sedat Canbaz

Post by Anton101 »

Here is the link to download...Eduard Nemeth helped me to compile: https://pixeldrain.com/u/JEPU1bDv

Best regards.

A.Ces.
User avatar
massimilianogoi
Site Admin
Posts: 394
Joined: Thu Aug 04, 2022 1:42 pm
Has thanked: 603 times
Been thanked: 647 times
Contact:

Re: Sedat Canbaz

Post by massimilianogoi »

Downloaded, thanks. Is this file also available?
Sedat Canbaz wrote:Eman exp (dated on 17.07.2023)
Anton101 wrote: Tue Jul 25, 2023 11:10 pm Here is the link to download...Eduard Nemeth helped me to compile: https://pixeldrain.com/u/JEPU1bDv

Best regards.

A.Ces.
People who have lost the hope.
Anton101
Posts: 73
Joined: Fri Jul 14, 2023 4:36 pm
Has thanked: 160 times
Been thanked: 206 times

Re: Sedat Canbaz

Post by Anton101 »

Well, I dont have that exp file...maybe by PM, Sedat can send it to you.
Post Reply