when a daily tester lies to people...

[deeds]

Between one who makes promises when he loses credibility at each tournament and the other who claims the paternity of engines whose source codes he does not know how to modify, we understand why they are in contact

With all the tools and knowledge shared for months/years in the area of ​​learning engines, a small minority still don't know the settings for training mode or tourney mode, still don't know how to get experience files from games.

Despite all their attempts to cling to the nextgen performance of learning engines, our greatest satisfaction is watching them fail again and again.

Finally I'm very happy not to read c/c++ because I avoid all its engine modifications and not to play online because I avoid all these people

[sarona]

To be clear: that Proteus experience file has very little Aurora data as I was/am training and saving files based on specific ECO codes - as opposed to one large one. I have individual files (A00, A01, A02, etc). He did not receive a large exp file from me. Just Aurora self play games with narrow opening lines. That experience file is most likely a merge of multiple files from the various chess fora trained with different engines with different evaluation/searches. I am not even sure if Igor was using one specific engine on the Playchess server as many of the various Proteus search sources were quite different. There were even mods to the PSQTs. It makes no sense to merge StockfishMZ, Aurora, Eman and other experience files together. That is pretty stupid.

I have been looking at Sedat's parameters specifically for book testing and a lot of what he is trying makes sense. Do not underestimate his approach when dealing solely with book testing. I am getting interesting results by not using NNUE in my preliminary tests with other polyglot books (Goi was mentioning this on Outskirts last week). This method is logical and seems to be an improvement over conventional approaches - so far.

[Sedat Canbaz]

Between one who makes promises when he loses credibility at each tournament and the other who claims the paternity of engines whose source codes he does not know how to modify, we understand why they are in contact

Hey Chameleon,

What are talking about..? what kind of promises?

Yes...that's true I am contact with Mr. Anton
But all our comments are publicly available...
Exception with RapTora's author, I'm contact also
With many other chess friends as well, so what?
Did I make mistake? correct Me please if I'm wrong..

Meanwhile, about some of the rest issues,
Sorry to say again, once more we noticed:
Really..you are far away to be as expert!
And you are not professional for sure...!

Do you know why? there are many reasons, e.g
You make BIG double standards to rest authors!!

And if you have a little bit honesty...then I suggest you:
Re-check, re-read all my tours...otherwise not fair..!!
Note also that Marco's engines are out of credibility!!
I assume you know why...if not M.Z does not deserve it!

Continuing...

Just in case of new credibility point stats calculations,
You can start checking e.g with these Championships:
Btw, also not so bad idea..to check all other my tours!
E.g which are as book tours, non-learning engines etc.
Just talking over exp files..would not be so much boring?

Anyhow, here are the ones, at least with Eng Learning:
Btw, as I stated before, Stability is the most part..so
Do not forget e.g in case of new credibility calculations
E.g Eman is not stable (via high number of concurrent..)

Eman 8.91: Winner of SCCT (2m+1s) NNUE Championship XXXIV

RapTora 2.2: Winner of SCCT (1m+0.6s) NNUE Championship XXXIII

RapTora 2.2: Winner of SCCT (1m+1s) NNUE Championship XXXII

Raubfisch X48e: Winner of SCCT (2m+1s) NNUE Championship XXI

Eman 6.91+SugaR AI 1.40: Winners of SCCT (3m+2s) Handicapped CS II

Eman 6.91: Winner of SCCT (3 min) NNUE Championship IV

By the way, (about Duel match: StockfishMZ vs ProteusSF)
What a joke)), you made BIG mistake: giving wrong points!
I mean e.g such as highest points as credibility: 92
Come on..1st. 400 games (per player) is not BIG data!
You are not aware of that?? error margin can be too high!!
Plus, bad idea...comparing only two players each other!!
Much better stats, if you check vs many players...!!
At least I can say: we noticed, StockfishMZ + Eman exp
How was buggy..in case of various test conditions...
If not clear, StockfishMZ lost many games in 22 moves!
Do you understand? and now...plus StockfishMZ lost..
And frankly I never seen a buggy one to get best points!

On other hand, there are better and stable chess engines,
Where you underestimated their work..well-done to you!!

And sorry to say again and again, so far
You are hopeless man..how many times you proved to be so!
If still not so clear:
Start to 'learn'...start to say 'sorry'...start to say 'thanks'
Start to be 'honest'..start to produce 'quality database' etc.
In other words,
Start to be a good human, but it's all in your hands..not in mine..
And please just do not get me wrong: I am trying to help you !

[Sedat Canbaz]

Deeds is not going to send anything, he just evades the challenges. It does not send me exp.cpp and modified exp.h... It does not send its experience file for a duel against books.

This living being is a waste of time.

Of course, Chameleon will not share both exp files.
Because, it will be clear his recent test is valid or not...

As you said, plus ZamChess, and I agree with both of you
A lot of noise, waste of time for nothing..sad but true..!

Greetings

[deeds]

...It makes no sense to merge StockfishMZ, Aurora, Eman and other experience files together. That is pretty stupid...

I like their "sample" list because people can clearly see how long Sedat hasn't understood that.

...I have been looking at Sedat's parameters specifically for book testing...

In their list, the customers told about learning engines with experience files with dangerous concurrent write accesses.
Nothing to do with Sedat's parameters for book testing with only safe concurrent read accesses.
That too he did not understand that.

[Sedat Canbaz]

Nothing to do with Sedat's parameters for book testing with only safe concurrent read accesses.
That too he did not understand that.

You can talk and talk...but
A lot of bla bla bla.. for nothing )

In other words,
You are not able to understand/learn that
I ran many concurrent matches plus on my
Ryzen7 4800H machine too..and we noticed
Even on my both machines, e.g the Learning
Engines are performed better..in next re-run..
Or via playing more games...sure I mean if we
Use to play small number of openings...e.g but
If using many various openings.. then very likely
We will need many more games (per player)
Sure depending on cond. but at least it's required
Thousands games...

[deeds]

It's amazing how different the view of learning engines is between daily testers and engine's trainers :



I doubt any engine's trainer will ever again waste time gauging where the learning engines stand in relation to the SF16...

Long live amateurs

[Sedat Canbaz]

It's amazing how different the view of learning engines is between daily testers and engine's trainers :

Hey Chameleon,

I see that you'll never stop..but no BIG problem )

On other hand, what we noticed also that,
You have 0 (zero) pride! and if I'm not wrong
You was banned at least 2 (two) times by Marco,
But again and again you preferred to give best
Points (92) to Marco's chess engine...hehe..LOL
Look..I knew that you are cheap man, but not so..

Btw, plus I'd not much surprise that,
You may sell anyone out for money!

And as I mentioned before:
I do not trust to your testings, your techniques are business
Techniques...You tried many times to fool the people with your
Bullshit testings...but now, the picture is perfectly clear...!
Otherwise, Depth4_Deeds book would be ranked much higher!

And as last,
I'd prefer to work daily, but via free and honest way!
And that's another main / BIG difference than you...!!

That's all

Bye, Bye )

[deeds]

When Sedat started with 100% credibility :

Hey DeeDs, look at this :


Sedat, the most incredible tester of learning engines !

But it's his choice to damage himself his credibility tourney after tourney by committing the same mistakes again and again. And still you are nice to have listed only few of samples otherwise his credibility would be even weaker. I think we can save our time/energy by ignoring the daily testers who have no knowledge/skill in learning engines.

[deeds]

...You was banned at least 2 (two) times by Marco...

No, at least banned 3 times but OCF's bans are useless because :
- I have no problem going at OCF when I want to get an OCF's file by myself
- From DeeDs to Chris via Unofficial I have always been in contact with MZ
- Several OCF's members can regularly send me specific OCF's files
- My methods are still present at OCF like
training's method on fishtest conditions
exchange about training's methods
training's method with books
etc.