SedatChess

[Sedat Canbaz]

Oh, ok!

I miss understood.

Great test!

When you come back from vacations, I'll send you a very updated RapTora with its own exp file.

Best regards.

Ok.. no problem )

Thank you..

See you later..

Greetings

[Sedat Canbaz]

When you come back from vacations, I'll send you a very updated RapTora with its own exp file.

Best regards.

Dear Mr. Anton,

Well, my previous posting was during late midnight..
So now I will try to be much more clear... )

About sending a new RapTora engine vers,
Sure if you are referring about B/E/N CS:
https://sites.google.com/site/computers ... t-ben-cs-1

Then, I will prefer testing only your exp file/s, the reason
In this championship, my target to test those engines, which
Are very close (in strength) to SF-PB 060623 chess engine!

Note also that,
If we will allow stronger engines or weaker etc. then simply
Will be not so clear, due to Engine strength or Exp strength!

What I can say more,
Very likely as main engine (for new exp files) will be
RapTora 2.2, plus BrainLearn 25.1, LittleBeast 2.2 sl
I mean, Eman 8.91 will be not used if new Eman exp entries
As reason (as I stated before) Eman 8.91 is not stable in
Gauntlet mode ---> Concurrent games: 64

Be sure also that,
I'm interested much in your newer eng versions,
But in other newer competitions...

Thanks for your interest and understanding


Best,
Sedat

[Anton101]

All right, I'll send you a new exp file in a few hours by email.

Best regards.

[Sedat Canbaz]

All right, I'll send you a new exp file in a few hours by email.

Best regards.

Good.. but we have plenty of free time,
I mean better in September...I will inform later,
When the tour is resumed...

Greetings

[Sedat Canbaz]

If you think that Eman "learned" something after a few 100 game matches, with parallel concurrent matches, and different openings you are quite wrong, believe me...

Eman usually needs 500 games or more per single opening line to learn something and this is something that even the Eman's author said, it is not something I am inventing.

Regards,

Ibai

Hello Ibai,

How are you ?) and I hope you are well...

By the way,
What is your opinion about my latest testings:
RapTora 2.2 vs SF-PB 060623 ?

E.g here (in 2nd, especially in 3rd test) we noticed that
RapTora 2.2 Eduard is even more impressive (than Eman 8.91)
And even via Concurrent games: 10

Do you STILL believe that e.g under SCCT conditions:
Learning Engines (in Concurrent) are not as VALID test?
If not so clear,
I am simply referring about RapTora 2.2 Eduard

Or maybe this is simply due to error margin..?

Thanks in advance

Best,
Sedat

[Anton101]

Ok...September will be...

Regards.

[IbaiBuR]

Hello Ibai,

How are you ?) and I hope you are well...

By the way,
What is your opinion about my latest testings:
RapTora 2.2 vs SF-PB 060623 ?

E.g here (in 2nd, especially in 3rd test) we noticed that
RapTora 2.2 Eduard is even more impressive (than Eman 8.91)
And even via Concurrent games: 10

Do you STILL believe that e.g under SCCT conditions:
Learning Engines (in Concurrent) are not as VALID test?
If not so clear,
I am simply referring about RapTora 2.2 Eduard

Or maybe this is simply due to error margin..?

Thanks in advance

Best,
Sedat

Eveyrthing fine here, thanks

Just checked the test....

Some friendly recommendations / advices from my side:

You should not use an experience file trained by other engine's data with Raptora, since the experience data is attached to the style / search of the engine it is training it, even if you have got a succesful result (still, very low number of games as you may know) it is not advisable really.

I also still maintain that with such a low number of games as we both said, the engine has probably learnt anything, needs more games, I hope you agree with me on that too

Also, to test experience files, it is normally important to tell the users the experienceRate of each opening in the suite, which can be obtained via this tool: https://github.com/chris13300/expRate, because if the experience file has a low rate on the opening it and you run the engine in training mode (e.g, default settings), the engine will try unusual / different moves. This is just a recommendation too.

This is similar to the book settings the book authors recommend for their books basically

Also, RapTora 2.2 and SF-PB are not the same engine so we dont know how much influence that had compared to the experience file, probably, since you say that the experience file is not so big I bet it did not have much influence in the results

This is all I can say about it, thanks for inviting me to discuss in a friendly way again, much better than before. Hope you are fine too. If you have any questions regarding my opinion just tell me.

Regards and greetings,

Ibai

[Sedat Canbaz]

Eveyrthing fine here, thanks
Just checked the test....
Some friendly recommendations / advices from my side:
You should not use an experience file trained by other engine's data with Raptora, since the experience data is
.....

Thanks for your replay, comments...

Hmm...as far I noticed,
There are some confusions between each other... sad..
At least it seems to be so..or.. and no much idea why)

However and once more to be more clear,
Some days ago.. you commented something like that
SCCT is not 'valid' because I run Concurrent matches

Now even after both testings, that means 700 games
Do you STILL believe that Learning engines are NOT
Capable to gain, learn (under Concurrent) etc. ?

About your friendly suggestion..
Thank you, but I already stated that
I plan to test new Eman exp entries via Raptora 2.2
If you missed,
Not so good idea if I will run via Concurrent games: 32
Because my tournament machine is 2x EPYC 7B12, which
Offers us: 256 Threads / 128 Cores

By the way, I spent so many EURO
Just for Chess FANS.. to produce quality games in shorter time!')
Sure they are NOT as best.. but at least we see some FUN for sure )

On other hand,
I respect you as good programmer, but as a tester
I have different view than you... SORRY...
And once more...and as I mentioned before,
I relay on testings, than comments...!

Continuing...

Sure, for a better conclusion, we need more games..
Regarding the real Elo differences...but anyhow
After checking the latest results, it's already clear that,
RapTora gains/learn much..even via Concurrent matches!
Actually this is already proved in many SCCT tours!
So talking more over these issues will be meaningless!
I just wanted to ask your opinion.. ) but it seems hard
To believe...but no BIG problem...I've done my best!)
Even on my Holiday days.. )

But on other hand, this is also true that,
SCCT B/E/N Championship is not as World Championship!
I mean the played results should not take too seriously...
Actually, we all learn at least many of us...right ? or is there
Anyone here who knows everything?) if so who will be 1st Champ?)

That's all for now...)

Greetings

[IbaiBuR]

However and once more to be more clear,
Some days ago.. you commented something like that
SCCT is not 'valid' because I run Concurrent matches

Now even after both testings, that means 700 games
Do you STILL believe that Learning engines are NOT
Capable to gain, learn (under Concurrent) etc. ?

About your friendly suggestion..
Thank you, but I already stated that
I plan to test new Eman exp entries via Raptora 2.2
If you missed,
Not so good idea if I will run via Concurrent games: 32
Because my tournament machine is 2x EPYC 7B12, which
Offers us: 256 Threads / 128 Cores

On other hand,
I respect you as good programmer, but as a tester
I have different view than you... SORRY...
And once more...and as I mentioned before,
I relay on testings, than comments...!

Continuing...

Sure, for a better conclusion, we need more games..
Regarding the real Elo differences...but anyhow
After checking the latest results, it's already clear that,
RapTora gains/learn much..even via Concurrent matches!
Actually this is already proved in many SCCT tours!
So talking more over these issues will be meaningless!
I just wanted to ask your opinion.. ) but it seems hard
To believe...but no BIG problem...I've done my best!)
Even on my Holiday days.. )

Greetings

It is sad that you are going to test Eman experience files with RapTora, it doesnt make much sense but if you want to, it is your choice basically...

RapTora may be able to learn "something" via concurrent matches, but, as I told you before, probably a lot of experience data is lost sadly because there are too much concurrent reads / updates in a single experience file...

Since you are right that it is better to rely on comments than on testings, when I have some free time I will try to support what I say with tests too...

I also wonder, why dont you rely on my settings / advices? Because when bookmakers give you settings for their books you dont ignore them? I am trying my best also to give you best settings possible...

Regards,

Ibai

[Sedat Canbaz]

Since you are right that it is better to relay on comments than on testings,

What a pity that
The confusions (between each other) are continuing...

If not so clear,
Really ? are you sure ? I said just opposite..and not once...
Many many times... if still not CLEAR:

I prefer to relay on TESTINGS (than comments)

And I wished to see your replay something like that:
SCCT is valid or NOT valid...

I mean in very simple: YES or NO ??))

Ok.. that all )

Best,
Sedat