SedatChess

[Sedat Canbaz]

Many thanks for the tour, I'll be working next three weeks in the engines, in my test CorChess and BrainLearn wasn't the best...but it's ok.

Congrats to the director tour, see you next week, let's see what's new.

Best regards.

You are welcome )

A good question, comment...
Regarding BrainLearn and Corchess etc. performances...

Well, as I stated earlier, it all depends on our test cond.
Sure for better conclusion..a lot of games are needed..
At least via latest SCCT tours, we have as serious data..

So..here is my view over these issues e.g just a few examples,
Which may play as important roles (during our engine testings):
-Some engines are doing better vs stronger/weaker opponents..
-Or some engines perform better via weaker/stronger books
-Or some are better at Bullet, or some better at Blitz etc.
-Even MP and Ponder ON or OFF can be as serious influences...
-Sure our hardware speed differences, hashtable sizes etc. can be
As other important influence about why BrainLearn is better etc.

Forget everything,
Pre-Learned EXP files can be as other important factors etc.
E.g in case of switching EXP from Bullet + 1 Core to Blitz + MP

If not so clear, please see Raid 2.73 and Stockfish 220823 sl:
https://sites.google.com/site/computers ... t-nn-cs-39

But as other test, where next time as without Pre-Learned EXP:
https://sites.google.com/site/computers ... t-nn-cs-40

So if comparing both tour results, we will notice that
Pre-Learned exp files may play as advantage or disadvantage etc!

Greetings )

[Sedat Canbaz]

Update (mainly for Super League):

Why I realized both Leagues to be over for several reasons: 1st. it's clear that
New entries have no BIG chances to perform better than those opponents, which
Are already playing via pre-learned EXP files.. 2nd. no much Improvements, here
Very likely the learned EXP files played some role over NON-Learning engines
On other hand, the engine 'Learning' is nothing bad at all.. .at least, in critical,
Weak opening positions, the EXP files may play as important influences / role...
Otherwise, I hardly doubt that about these beasts learning engines would gain
Better points than the ones, which are not capable to learn in game play mode
3rd. giving 5m+1s is too much for running under these cond..it takes too much
Time, 2m+1s seems to be better..especially if no Elo progress in latest months..
Anyhow, this time BranLearn 26 is said the last word...2nd is Raid 2.73, where
3rd rank belongs to the latest entries: Tactical and DarkSister..very good really..
Ok.. that's all for now.. and I hope to see all of you plus new incıomers as well!

Super League's Champion: BrainLearn 26 - Congrats to K. Kiniama, A. Manzo, SF team!

Super League's Final Rankings:
https://sites.google.com/site/computers ... n-42-super

Games:
https://mega.nz/file/f8h0jCrR#QChgPHjS- ... DtFXET3Tko

Best,
Sedat

ahh there's my prediction

Thank you for the tournament Mr Sedat . i see Awol Z10 is capable of handling learning engines .
Honestly didn't expect Raid to be in the Top 3, being old as he is, am touched .

Congratulations to manzo and ivec

Not at all, it is my pleasure Mr. Tanick )

Agreed..Awol Z10 (as NON-learning engine) is a very serious opponent!
But in my guess..if adding learning future then would perform better..
You know, it all depends.. but as usual, the EXP files really helps a lot ...

About Raid 2.73 performance,
To be 2nd..very very good plus in some older tours, the same Raid vers
Managed to be 'DOUBLE' times as Champion...so no BIG surprise !!

Greetings )

[OrgZ]

I never thought of giving him a learning feature bcz i designed him specifically for analysis. if you would like me to add i could do that its no prob

[Sedat Canbaz]

I never thought of giving him a learning feature bcz i designed him specifically for analysis. if you would like me to add i could do that its no prob

Thanks a lot ...

Sure... I wish... in this way,
We may compare the influences...


Greetings

[OrgZ]

I never thought of giving him a learning feature bcz i designed him specifically for analysis. if you would like me to add i could do that its no prob

Thanks a lot ...

Sure... I wish... in this way,
We may compare the influences...


Greetings

mail

[Sedat Canbaz]

I never thought of giving him a learning feature bcz i designed him specifically for analysis. if you would like me to add i could do that its no prob

Thanks a lot ...

Sure... I wish... in this way,
We may compare the influences...


Greetings

mail

[Sedat Canbaz]

Breaking News (01.11.2023):
Another new CS series, where this time the target will be to test mainly more
New 3700+ engines, sure depending on my free time I may test some oldies..
The tour is active...very soon more updates..Btw, no pre-learned exp is used

Leaders: AWOL+DarkSisTer - Congratulations to T.Ramz, A.Ces, SF team!

For Latest Standings:
https://sites.google.com/site/computers ... t-nn-cs-44

Greetings )

[Sedat Canbaz]

Meanwhile, a small note:
AWOL Z11 is a new version, which is capable to learn,
Where the previous (AWOL Z10) was not able to learn...

[Sedat Canbaz]

Update

Via latest entries, standings changed where Tactical managed to take the lead..!!
Meanwhile, any known author, who is interested...can submit his new engine..
And just in case of interest..please do not be too late...thank you in advance

The Current Leader Tactical 281023 - My Congratulations to Anton Ces, SF team!

For Latest Rankings:
https://sites.google.com/site/computers ... t-nn-cs-44

More Details
• Mostly of NNUE engines are AVX2 and tested via theirs default Evalfiles
• All engine matches are played via CuteChess GUI concurrent games 64
• Sure where possible: all of engines are played via move overhead: 100
• In this new active Championship: max. up to 76% (in similarity) allowed
• All of matches are played via new Unique suite (based on 372 openings)
For more fair conditions, each of used opening line is repeated as twice
• Current overall draw percentage (based on 6000 games) is normal: 89%
Btw, via weaker openings I may reduce even less than 50%...just in case
Then we will need many thousands of games (per player), otherwise Elo
Error margin can be high..simply will be as lottery to see which is stronger
• Each participant will contain min. 750 games... but the high-rated 1000+
• Any Engine Leader after 1000 games (per player) will be titled Champion
• As usually, all the played games will be available after the end of tourney

Greetings

[Sedat Canbaz]

One thing more, with very low draw perc. no way to see wins by both sides
E.g via Unique openings, each line includes wins by White + Black of 3700+

Btw, if not today, tomorrow will be clear the 1st Champion...

Good luck to all )