Chess Forum

I'm just a member like the others.

Anyway, I think I've seen a paradox: you can't have both a strong engine and a low draw ratio. Draw games happen when the engines are strong and chose the best moves leading to a draw even when the score is ~|0.60|. When an engine is weaker there is a higher possibility that worst moves are chosen and that the final result is a loss/win, hence the draw ratio is smaller.

Sedat Canbaz wrote: ↑Sun Jul 16, 2023 7:16 pm

massimilianogoi wrote: ↑Sun Jul 16, 2023 5:20 pm First time I see you making a head to head match book vs book (but I could be wrong).

I borrowed your ideas for a tournament (posted on the Outskirts Forum too): 1 single thread and NNUE disabled. I also added very short time. I'm sure there will be engine errors, but in this way it's very easy to discover book weaknesses: with all threads, NNUE enabled and long time, the engine would have corrected the weakness and bring the game to a draw.

I like it .. nice to see various book tours!

Interesting to point out that (in your tour):
3 Books (LesChats, DON, Bloom) are leading...
And it seems all books are very strong players...

On other hand, after checking your played games..
And as you say...if your tour is NNUE disabled,
I can say that 97% draw-ratio is very high!

E.g if we use Raubfish x48c SC even as NNUE is enabled,
Less draws may appear...I mean than your current tour's draws

The main difference,
You will get better games in quality!

Why I say like this...
E.g if I am not wrong, via Bullet/Blitz 1 Core:
Between Raubfish SC NNUE vs Cfish 030820 NON-NNUE,
There is more than 100+ Elo

What I can say more, If you wish to test as NNUE is disabled,
I suggest Cfish 030820 to be used...sure via Contempt 40
Not sure exactly, but less than 90% draws are expecting...

And if you wish to test all books as NNUE is Enabled
Then I suggest Raubfish SC NNUE to be used...-

Sure after all, it is all up to you )
You are the BOSS )

Greetings

massimilianogoi wrote: ↑Sun Jul 16, 2023 8:56 pm I'm just a member like the others.

Anyway, I think I've seen a paradox: you can't have both a strong engine and a low draw ratio. Draw games happen when the engines are strong and chose the best moves leading to a draw even when the score is ~|0.60|. When an engine is weaker there is a higher possibility that worst moves are chosen and that the final result is a loss/win, hence the draw ratio is smaller.

Yes..you have right...
And that's true...draw-ratio depends on strength!

But anyhow,
Cfish 030820 is one of the strongest out there,
Sure If NON-NNUE, at least my tours confirm this...

Well, and just to be more clear,

Sure I am talking for my side:
Mostly my tours are organized mainly for FUN

If my target is to produce very high-quality games,
Then I'd prefer 40/120 and 128 Cores (per engine)

But just in that case,
I expect the overall draws to be 99.9999999%
And very likely, I will lost my interest...
Do you know why ?)
If the results will be equal, then what is the point?

Regarding 97% draws as 1 Core, 1n+0.5s NON-NNUE etc,
I can't say very high-quality, but they are not bad..
And now why we have to use close engine in strength
But which is more DRAWISH...?)

I'd prefer Cfish 030820 C40 NON-NNUE,
Where it is expecting to see less draws..
Because I stated before,
FUN is one of the most important parts!

Sure this is my view...
And I respect any other testings as well

Btw, if we were all same,
The world would not be so boring ?)

One thing more,
Regarding the latest Top NNUE engines and testings etc.
Engines prefer to use data on previous played games!
So the overall draws are going higher and higher...
Sure in same time, tge quality as well...I admit..

Meanwhile,
Just imagine, Humans are started to check, gain from
Previous played games etc... then
What will be our rating and our overall draw-ratio?
Definitely we will increase our quality of game play.. plus
Our Elo rating will be increased much higher as well..
Of course I expect that our draw ratio will go much higher...
Btw, just in case ...not suggested to use fast time controls ))
Because we may miss to pick/choose the right moves !)

Greetings

One more suggestion, (in case of tours such as 1 core)
To be played our tour games: via parallel matches...
E.g in this system way of testing,
We will produce more games in shorter time...!
In same time that means,
We will save our electric bills as well !

Anyhow, this is just my opinion and recommendations..
And anyone is free to use any conditions as he/she wish..

Greetings

Meanwhile, what I can add more,

For less error-margin,
- I recommend more than 500-600 games (per player)
- Playing vs many various opponents, min 20 players

But in case of direct matches (if only two players),
200 games can be enough.. sure more better, but
This time may appear many more doubles... plus
Running so many games each other is not required much..
Reason.. usually the Top books prefer to play close styles
Trend openings...and final results will be not changed much!
Sure no any guarantee..but according to my overall tests,
In most cases, the played tour statistics confirm this...

Btw, it is not so bad idea (for more fair Duel coınditions)
The event: BIN vs CTG to be played as Book Learning OFF

About Time Controls to be used,
If NON-NNUE: at least Blitz e.g such as 2m+1s, Ponder ON
If NNUE enabled: even Bullet can be used, Ponder OFF etc.

But again and again,
It all depends on main engine which will be used...

That's all for now... )

Greetings

Hello there,

After so many comments about Eman (on OpenChess forum)
I realized to run several quick Eman 8.91 testings...
Sure as main target to see the learning influence..
E.g Eman's and GUI concurrency option as 1 (one)
Plus Eman's and GUI concurrency option as 8 (eight)
And Eman's Experience MultiPV is Disabled / Enabled

1st Test: Ruleeee exp + Eman 8.91 vs SF-PB 060623
Note: Cutechess GUI's concurrent games: 1

2nd Test: No Pre Learned Exp + Eman 8.91 vs SF-PB 060623
Note: Cutechess GUI's concurrent games: 8

3rd Test: Pre Learned Exp + Eman 8.91 vs SF-PB 060623
Some Notes about 3rd test:
Cutechess GUI's concurrent games: 8
Eman's Experience MultiPV is Disabled
Pre Learned Exp is the same one of 2nd test...

4rd Test: Pre Learned Exp + Eman 8.91 vs SF-PB 060623
Some Notes about 4rd test:
Cutechess GUI's concurrent games: 8
Eman's Experience MultiPV is Enabled
Pre Learned Exp is the same one of 2nd + 3rd test...

Both (1st+ 2nd) testings are ended...
Right now the 3rd test is running..

Sure I hope to publish all results via more details very soon

And stay tuned please.. )

Greetings

never would have occurred to me that sedat would test my experience file

Halcyon wrote: ↑Mon Jul 17, 2023 10:58 am never would have occurred to me that sedat would test my experience file

ruleee is your exp file name ?

Yes. I uploaded this deeds file so he could test it with his experience file in three openings

Halcyon wrote: ↑Mon Jul 17, 2023 11:23 am Yes. I uploaded this deeds file so he could test it with his experience file in three openings

Hello halcyon

Great...

I used it because, rulleeeee exp is performed not so bad so far:
https://sites.google.com/site/computers ... t-ben-cs-1

13th place (unlucky);p

but at least i'm not the last :P

I can send you an updated .exp file in two months