when a daily tester lies to people...

General discussion about computer chess...
ConRog
Posts: 32
Joined: Thu Jun 15, 2023 6:12 am
Real Name: Roger Constantinesco
Location: Tahiti, French Polynesia

Re: when a daily tester lies to people...

Post by ConRog » Wed Aug 16, 2023 8:36 pm

Hi,

I confirm : my post was in " LesChats110423" topic in another forum, and referring to a tournament organized by Goi with non NNUE engine, and bullet games 1+0.5.

SCCT tournaments are tuned to create good games in all conditions, even with non NNUE engines.

Best regards.

User avatar
deeds
Posts: 1008
Joined: Wed Oct 20, 2021 9:24 pm
Location: France
Contact:

Re: when a daily tester lies to people...

Post by deeds » Wed Aug 16, 2023 8:44 pm

deeds wrote:
Tue Aug 15, 2023 9:01 pm
...i wonder how many days/weeks it will take to someone to find it and maybe tell it to Sedat :lol: :lol: :lol:
Already 1 day :!:

User avatar
deeds
Posts: 1008
Joined: Wed Oct 20, 2021 9:24 pm
Location: France
Contact:

Re: when a daily tester lies to people...

Post by deeds » Wed Aug 16, 2023 9:03 pm

Too late, Conrog ! Sedat already understood :
Sedat wrote: ...But I admit also that (when NNUE as enabled)
Some of my tours are Blitz + MP + Ponder ON
Where in most cases, their draw-ratio was 99%
The old era of static chess with opening books is dying since people train learning engines. In computer chess era, it became absurd to run engines from starting position. Fishtest uses opening lists, Stockfish doesn't handle books, we have tools to reveal openings handled or not by books, it's the end.

Sedat Canbaz
Posts: 1978
Joined: Wed Jun 21, 2023 6:29 am

Re: when a daily tester lies to people...

Post by Sedat Canbaz » Thu Aug 17, 2023 1:06 am

ConRog wrote:
Wed Aug 16, 2023 8:36 pm
Hi,

I confirm : my post was in " LesChats110423" topic in another forum, and referring to a tournament organized by Goi with non NNUE engine, and bullet games 1+0.5.

SCCT tournaments are tuned to create good games in all conditions, even with non NNUE engines.

Best regards.
Many thanks for your valuable comments!

Greetings

Sedat Canbaz
Posts: 1978
Joined: Wed Jun 21, 2023 6:29 am

Re: when a daily tester lies to people...

Post by Sedat Canbaz » Thu Aug 17, 2023 1:36 am

deeds wrote:
Wed Aug 16, 2023 9:03 pm
Too late, Conrog ! Sedat already understood :
Sedat wrote: ...But I admit also that (when NNUE as enabled)
Some of my tours are Blitz + MP + Ponder ON
Where in most cases, their draw-ratio was 99%
The old era of static chess with opening books is dying since people train learning engines. In computer chess era, it became absurd to run engines from starting position. Fishtest uses opening lists, Stockfish doesn't handle books, we have tools to reveal openings handled or not by books, it's the end.
Hello Chameleon )

Why too late ?? I think that
Conrog is made as important statement, plus just on time!

By the way,
What kind of personalized BRAIN do you have? ))
Really your have BIG problems to understand, learn etc.!

If not so clear, I admit that some SCCT 3700+ NN tours are
Played on fast hardware as Blitz, MP, Ponder ON etc.
But I was quite disappointed and I hardly doubt to run again!
As main reason, too BORING chess !! over than 99% draws!

On other hand,
Very likely again..you will have no clue but...maybe 0.0001%
You may have a chance to learn...in short, time will tell )))

And again and again, Deeds does not learn..
E.g here is another prove... but not once, MANY TIMES!
Because I stated many times that almost all my ratings are
Played as Hyper-Threading (SMT) as disabled...so OOPS ))

One thing more (about Marco's comment such as 1 Thread..)
It seems that M.Z was not referring SCCT tours, because
Almost all SCCT ratings are NOT played as 1 Thread, e.g
SCCT's conditions are mainly as MP or via 1 Core...
If still not so clear: Do Not Mix Apples with Oranges!

Meanwhile, OPEN your eyes, because
Stockfish doesn't handle learning too!!

And as last words,
If you believe that the opening theory is dead...if so...
Why the opening books performed better than exp files?

Code: Select all

The Current Leader: HCanFree Upd -  Congrats to Hamdullah Can!

 Rank Name                  Elo    +    - games score oppo. draws 
   1 HCanFree Upd          3789    2    3   600   52%  3781   96% 
   2 V by Kerveros         3788    2    3   600   51%  3781   97% 
   3 CachoSCCT             3787    3    3   600   51%  3781   96% 
   4 ChessBot 170623       3787    3    3   600   51%  3781   95% 
   5 OPTIMUS 3.1           3787    3    3   600   51%  3781   97% 
   6 Hamish 4.1            3787    3    3   600   51%  3781   96% 
   7 SkynetSCCT 10.1       3785    3    3   600   51%  3782   96% 
   8 Leschats 140623       3783    3    3   600   50%  3782   98% 
   9 NoPreExp, BrainLearn  3782    3    3   600   50%  3782   99% 
  10 AlexChess, RaptoRa    3782    3    3   600   50%  3782   98% 
  11 Ding-Bat, RaptoRa     3782    3    3   600   50%  3782   97% 
  12 Anton, RaptoRa        3781    3    3   600   50%  3782   98% 
  13 Rulleeeee, Eman       3780    3    3   600   50%  3782   99% 
  14 NoPreExp, L.Beast     3780    3    3   600   50%  3782   97% 
  15 Cubail, Eman          3779    3    3   600   50%  3782   97% 
  16 NoBook, SF-PB.        3778    3    3   600   49%  3782   97% 
  17 NoPreExp, Eman        3778    3    3   600   49%  3782   97% 
  18 NoPreExp, RapTora     3778    3    3   600   49%  3782   98% 
  19 Hurnavich, L.Beast    3775    3    2   600   48%  3782   96% 
  20 Africaata, Eman       3774    3    2   600   48%  3782   96% 
  21 Khalid, Eman          3773    2    2   600   48%  3782   95% 

• Calculated by Bayeselo / SkynetSCCT fixed to 3785 Elo 
As another REMİNDER:
If you again will NOT learn: please STOP your NONSENSE !!

Sedat Canbaz
Posts: 1978
Joined: Wed Jun 21, 2023 6:29 am

Re: when a daily tester lies to people...

Post by Sedat Canbaz » Thu Aug 17, 2023 1:51 am

deeds wrote:
Wed Aug 16, 2023 8:44 pm
deeds wrote:
Tue Aug 15, 2023 9:01 pm
...i wonder how many days/weeks it will take to someone to find it and maybe tell it to Sedat :lol: :lol: :lol:
Already 1 day :!:
No comments, exception:
You need a good doctor!

User avatar
deeds
Posts: 1008
Joined: Wed Oct 20, 2021 9:24 pm
Location: France
Contact:

Re: when a daily tester lies to people...

Post by deeds » Thu Aug 17, 2023 5:41 am

book's creator #3 wrote: "yes like many (2015)... there are too many draws, no interest from startpos in computer chess era..."
Sedat just confirmed it so you were right, thanks again !

As Sarona did at Chess2U, now is the time to ignore the AlexChess team and his sponsors (Sedat, Anton101, Homayun, ZamChess, Conrog, etc).

Our recent contacts among the creators of opening books will bring us new sparring-partners for learning engines. In return, i offered tools, new v3.x file formats. I saw many book's creators already fixed their books attached to my "exp vs book" training games, good point.

We will still have a lot of fun seeing the Sedat-like fidgeting, not understanding the advice, ignoring the evidence, remaining unanswered, they sabotaged their results in 2022, I bet it will be the same in 2023...

That's all for me, good luck !

Sedat Canbaz
Posts: 1978
Joined: Wed Jun 21, 2023 6:29 am

Re: when a daily tester lies to people...

Post by Sedat Canbaz » Thu Aug 17, 2023 8:56 am

deeds wrote:
Thu Aug 17, 2023 5:41 am
book's creator #3 wrote: "yes like many (2015)... there are too many draws, no interest from startpos in computer chess era..."
Sedat just confirmed it so you were right, thanks again !

As Sarona did at Chess2U, now is the time to ignore the AlexChess team and his sponsors (Sedat, Anton101, Homayun, ZamChess, Conrog, etc).

Our recent contacts among the creators of opening books will bring us new sparring-partners for learning engines. In return, i offered tools, new v3.x file formats. I saw many book's creators already fixed their books attached to my "exp vs book" training games, good point.

We will still have a lot of fun seeing the Sedat-like fidgeting, not understanding the advice, ignoring the evidence, remaining unanswered, they sabotaged their results in 2022, I bet it will be the same in 2023...

That's all for me, good luck !
Hehe.. do you give up really ?? )))
It looks like I managed to break down Mark Twain's saying
“Never argue with an idiot. They will drag you down to their level and beat you with experience.”

And as we see,
Deeds can not beat... with his experience (Depth4 exp) LOL )))
Sure there is a possibility., but before..a new system is required..
E.g Depth4 exp to be (as player) in serious tourneys, ratings ..!!

Continuing....

Give my regards to your 'book's creator #3'
But I totally disagree with both of you!
Actually I hardly doubt that there is such:
book's creator #3 = very likely a fake name))
But let's assume that there is such ...then I say:
In 2015 year: the situation was not same as
Nowadays, e.g 2015's engines were much weaker...

Meanwhile, according to SCCT rating's 2015 results:
https://sites.google.com/site/computerschess/scct-3m-1s
The best Top engines were minus plus 3300-3360 Elo
That means in same time if comparing 2015 vs 2023:
https://sites.google.com/site/computers ... t-nn-cs-32
We see about 400 Elo difference !!

In same time,
2015's produced draw-ratio was much lower than
The latest Top engines of 2023, one of reasons:
Many engines are already capable to gain from
Previous played data (Evalfiles, Exp files etc)
So no BIG surprise that we see higher draw perc.

But this is also true and very sad news is that,
The guys (like Deeds) will never learn and see it!
These kind of BRAINS will accuse / bombard those
Testers, when Eman exp files will be not as Winner!
This is already as 100% fact..proved so many times!!

Regarding the question from which engine/book team I am...
Frankly, over the long past years, I received many many times
Invitations to join as main book author (including TCEC etc.)
Or to be Operator (via my hardwares) in serious online tours..plus
I received invitations from CEGT, CCRL etc. to join in their teams!
Where unfortunately I did not accept all their invitations, due to
As usually, I've no much free time for all..SORRY..in other words:
I am just as TD of my own organized competitions (SCCT), sure
In same time, I am author of Perfect, Balsa and Unique books!
Plus I have many other chess activities..but no free time to
Mention all, but there is one true: too much for sure...!!

As last words, I assume also that in the next months/years:
Book vs Exp matches will be exiting and will bring more fun!
If nothing else, I hope so...

Thanks for reading!

ConRog
Posts: 32
Joined: Thu Jun 15, 2023 6:12 am
Real Name: Roger Constantinesco
Location: Tahiti, French Polynesia

Re: when a daily tester lies to people...

Post by ConRog » Thu Aug 17, 2023 10:07 am

deeds wrote:
Wed Aug 16, 2023 9:03 pm

The old era of static chess with opening books is dying since people train learning engines. In computer chess era, it became absurd to run engines from starting position. Fishtest uses opening lists, Stockfish doesn't handle books, we have tools to reveal openings handled or not by books, it's the end.
Hi deeds,

I don't think the era of engines, plying with well tuned opening books, is dying.

The goal of opening books is to go fast, and without loosing any microsecond, on maximum depth and on the best position, so that the engine will keep all remaining time to play maximum strength VS opponent which will have little time left because of losing precious seconds before the end of book line.

Good books will always beat a learning engine at all time controls.

Best regards.

User avatar
deeds
Posts: 1008
Joined: Wed Oct 20, 2021 9:24 pm
Location: France
Contact:

Re: when a daily tester lies to people...

Post by deeds » Thu Aug 17, 2023 10:56 am

ConRog wrote:
Thu Aug 17, 2023 10:07 am
I don't think the era of engines, plying with well tuned opening books, is dying.
I'm not surprised if most book's authors still believe it too :arrow: no one want to believe that his daily work is a dying activity, especially those who have been doing this for decades. But hey, it's not as if it wasn't obvious and as if we hadn't warned them.

ConRog wrote:
Thu Aug 17, 2023 10:07 am
The goal of opening books is to go fast, and without loosing any microsecond, on maximum depth and on the best position, so that the engine will keep all remaining time to play maximum strength VS opponent which will have little time left because of losing precious seconds before the end of book line.
This is exactly what the other book's authors have explained to me :arrow: which is why, now, they want to collaborate with engine's trainers.
With neural networks, the engines guess the same moves as the book's moves (100% ponder) so when the book is out-of-book :arrow: the time saved by its engine is already useless :arrow: this technique leads to 99% draws :arrow: => it's a technique dying.

ConRog wrote:
Thu Aug 17, 2023 10:07 am
Good books will always beat a learning engine at all time controls.
Even under stupid conditions or sabotaged results, SCCT showed the opposite and several times :!:
By reinforcing experience data, we already showed how a learning engine can resist to odd tc (=opponent with 4x, 5x and even 6x longer TC).
When a learning engine wins more games than an engine with an opening book, it is absurd to believe that the opening book was not good or that the TC was not favorable.
As Ibaibur explained, those who still believe this have a complete misunderstanding of the performance of experience data. Nothing surprising because this lack of knowledge is carefully maintained by daily testers.

Post Reply