SedatChess

[Amos 4ever]

If the reader already knows it is enabled by default in the code, then I think the user can also suspect that I am talking about the OS part, right? It is logical! But ok...

In any case Lp does not affect the result but only increases the speed, but as you know with NNUE it has no influence
Previously gave a little help with engines without NNUE

[IbaiBuR]

If the reader already knows it is enabled by default in the code, then I think the user can also suspect that I am talking about the OS part, right? It is logical! But ok...

In any case Lp does not affect the result but only increases the speed, but as you know with NNUE it has no influence
Previously gave a little help with engines without NNUE

Yeah, I know, but for experience files for example it shold increase performance a bit too since allocating LargePage memory for the experience data will speed up I/O operations during training for example. Or am I wrong here?

[Sedat Canbaz]

This test is with Large Pages enabled or not?

A good question

Well, current Hash test plus almost all my tours over the past years are
Played with Large Pages as Disabled (directly not allowed from Windows)

And if you are asking Me why (or for anyone interested, missed...),
Normally LP as enabled may bring a good speed up for some chess engines,
Especially in analyzing mode, but in game-play mode... be careful please!
And If you are overall engine tester...and in case of long-running matches,
Then LP as enabled may down the hardware speed... use it in your own risk!
For more details about LP issue:
https://learn.microsoft.com/en-us/windo ... ge-support

Btw, in same time, all my tours are played as
Hyper Threading / SMT is Disabled as well

Greetings

[Amos 4ever]

If the reader already knows it is enabled by default in the code, then I think the user can also suspect that I am talking about the OS part, right? It is logical! But ok...

In any case Lp does not affect the result but only increases the speed, but as you know with NNUE it has no influence
Previously gave a little help with engines without NNUE

Yeah, I know, but for experience files for example it shold increase performance a bit too since allocating LargePage memory for the experience data will speed up I/O operations during training for example. Or am I wrong here?

Yes you are right even if the risk with a very large set of games (above 1000) defragments the memory too much because the physical space for each large page must be contiguous. Allocating large pages under these conditions can significantly affect system performance.

[Sedat Canbaz]

Hello Chess Friends,


New similarity data based mainly on latest Top engines:
https://sites.google.com/site/computers ... -july-2023

Actually, this is known that many of Top engines are
Very close to each other, especially which are 3700+

Note that these values are generated with time 100 ms
Just imagine, e.g 10 Sec (per move), then very likely
Their percentages would be much close to each other!

On other hand, what is interesting to point out that,
Usually SF based engines have almost same bug/s!
E.g I am referring about the latest CTG buggy code,
If not so clear, you may see what is gong on: Details

As another shocking news is that,
SF PB 190623 vs Polyfish 210623 has a similarity: 99%
It seems that both engines are based on same source!
And this is other question of course, anyhow as a tester,
In my next tours, I wonder which one to test...you know
Testing almost same identical ones will be meaningless!

By the way, I've tested RapTora 2.2 by Antonio Ces plus
RapTora 2.2 by Eduard Nemeth, where their results: 92%
What does it mean? not exactly same vers..but very close!

As final words (about all SF based engines),
The Mother is clear... but I wonder who is the Father?)

Ok..that's all for now and thanks for your interest )



Best,
Sedat

[Sedat Canbaz]

Hello there )

I've realized to check again... and what a pity that
Even with new releases: same CTG bug is still on scene!
E.g today I tested Brainlearn 25.1 vs Polyfish 210623
And unfortunately the error happened again, btw actually
SugaR and CatroPoly are the ones with same CTG bug too!

And to be more clear,
I ran a few quick games, using a CTG book, based on only wins!
And as in previous test, the lost lines are played again..sad..

On other hand, (as I stated before):
In case of Polyglot book (if based on only wins) then
The Engine is out of book e.g which contains lost lines!

It's clear that there is something wrong in their CTG code!
And I still hope that,
The Top programmers will fix it (in next releases) and who knows?
Then we may run and test serious CTG books under Cutechess too!

Thank you in advance!

By the way (for anyone interested),
Here are the recent played games + my used wins ctg book:
https://mega.nz/file/204yHKwA#n5Xe32Ewr ... 1nRXxuAnKA

Greetings

[Sedat Canbaz]

Hello again )

I am very pleased to announce that,
A new, but a great idea is born...!!

Very soon I'm going to create a BIN book, which will be
Tuned to prefer latest Depth4 exp series (by Deeds)

To be more clear, according to Deeds tester,
His Depth4 exp managed to win all opponents...but what about under
SCCT- Book/NNUE CS XIV's conditions? Can we expect such performance?
We will see... after all, time will tell...!)
And all this mystery will be solved soon...!!

Greetings

[Sedat Canbaz]

Update

Depth4_Deeds book test is started, and here are 1st played results
Btw, unfortunately so far, not so good start by Depth4_Deeds book
Sure we are still in the beginning...let's see after more games...

Code: Select all

Rank Name                          Elo     +/-   Games   Score    Draw 
   0 Depth4_Deeds                  -43      19     177   43.8%   85.3% 
   1 HCanFree Upd                  120     127       9   66.7%   66.7% 
   2 Lotusblume                    102     134       7   64.3%   71.4% 
   3 Mustafacanseven                79     102       9   61.1%   77.8% 
   4 OPTIMUS 3.1                    79     102       9   61.1%   77.8% 
   5 ChessBot 170623                79     102       9   61.1%   77.8% 
   6 Chucaro 4 Upd2                 70      92      10   60.0%   80.0% 
   7 Cute by Cevdet                 44      82       8   56.3%   87.5% 
   8 NebulaRing178                  44      82       8   56.3%   87.5% 
   9 KainNew                        44      82       8   56.3%   87.5% 
  10 V by Kerveros                  39      73       9   55.6%   88.9% 
  11 SkynetSCCT 10.1                39      73       9   55.6%   88.9% 
  12 CachoUPD2                      39      73       9   55.6%   88.9% 
  13 OlisFish 220523                35      66      10   55.0%   90.0% 
  14 Agarta                         35      66      10   55.0%   90.0% 
  15 Leschats 140623                35     122      10   55.0%   70.0% 
  16 Predator 2.3                    0       0       9   50.0%  100.0% 
  17 StudentChess                    0       0       8   50.0%  100.0% 
  18 Dara 150222                     0     125       8   50.0%   75.0% 
  19 Goi 241222                      0       0       9   50.0%  100.0% 
  20 OlisFish 040623                 0       0       9   50.0%  100.0% 

177 of 1000 games finished.

[Sedat Canbaz]

Breaking News - 05.07.2023:
The tour resumed, but for short temporary time..and it's paused again!
Well, via this update: a new book entered: Depth4_Deeds and frankly
I expected much better performance because according to Deeds tests:
viewtopic.php?f=4&t=3388&start=40
His Depth4 exp managed to win many..but it seems under these cond.
Depth4_Deeds book was not so lucky...or simply his used lines will be
Stronger on Deeds's hardware? ? What I can say more over this issue:
SCCT is 'hard'.. not so easy as it looks! but there is nothing impossible!
And who knows? maybe next time Deeds newer lines will be Champ?
But before, I suggest harder work...because 'success' is no accident!

For Latest Standings:
https://sites.google.com/site/computers ... t-bn-cs-14

Best,
Sedat

[EmptikBest]

@deeds, can you please send a link to the old experience file Khaled shared at infinity chess as you say?