SedatChess

[Sedat Canbaz]

UPDATE 3

The Duel (between both recent Winners) ended!
And yes, Plentychess, Plentychess, Plentychess..
As we see, with clear margin ahead: 161+ Elo!
In other words, just unbelievable performance!

Grand Champion: Plentychess- Congrats Yoshie2000!

Code: Select all

Score of Plentychess 5.021 vs Booot 7.4: 88 - 1 - 111 [0.718]
...      Plentychess 5.021 playing White: 54 - 1 - 45  [0.765] 100
...      Plentychess 5.021 playing Black: 34 - 0 - 66  [0.670] 100
...      White vs Black: 54 - 35 - 111  [0.547] 200
Elo difference: 161.9 +/- 30.1, LOS: 100.0 %, DrawRatio: 55.5 %
200 of 200 games finished.

Duel's Conditions:
2x Epyc 7B12, CuteChess, Ponder OFF, 1m+1s, 1 Core, 128 Hash, 4 MEN
Both engines used to play via Unique suite: each opening played twice

GAMES:
https://mega.nz/file/rs5knLqR#FpaOLo03l ... VIXzcvcMmg

Best,
Sedat

[Sedat Canbaz]

Hello Friends )

Well, a new 3800+ Elo strength test is organized,
Sure for the next new future book competitions..
So I realized to share my recent chess activities
With all of you..just for computer chess progress..

More Details:
The target to check which Top engines are:
- Less Drawish (via strong opening lines)
Actually known..but worth to gave a try )
- Stronger (in strength between each other)
- Better (in opening book settings etc.)
- More stable (if using fast Bullet TC)

Just a few examples, according to my experience:
Some come with poor book settings or don't follow
Strictly settings (if various book moves enabled)

In short,
I will test and I will see..if it's all right, then
I'll pick the best one, which is more suitable!

By the way, here are the FINAL Standings:
SF-X 130425 wins..Congratulations to Vasid Chouhan!

Code: Select all

Rank Name                          Elo     +/-   Games   Score    Draw 
   1 SF-X 130425                     4       6    1000   50.5%   91.9% 
   2 Kookaburra 3.10                 3       6    1000   50.4%   92.5% 
   3 SF-PRO2 020525                  0       6    1000   50.0%   92.5% 
   4 SF-PB 250525                    0       6    1000   50.0%   91.1% 
   5 Omega 6M                       -1       6    1000   49.9%   91.0% 
   6 DarkSeid 6.0                   -7       6    1000   49.0%   91.2% 

GAMES:
https://mega.nz/file/CwAVWTDL#pRw1xQDy6 ... v63OtRGDhU

Conditions:
2x Epyc 7B12, CuteChess120, PonderOFF, 15s+0.6s, 1 Core, 32 Hash, 4-MEN
All engines are used to play via Unique suite: each opening played twice

And please stay tuned.. very soon new updates..

Greetings

[Sedat Canbaz]

UPDATE

Just to be more clear,
Why I've preferred fast Bullet TC + Unique suite etc?
Reason the planning official book tour is expecting
Close draw ratio...Moreover, my target to keep
The draws less than 90-92%Why? for more FUN!

In same time, the planning tour games are expecting
To be not so bad in quality..you may know, if running
3800+ Elo tours, no much ways to reduce the draws..

But this is also true... no matter..how is good..
In case of close to 95% + and more draws..it is
Really boring, waste of time, because I think that
FUN should be one of most important parts, right ?

Otherwise, if almost all games will be ended DRAW..
Believe it or not, many of us will loose interest!
Besides, no much ways to see which book is best..

Or if giving much NN assistance, Slow TC, MP etc.
Then if equal or 99% draws..How you'll determine?
Btw, see please: 3 Champs, if under better, optimal cond:
https://sites.google.com/site/computers ... s-26-blitz
Plus..in case of running with more optimal cond.
We'll need a lot of number of games (per player),
Otherwise, the error margin will be too high!

In other words, if I am going to summarize:
In case of 95%+ draws + more...SORRY...
I'm good in my job, but ask Me other wish..that I may...

Btw, as next, soon as possible:
Book settings test will be started...

Best,
Sedat

[Sedat Canbaz]

UPDATE 2

The Book Strength Test is ended...

1st of all,
I've picked two special books via more moves as enabled!
In this way, a affective way to check which are better...

After running too many various testings..FINELY:
SF-PB 250525 eng seems to be One of Best..
Really a good job, Congratulations to Anton CES!
Congrats to all rest Book/Eng Authors as well!

So in the planning next/new Championship:
All Books will be played by SF-PB 250525

Reason?
SF-PB's book settings worked flawlessly, plus as
Other important issue is that less Drawish: 84%

About Kookaburra 3.10,
One of best too, but slightly more Drawish: 86%

About SF-X 130425,
One of strongest for sure but Drawish: 91%, very likely
Due to its book settings...

About Raubfish SC,
Not so Drawish: 84% plus closest (in Elo) to overall!
But a bit boring...if each time via Raunfish SC, right ?)

About Cfish 030820 NON-NNUE,
Less Drawish engine than all: 75% but due to more blunders..
But who knows?) perhaps in other future tours I may use it...

And here are the recent book setting strength results:

Code: Select all

Overall (all in one) performance: + 9 Elo / 86% Draws

1   Cevdet 131024   +258/-178/=2764 51.25%  1640.0/3200
2   Everest 300923  +178/-258/=2764 48.75%  1560.0/3200
--------------------------------------------------------
By SF-PB 250525: + 16 Elo / 84% Draws
                 
1   Cevdet 131024   +98/-53/=849 52.25%  522.5/1000
2   Everest 300923  +53/-98/=849 47.75%  477.5/1000

-------------------------------------------------------
By Kookaburra 3.10: + 13 Elo / 86% Draws
            
1   Cevdet 131024   +74/-43/=683 51.94%  415.5/800
2   Everest 300923  +43/-74/=683 48.06%  384.5/800

----------------------------------------------------
By SF-X 130425: 0+ Elo / 91% Draws
             
1   Everest 300923  +45/-44/=911 50.05%  500.5/1000
2   Cevdet 131024   +44/-45/=911 49.95%  499.5/1000
******************************************************
Plus as alternatives with other well-known engines:

By Raubfish SC: + 7 Elo / 84% Draws

Note: played via TC: 30s+0.6s 
                
1   Cevdet 131024   +25/-24/=151 50.25%  100.5/200
2   Everest 300923  +24/-25/=151 49.75%   99.5/200
----------------------------------------------------
By Cfish 030820 NON-NNUE: + 2 Elo / 75% Draws

Note: played via TC: 40s+0.6s 
                    
1   Cevdet 131024   +25/-24/=151 50.25%  100.5/200
2   Everest 300923  +24/-25/=151 49.75%   99.5/200

GAMES:
https://mega.nz/file/ro5F2a4Q#AMJB5BCDo ... 0eBLcrtSZA

Conditions:
2x Epyc 7B12, CuteChess120, PonderOFF, 15s+0.5s, 1 Core, 32 Hash, 4-MEN
Best Move: Disabled / Book Depth: 255

More Details:
-SF-PRO engine is not tested... due to is poor in book options...e.g
Includes one book, plus no way to disable/enable: Best Move OFF/ON
-Omega 6M does not include book settings, sad indeed, so not tested,
Actually it's possible..but required a polyglot adapter...but complicated..
-DarkSeid 6.0 include book settings but was out due to its performance..

And I hope all these new testings to be useful...
Sure we need more of them..but that's all for now )

Greetings

[Sedat Canbaz]

UPDATE 3

Here is another new Duel strength testings,
Organized just out of curiosity and for FUN!
But note that played in Handicapped TC cond
Moreover, running, sharing more..better, right?
And I hope you will enjoy it..

1st of all, and to be more clear,
SF-PB 250525 is played via much faster Bullet TC 15s+0.5s,
Where its opponent (Hypnos) via slower Bullet TC 60s+0.5s

Just GREAT performance, despite on HUGE Disadvantage,
SF-PB 250525 is managed to win..that means a lot! and
It seems Hypnos needs slower TC to be better/stronger..!

SF-PB 15s+05s vs Hypnos 60s+0.5s:+5 Elo for SF-PB!

The Winner: SF-PB 15s+0.5s - Congrats to Anton CES!

Code: Select all

1   SF-PB 250525   +15/-11/=274 50.67%  152.0/300
2   HypnoS++ 1.02  +11/-15/=274 49.33%  148.0/300

GAMES:
https://mega.nz/file/XwQiUZQS#ohHR89XZa ... YwympW8y6g

Meanwhile,
Who said via fast Bullet TC are as bad quality ?)
Maybe true on their old-dated, jalopy hardwares?
But here SCCT, which serves a Super Faat Machine:
2x EPYC 7B12, 128 Cores (when HT Disabled), but
When Hyper Threading is Enabled: 256 Threads,
For more details (see and learn about its power ...):
https://sites.google.com/site/computers ... benchmarks

By the way, Why so many tests over and over?
Because some Outskirts members have no clue!
Remember also that SCCT's played 1 Core can
Be counted as many Threads (as MP, Blitz, Rapid on
Slower, old hardwares), Proved and Guaranteed!

Actually we're in 2025 but still some are not
Aware that Core/Thread are 2 different things!
Anyhow sooner or later, alright on the night..

Btw, please stay tuned,
As next another interesting handicapped match:
SF-PB 250525 15s+0.5s vs Raubfish SC 60s+0.5s

Greetings

[Sedat Canbaz]

UPDATE 4

After checking more closely the played games...
Unfortunately Hypnos lost 4 (four) games on time,
Not sure as Eng or CuteChess GUI bug...but rather
I think as GUI bug..because according to my
Other GUI testings.. now I remembered that
Some Cutechess GUI series are not so stable...

And for more accurate and fair conclusions,
Here are the standings..as without those 4 buggy games:
Again..despite on everything...still Hypmos is not better:

SF-PB 15s+0.5s vs Hypnos 60s+0.5s: O (zero) Elo diff.

Code: Select all

1   HypnoS++ 1.02  +11/-11/=274 50.00%  148.0/296  5520.00
2   SF-PB 250525   +11/-11/=274 50.00%  148.0/296  5520.00

Note also that for the current handicapped matches,
I used a special Cutechess GUI ver, which downloaded:
https://rwbc-chess.de/

Otherwise,
if using the official and stable Cutechess GUI 1.2.0,
No way to run the matches via handicapped TC cond.

Greetings )

[Sedat Canbaz]

UPDATE 5

Duel handicapped series are on... and again
Organized just out of curiosity and for FUN!
Plus..I think that these tests are quite useful..e.g
Many of us can imagine the power of the planning
Future tours that even via fast Bullet TC can be
Counted as not so bad in quality for sure !

And great performance again...and as see even via
So much disadvantage cond. SF-PB still unbeatable!

On the other hand,
No any BIG surprise that Raubfish performs weaker...
Since much old-dated, released several years ago but
I can't say as same words for Hypnos..since a new one,
Belonging to this year (2025) !

Anyhow, here are the FINAL results:

SF-PB 15s+05s vs Raubfish 60s+0.5s:+45 Elo for SF-PB!

The Winner: SF-PB 15s+0.5s - Congrats to Anton CES!

Code: Select all

1   SF-PB 250525       +51/-12/=237 56.50%  169.5/300
2   Raubfish X48c3 SC  +12/-51/=237 43.50%  130.5/300

GAMES:
https://mega.nz/file/CwgVBZ7S#0lLYl7MlT ... 6SYgwV4HXM

Conditions (for both recent matches):
2x Epyc 7B12, CuteChess, Ponder OFF, 1 Core, 64 Hash, 4 MEN
Both used to play via Unique suite: each opening played twice

Note also that,
No any game is lost on time...in same time that means
Both engines Raunfish And SF-PB are more stable than
Hypnos 1.02++

Have fun and enjoy

Best,
Sedat

[Homayoun]

Thank you very much Sedat.

[Sedat Canbaz]

Thank you very much Sedat.

Not at all, it's my pleasure Homayoun )

[Anton101]

Dear Sedat, I have no credit for SF PB. I only added the book options and modified the Polybook code to optimize it, but nothing more. Full credit goes to the Stockfish developers. Thanks for your kind words and tour.

Best regards!!