OpenChess

Posted: **Tue Mar 05, 2024 2:15 pm**

Indeed. From version 25.2 brainlearn had my attention. Which ended up swapping shash with him. Its has been a very good adversary against my derivatives. Been loving Brainlearn from that version shash is no longer the shash that i know, so i use brainlearn now.

Point is, the latest brainlearn versions are not that slow and boring, they have been exciting to was and with very interesting choice of moves in both bullets and blitz. Zues 26.5 is old the last version that saw live was version 50 (if not mistaken) so you see the difference in versions?.

Posted: **Tue Mar 05, 2024 2:45 pm**

OrgZ wrote: Tue Mar 05, 2024 2:15 pm Indeed. From version 25.2 brainlearn had my attention. Which ended up swapping shash with him. Its has been a very good adversary against my derivatives. Been loving Brainlearn from that version shash is no longer the shash that i know, so i use brainlearn now.

Point is, the latest brainlearn versions are not that slow and boring, they have been exciting to was and with very interesting choice of moves in both bullets and blitz. Zues 26.5 is old the last version that saw live was version 50 (if not mistaken) so you see the difference in versions?.

I see the differences of course.. )

Actually my question was not exactly that but.. anyhow ))

Well, and frankly
In most cases...my target and pay attention more
Over public chess engines, because private engines
Are not so useful for overall chess community, right ?)

Note also that in past, from Zeus author (Kerveros)
I was getting frequent the latest Zeus releases, but
Later I switched testing mainly to public releases..
So I tested Zeus 26.5 version as last public one... sure
If there is newer Zeus to public..just let me know please
Note: I am aware that there is Zeus 27..but close to v 26.5
Even I can say Zeus 27 performed weaker than Zeus 26.5

Btw, as you know, I test Private books too, but situation
Is not exactly same as private engines, because from
Book Authors view...mostly wish that their strongest lines to
To be remained secretly.. if will be publicly available..
Then quite easy to be blocked... sure I am talking for
Experienced Top book makers...

Best,
Sedat

Posted: **Tue Mar 05, 2024 4:44 pm**

UPDATE

Well, here are my latest organized strength testings,
And as we see, mostly are quite close in strength...

SCCT - Bullet Rating (30s+0.5s):

Code: Select all

Rank Name              Elo    +    - games score oppo. draws 
   1 Brainlearn 27    3780    3    3  1500   51%  3775   95% 
   2 CoolIris 11.80   3778    3    3  1500   50%  3775   96% 
   3 SF-PB 080124     3776    3    3  1500   50%  3775   96% 
   4 Polyfish 140124  3776    3    3  1500   50%  3775   96% 
   5 Raid v3.4        3775    3    3  1500   50%  3775   96% 
   6 CoolIris 11.90   3775    3    3  1500   50%  3775   96% 
   7 SF POLY 220723   3775    3    3  1500   50%  3775   95% 
   8 SF POLY 261123   3775    3    3  1500   50%  3775   96% 
   9 Eman 9.90        3775    3    3  1500   50%  3776   95% 
  10 Incognito 5 Pro  3774    3    3  1500   50%  3776   94% 
  11 ShashChess 34.6  3771    3    3  1500   49%  3776   94% 

Rating's Draw Ratio (based on 8250 games) is high: 95%

Here are the next new Duels...Bullet 2024 Vs Blitz 2022
And all of below Duels are played just out of curiosity... enjoy )

Code: Select all

Test 1: Bullet (30s+05s) Vs Blitz (2m+1s): 6 Elo difference
                           
1   Zeus 26.5 2m+1s        +17/-9/=474 50.80%  254.0/500
2   Brainlearn 27 30s+05s  +9/-17/=474 49.20%  246.0/500

-----------------------------------------------------------

Test 2: Bullet (30s+05s) Vs Blitz (2m+1s): 6 Elo difference
                                            
1   Kayra 1.5 2m+1s         +15/-7/=478 50.80%  254.0/500
2   CoolIris 11.80 30s+05s  +7/-15/=478 49.20%  246.0/500

-----------------------------------------------------------

Test 3: Bullet (45s+05s) Vs Blitz (2m+1s): No Elo difference
                                                                  
1   Kayra 1.5 2m+1s        +9/-9/=482 50.00%  250.0/500 
2   Brainlearn 27 45s+05s  +9/-9/=482 50.00%  250.0/500

GAMES:
https://mega.nz/file/uhhFxTJQ#BFHcwaBRA ... Kr0TZ1C_Hc

As final words,
For a better conclusion we need more various strength testings
But anyhow, now we have some data, which seems to be not bad
E.g 2024's Bullet games can be counted as Blitz of 2022 year!!

In other words,
If I will run Bullet Books Tours (by latest SF- Based NN Engines)
The planning games are expecting not so bad in quality too!
Sure we'll see plenty of draws..but if we want to be in trend
For right now I can not see any other way...what about you?

And I hope current new strength results to be useful too!

Best,
Sedat

Posted: **Tue Mar 05, 2024 6:38 pm**

UPDATE 2

NEW HASH TEST (1m+0.5s): 64 MB Vs 128 MB

And as we see, 2.3 Elo (in favor for 64 MB Hash)

Code: Select all

Score of Incognito 128mb vs Incognito 64mb: 5 - 9 - 586 [0.497]
...      Incognito 128mb playing White: 4 - 1 - 295  [0.505] 300
...      Incognito 128mb playing Black: 1 - 8 - 291  [0.488] 300
...      White vs Black: 12 - 2 - 586  [0.508] 600

Elo difference: -2.3 +/- 4.2, LOS: 14.3 %, DrawRatio: 97.7 %
600 of 600 games finished.

Conditions:
2x Epyc 7B12, CuteChess, 1 Core, Ponder OFF, 1m+0.5s, Balsa, 4-MEN
NOTE: Incognito 5 PRO is used (for both players 2 copies installed)

Hope this new HASH test helps too!

Ok dear chess friends, see you later.. )
Sure the games will be shared later too)

Best,
Sedat

Posted: **Tue Mar 05, 2024 7:08 pm**

Sure, help a lot.

Posted: **Tue Mar 05, 2024 8:58 pm**

Thank you very much for tests updates.
Best regards

Posted: **Tue Mar 05, 2024 11:17 pm**

Anton101 wrote: Tue Mar 05, 2024 7:08 pm Sure, help a lot.

I am glad )

Posted: **Tue Mar 05, 2024 11:18 pm**

Homayoun wrote: Tue Mar 05, 2024 8:58 pm Thank you very much for tests updates.
Best regards

Not at all.. it is my pleasure Homayoun )

Greetings

Posted: **Wed Mar 06, 2024 2:30 pm**

Hello there )

Just I'd like to inform you also that,
I managed to run another new testings, played mainly to see
About Draw influences e.g if we going to increase Bullet TC
From 30s+05s to 1m+05s, then the DRAW difference is: 2
Yes.. the draws with 1m+0.5s is higher, but normal..since all
Of the engines have more time to think, so more draws on scene!
In same time that means less blunder moves, better quality games
But what about FUN?.. who cares ?) at least I care....you see)
E.g Raunfish SC or using Cfish 030820 was as clear prove !)
But sometimes we have to move on..boring if each time same..

Meanwhile, what is unusual, actually no any surprise that
Even under 1 Core, Ponder OFF, Bullet etc we see high draws!
To be more clear, if going to compare both strength results:
Bullet 30s+0.5s = 95% draws / Bullet 60s+0.5s= 97% draws

Btw, just if we using MP, PONDER ON, Blitz etc: 99% draws!
Sure I refer over my testings+Balsa...but I see sometimes
In other sources...these DRAW values are lower..and I assume
You know why..if not clear, here are one of main influences:
- Using weak openings play serious role over draw percentages
- Using wrong Eng, Hardware configurations play other role too
I mean for those.. who are beginners in Eng, Book testings..
- Even if using NON-stable chess GUIs... can play role too!
- Sure many other factors may play over the draw influences
E.g if running matches on unstable operating systems etc.

But just if matches on more optimal cond.. let's say Rapid,
Forget about to see wins...maybe in 200-300 games...?)
Sure I refer about NN engines with over than 3750+ Elo

In other words, what I'm trying to tell you..
I am doing my best to see more FUN... !)

And please stay tuned..soon as possible
New Rankings + Games will be shared...

Greetings )

Posted: **Wed Mar 06, 2024 4:16 pm**

Well, here is mentioned new Bullet tour (1m+0.5s): with 97% draw ratio!
That means in same time, in 100 games: there are minus plus 3 wins...
So very likely I will set 30sec+0.6sec for the next planning Book CS...
At least, via 30s+06s it's expecting not so bad games, plus less draws...
Sure we will see more errors (blunder moves) via Bullet...but here is the key !
Otherwise, let's choose, use such conditions.. that engines will never make mistake )
Besides, I assume that you know already about my recent organized Duels:
Brainlrearn+Cool Iris Bullet Vs Kayra+Zeus Blitz
If you not aware.. Duel's played strength results are ended close to each other!
Sure some may not be so happy...but just in case of more time, cores etc
This time, FUN is going to be dead.. sorry..FUN comes as most important!
And as usual, I say over (who loves to see a lot of draws such as 97%-100% etc.)
What is point of using BIG umbrella..if no much rain on the horizon ?)

Btw, I realized to test some older versions, to see the influences...
So it is clear that even the former Champions are still not dead...!!

Code: Select all

                                                           
1   Brainlearn 26.5     50.0 - 50.050.0 - 50.050.0 - 50.050.5 - 49.550.5 - 49.551.5 - 48.5**       302.5/600
2   SF POLY 220723      50.0 - 50.050.0 - 50.050.5 - 49.549.5 - 50.551.0 - 49.051.0 - 49.0 **      302.0/600
3   CoolIris 11.90 RM1  50.0 - 50.050.0 - 50.050.0 - 50.050.5 - 49.549.5 - 50.551.5 - 48.5  **     301.5/600  -7933.75
4   SF-PB 051123        50.0 - 50.049.5 - 50.550.0 - 50.050.5 - 49.550.5 - 49.551.0 - 49.0   **    301.5/600  -7935.50
5   Eman 9.90           49.5 - 50.550.5 - 49.549.5 - 50.549.5 - 50.550.5 - 49.549.0 - 51.0    **   298.5/600
6   SunLight 3          49.5 - 50.549.0 - 51.050.5 - 49.549.5 - 50.549.5 - 50.549.5 - 50.5     **  297.5/600
7   Incognito 5 Pro     48.5 - 51.549.0 - 51.048.5 - 51.549.0 - 51.051.0 - 49.050.5 - 49.5      ** 296.5/600

Conditions:
2x Epyc 7B12, CuteChess, 1 Core, Ponder OFF, 60s+0.5s, Balsa, 64 Hash, 4-MEN

GAMES (contains recent hash test games too):
https://mega.nz/file/rtpAFYLD#bYR55cvpL ... vsOAzviyaA

Greetings

OpenChess

SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess

Re: SedatChess