New Berserk soon?

[chrisw]

Torch 2, Berserk 13 and Rubichess have all almost reached Stockfish in strenght

Completely wrong.
ipmanchess:

Code: Select all

RubiChess 20240112 bmi2	Stockfish 20230922 avx512	313.5	686.5

Stefan Pohl:

Code: Select all

Stockfish 240322 avx2  vs. RubiChess 240112 avx2    : 1000 (+510,=468,- 22), 74.4 %

What is SF doing that Rubichess is not? Serious question. I thought you had implemented L1 sparsity algorithm and done SPSA tuning?

[Chessqueen]

Why is that version of Berserk 20240324 playing in TCEC, it is NOT the latest and strongest version ? ===> https://tcec-chess.com/

I understand they did not change the berserk version that played in L2 because they plan to continue L2.

Torch 2, Berserk 13 and Rubichess have all almost reached Stockfish in strenght and they are developed by much less contributors!

http://www.cegt.net/40_40%20Rating%20Li ... liste.html

I'm wondering if Torch that runs on Chess.com is the same of the version given to ranking lists (they use only few CPU and modest hardware) or is a different cluster version. Maybe the full-powered Torch 2 is better than Stockfish on top hardware. Where can I find detailed informations about the new Andrew Grant's engine?

I believe that the newer Obsidian 11.5 is 10 elo stronger than Rubichess

Rank Engine Score Ob Ru S-B
1 Obsidian11,5-avx2 17.5/30 · ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· =1=0=1==11==1======1========== 218.75
2 RubiChess-20240112_x86-64-bmi2 12.5/30 =0=1=0==00==0======0========== · ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· ·· 218.75


30 of 50 games played

Tournament start: 2024.03.31, 06:32:59
Latest update: 2024.04.03, 03:53:19
Site/ Country: DESKTOP-4QNC0GS, United States
Level: Blitz 0/5
Hardware: Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz with 15.9 GB Memory
Operating system: Windows 10 Home Home Edition (Build 9200) 64 bit
Table created with: Arena 3.5.1

[Uri Blass]

Why is that version of Berserk 20240324 playing in TCEC, it is NOT the latest and strongest version ? ===> https://tcec-chess.com/

I understand they did not change the berserk version that played in L2 because they plan to continue L2.

Torch 2, Berserk 13 and Rubichess have all almost reached Stockfish in strenght and they are developed by much less contributors!

http://www.cegt.net/40_40%20Rating%20Li ... liste.html

I'm wondering if Torch that runs on Chess.com is the same of the version given to ranking lists (they use only few CPU and modest hardware) or is a different cluster version. Maybe the full-powered Torch 2 is better than Stockfish on top hardware. Where can I find detailed informations about the new Andrew Grant's engine?

Even if I ignore rubichess that is the weaker of them it seems that they are not close to stockfish's strength.
Results in FRC CCRL list:

Stockfish16.1 vs Torch2 +27−10=158
Stockfish16.1 vs Berserk13 +42−3=105

Even 27-10 is a significant result.


https://stattrek.com/online-calculator/binomial
Assuming equal probabilities
I get P(X<=10)=0.00382 and I ignore the fact that white has better chances to win than black that probably only increase the significance of the results.

[Chessqueen]

Why is that version of Berserk 20240324 playing in TCEC, it is NOT the latest and strongest version ? ===> https://tcec-chess.com/

I understand they did not change the berserk version that played in L2 because they plan to continue L2.

Torch 2, Berserk 13 and Rubichess have all almost reached Stockfish in strenght and they are developed by much less contributors!

http://www.cegt.net/40_40%20Rating%20Li ... liste.html

I'm wondering if Torch that runs on Chess.com is the same of the version given to ranking lists (they use only few CPU and modest hardware) or is a different cluster version. Maybe the full-powered Torch 2 is better than Stockfish on top hardware. Where can I find detailed informations about the new Andrew Grant's engine?

Even if I ignore rubichess that is the weaker of them it seems that they are not close to stockfish's strength.
Results in FRC CCRL list:

Stockfish16.1 vs Torch2 +27−10=158
Stockfish16.1 vs Berserk13 +42−3=105

Even 27-10 is a significant result.


https://stattrek.com/online-calculator/binomial
Assuming equal probabilities
I get P(X<=10)=0.00382 and I ignore the fact that white has better chances to win than black that probably only increase the significance of the results.

Has Berserk 13 surpassed Dragon 3.3 YET ===> https://tcec-chess.com/