Ode to Berserk

[Jouni]

UHO book 60 + 0.6 level

Code: Select all

                           
1   obsidian150       +31  +89/=258/-53 54.50%  218.0/400
2   Berserk20250203   -31  +53/=258/-89 45.50%  182.0/400

[Chris Formula]

Berserk 20250307b: https://drive.google.com/file/d/146rlR7 ... yJNjJ/view
Changes: https://github.com/jhonnold/berserk/commits/main/

This version uses the new network berserk-deb80bf0ba9d.nn. I hope Jay will continue on developing Berserk, another great chess engine!

Chris

[peter]

Score of Obsidian15 vs Berserk 250307: 154 - 108 - 238 [0.546]
Elo difference: 32.1 +/- 22.1, LOS: 99.8 %, DrawRatio: 47.6 %
500 of 500 games finished.

3'+1" single thread (3.5GHz), UHO 2024 900-990cp 6mvs, regards

[Jouni]

I got better result for Berserk old UHO book 60 + 0.6:

Code: Select all

                           
1   obsidian150       +12  +105/=204/-91 51.75%  207.0/400
2   Berserk20250307   -12  +91/=204/-105 48.25%  193.0/400

There are already 4 gainers after March 7 in github .

[rulleeeee]

Where can I download the berserk-deb80bf0ba9d.nn network?

I use the network included with version 13 and Berserk plays so poorly that I beat it myself.

[Chris Formula]

Berserk 20250606: https://drive.google.com/file/d/1SX7PVA ... ok8Ue/view
Source: https://github.com/jhonnold/berserk/com ... 82a6790e5d

This version is based on a new neural network berserk-9b84c340af7e.nn.
Package includes AVX2 and SSE4 binaries.
Enjoy!

Chris

[Chacal X]

Good afternoon,
I will never tire of thanking you and Jim for the excellent work you do in favor of chess and thinking of the less fortunate, with the appropriate compilations.
Thank you very much!

[Chris Formula]

all good Flavio!