top of page

Berserk 11.1 - released (2023-02-21 update).

They waited and finally got it !!

This is not the official logo of the Berserk chess engine.


This is about fans and sympathizers of the Berserk chess engine, to which I also belong :-)


If you haven't yet heard of the Berserk chess engine, then I invite you to visit a special page dedicated to the Berserk chess engine - here.


The latest eleventh release of the Berserk chess engine brings a lot of changes, improvements and new features in the source code of this popular and extremely powerful chess engine.


Berserk is a free open source chess engine under GNU GPL 3.0 license.


* I hope that no but no company will get the idea to use the source code of the Berserk chess engine contrary to the provisions of this license ;-)

"

What's Changed

  • Don't increase reductions for cutnodes when the move is tactical by @jhonnold in #417

  • Between search optimizations by @jhonnold in #418

  • Disable SE Extensions when clearly winning by @jhonnold in #419

  • Fast Range over modulo for TT indexing by @jhonnold in #420

  • Setup Pyrrhic submodule by @jhonnold in #421

  • Remove depth requirement on eval adjustment by @jhonnold in #422

  • Fix recapture extension extending non-captures by @jhonnold in #423

  • Introduce negative reductions when hash move is singular by @jhonnold in #424

  • Update Github Workflow to force Ubuntu 20.04 by @jhonnold in #425

  • Standard LMR + PVS Logic by @jhonnold in #426

  • Search Stack by @jhonnold in #428

  • Combined Patches by @jhonnold in #430

  • Cleanup by @jherrera80 in #429

  • MovePicker Refactor by @jhonnold in #433

  • Prevent positive score differences from impacting TM by @jhonnold in #435

  • 12 Byte TT Entries by @jhonnold in #437

  • Accumulator Cleanup by @jhonnold in #438

  • QSearch ordering strictly based on Capture History by @jhonnold in #439

  • Utilize 128bit multiplication for TT indexing by @jhonnold in #440

  • Train with a mix of Berserk and Koivisto data by @jhonnold in #445

  • Track bestmove using "Root Moves" by @jhonnold in #446

  • Move Type in Continuation History by @jhonnold in #447

  • Add simple protections for search explosions due to extensions by @jhonnold in #448

  • History values fit within int16_t by @jhonnold in #449

  • Remove SF Chess960 Cornered Bishop Logic by @jhonnold in #450

  • General Cleanups by @jhonnold in #451

  • Speedups by @jhonnold in #452

  • Optimize Apply Updates by @jhonnold in #453

  • Additional continuation histoy for follow-follow up by @jhonnold in #454

  • Separate Move Pickers by @jhonnold in #455

  • Hold TT Entry by @jhonnold in #456

  • Increase hidden layer to 768 neurons by @jhonnold in #458

  • No TT eval correction in QSearch by @jhonnold in #459

  • Updates by @jhonnold in #460

  • Add PV TT Eval correction in QSearch by @jhonnold in #461

  • Identify upcoming repetitions by @jhonnold in #462

  • Add WDL Output and Normalize CP to 50% at 100cp by @jhonnold in #463

  • Utilize LMR Depth when move pruning quiets by @jhonnold in #465

  • History Pruning by @jhonnold in #466

  • Split Good and Bad captures with history and SEE by @jhonnold in #467

"

Source: GitHub



A lot of it, right ?


Berserk chess engine author Jay Honnold and a new contributor nicknamed jherrera80 have been very busily over the last few months, as I witnessed by following their work on GitHub.


And there was certainly plenty to do. Berserk is one of the world's top chess engines, and undoubtedly the development of an already advanced and sophisticated chess engine is not a trivial thing and requires advanced programming techniques and extensive knowledge.

The official logo of the Berserk chess engine, source: GitHub



Have these so many changes been followed by the strength of Berserk as well?


Let's see its performance:

Source: MCERL (ongoing)


After playing 1100 games: 3652 Elo.

36 Elo more than Berserk 10.


I'm happy to say that Berserk is growing in strength and consolidating on the third place of the MCERL rating list :)

(Dragon 3.2 and Dragon 2.6.1 are the same chess engine by Komodo Chess in different versions).


It is noteworthy that:

  • There is an increasing difference in playing power between Berserk and the next Koivisto, even though it was recently released just 1 month earlier.

  • Berserk 11 plays at a very similar level to the excellent Dragon 2.6.1 by Komodo Chess, which is a commercial chess engine (6 Elo difference).


Big kudos to Jay Honnold and jherrera80 and hearty congratulations to the significant increase in the strength of the Berserk 11 chess engine.


I also thank you for such a universal source code that allows compilation for different systems. Respect!



Update 2023-02-21


Release: Berserk-11.1

"

Note: TCs with an increment are mostly un-impacted.


This is a bug fix release to resolve a major issue with TM in cyclical and no-increment TCs. Berserk would use all of it's time until playing completely random moves. Apologies for all testers who potentially wasted their time and CPU hours on a buggy Berserk.

"

Source: GitHub



Download area is at your disposal:


Berserk 11.1 source code & neural network by Berserk Devs

Berserk-11.1_Source_Code_&_NNUE
.7z
Download 7Z • 10.68MB

Android – Compiled by Archimedes

Berserk-11.1_Android
.7z
Download 7Z • 21.29MB

Linux arm64 & x64 – Compiled by Darius

Berserk-11.1_Linux_arm64_&_x64
.7z
Download 7Z • 22.01MB

Mac Apple Silicon & Intel – Compiled by Darius

Berserk-11.1_Mac_Apple_Silicon_&_Intel
.7z
Download 7Z • 21.53MB

Windows x64 – Compiled by Darius

Berserk-11.1_Windows_x64
.7z
Download 7Z • 11.43MB


 

In the Files area you will find free chess engines for download.

 
 


bottom of page