Drofa_4.1.0_fix

This fixes an issue with RAM allocation.
Everything else should be the same.

Drofa_4.1.0

New release.
Some work was made, mainly to the search and overall binary speed.
At least one more release can be expected some time in future.

Major gainers:

• Tuning of the pre-move-loop pruning
• SEE rewrite, in accordance to what is used in top engines
• Using ALPHA cutoffs in the TT
• SPSA of the incremental TM (big thanks for AGE for implementing SPSA for openbench)
• Rewriting TT allocation and probing

Regression on the Pohl.pgn book

Elo   | 51.91 +- 3.55 (95%)
Conf  | 40.0+0.40s Threads=1 Hash=64MB
Games | N: 20000 W: 6892 L: 3926 D: 9182
Penta | [150, 1510, 4217, 3470, 653]

You can expect much less elo on a balanced books and/or while using non-incremental TM

As per usual, choose the binary with best speed, but building using make should give a better binary

Drofa 4.0.0

It was a long way since Drofa 3.0.0 release (+375 elo), and I am ready to present Drofa 4.0.0.

Improvements:
Compared to-pre-released version (Drofa_3.3.22), several improvements were made:

1. SMP bugfix
2. Addition of the supplementary King-Pawn NN (it is NOT replacing existing hand-crafted eval)
3. Some improvements on UCI

In single core tests, this results in ~10 elo on a high-bias book, however on 8CPU up to +40 elo can be expected (compared to 3.3.22).

Future
As of Drofa 4.0.0 the engine reached incredible strength I never thought I would be able to achieve.
However, introduction of Efficiently Updatable Neural-Network-based evaluation functions (NNUE) allowed
new engines to eclipse Drofa in strength, and made long developed engines strength basically unreachable for me.

I dont want Drofa to become just another NNUE engine. However I do not have time to progress at current pace with releases.
Thus:

• In future Drofa would receive releases every 15 elo-gaining patches or every +20 elo gained on high-bias book (whichever comes first).
• If NNUE is ever to be ported on top of Drofa`s search, it will be developed as a separate project. All search patches of said project would be attempted in Drofa.

Selecting a binary:
chose one that dont crash and shows the highest speed

Drofa_3.3.22

This is a Drofa pre-release.
Drofa_3.3.22 will be a last version that is using ONLY hand-crafted evaluation (HCE).
I became harder and harder to improve HCE for me, and I dont have much time for it, thus HCE development will be stalled from now on.
Further release will include King-Pawn NN and thus wont be full HCE.

Improvements:
PEXT - movegen
Draw-randomization
Improvements in the counterMove History (cmh)
Implemented cmh-pruning
Winnability evaluation
Endgame-specific evaluation

Self-play results (LTC, 8moves_noob book):

ELO | 85.21 +- 3.32 (95%)
CONF | 40.0+0.40s Threads=1 Hash=64MB
GAMES | N: 20000 W: 7203 L: 2394 D: 10403

Selecting a binary:
chose one that dont crash and shows the highest speed

Drofa_3.3.0

Hello all!
Drofa 3.3.0 released
As usual, it shows about +100 elo improvement in the self-play.
Contains improved Hand-Crafted evaluation (~+40 elo), various search improvements and a couple of bugfixes

STC 8'' + 0.08''

ELO   | 109.76 +- 3.63 (95%)
CONF  | 8.0+0.08s Threads=1 Hash=8MB
GAMES | N: 20000 W: 8648 L: 2532 D: 8820

LTC 40'' + 0.4''

ELO   | 115.92 +- 3.40 (95%)
CONF  | 40.0+0.40s Threads=1 Hash=64MB
GAMES | N: 20000 W: 8235 L: 1799 D: 9966

Drofa 3.2.0

Since the last release Drofa got ~100 self-play elo.

The most important changes:

• Rewrite of history bonuses (now they are capped, plus other tweaks)
• Added capture history
• Added countermove history
• Added late capture extension
• Scale evaluation based on winning side pawn count
• Adjust some of the extensions if null move failed
• Scale evaluation for OCB + Rook / Knight endgames
• Some evaluation tweaks / a couple of new eval terms
• Much more agressive LMP, LMR and SEE
• Retuned evaluation with Drofa data

STC 8'' + 0.08''

ELO   | 94.21 +- 5.29 (95%)
CONF  | 8.0+0.08s Threads=1 Hash=8MB
GAMES | N: 10000 W: 4273 L: 1626 D: 4101

LTC 40'' + 0.4''

ELO   | 109.37 +- 5.06 (95%)
CONF  | 40.0+0.40s Threads=1 Hash=64MB
GAMES | N: 10000 W: 4249 L: 1201 D: 4550

Drofa 3.1.0

Since the last release i got ~100 self-play elo.

Changes include, but not limited to:
-Tuning pruning margins
-Handtuning and Optuna tuning of a Safety values
-Evaluation of threats (pieces attacking other pieces)
-Improving evaluation of the pawns
-Improved evaluation of the KingPawnShield
-Improved evaluation of passed pawns (Free, FreeToAdvance, Distance between passers)
-Singular extentions
-Some other minor search tweaks

STC: 10s+ 0.1s
Score of Drofa_3.1.0 vs Drofa_3.0.0: 781 - 238 - 481  [0.681] 1500
Elo difference: 131.74 +/- 14.97

LTC: 60s + 0.6s
Score of Drofa_3.1.0 vs Drofa_3.0.0: 666 - 225 - 609  [0.647] 1500
Elo difference: 105.25 +/- 13.70

Drofa 3.0.0

Changes include, but not limited to:

• Tuning of the evaluation features
• Several new evaluation features added
• SEE implemented
• Aspiration window implemented
• Lazy SMP implemented

Drofa 3.0.0 is ~250 - 300 elo stronger than Drofa 2.2.0:

STC	(10``+ 0.1``):
Score of Drofa_3.0.0 vs Drofa_2.2.0: 386 - 35 - 79  [0.851] 500
Elo difference: 302.70 +/- 35.74

LTC	(60`+ 0.6`):
Score of Drofa_3.0.0 vs Drofa_2.2.0: 205 - 19 - 76  [0.810] 300
Elo difference: 251.89 +/- 38.73

Drofa 2.2.0

Drofa 2.2.0 release.

~40 elo stronger than 2.1.0 version
~140 elo stronger than 2.0.0 version.

Improvements include:

• better move ordering
• tempo evaluation
• king proximity to passer evaluation
• refactoring of the transposition table.
• "Hash" UCI option added
• LMR and NMP improvements
• PV no longer getting cut off during long searches
• several bugfixes

Drofa 2.0.0

Drofa v. 2.0.0

Roughly 30 elo-gaining patches since Drofa 1.0.0 release.

~50 - 90 elo above Drofa 1.3.1 (2402 elo in CCRL blitz).

Score of Drofa_1.3.8 vs Drofa_1.3.1: 180 - 75 - 231 [0.608] 486
Elo difference: 76.27 +/- 22.41

Special thanks to Guenther Simon for helping me with Drofa windows compiles.