fairystockfish.git
4 years agoSome more variant descriptions
Fabian Fichter [Fri, 2 Apr 2021 18:32:12 +0000]
Some more variant descriptions

No functional change.

4 years agoSimplify piece value calculation
Fabian Fichter [Thu, 1 Apr 2021 17:14:22 +0000]
Simplify piece value calculation

karouk STC
LLR: -2.99 (-2.94,2.94) [-10.00,5.00]
Total: 9432 W: 4175 L: 4305 D: 952
http://www.variantfishtest.org:6543/tests/view/60620a8b6e23db21bb8bca92

karouk LTC
LLR: 2.96 (-2.94,2.94) [-10.00,5.00]
Total: 4812 W: 2122 L: 2095 D: 595
http://www.variantfishtest.org:6543/tests/view/6062c9c06e23db21bb8bca97

4 years agoSmall clean-up
Stéphane Nicolet [Wed, 24 Mar 2021 20:55:49 +0000]
Small clean-up

Bench: 4321677

4 years agoSimplify King Evasion
bmc4 [Tue, 30 Mar 2021 03:42:45 +0000]
Simplify King Evasion

Simplify away the removal of some illegal `KING`-evasion moves during move
generation. Verified for correctness by running perft on the following positions:

```
./stockfish
bench 16 1 6 default perft
Nodes searched: 71608931810

./stockfish
position fen 4rrk1/1p1nq3/p7/2p1P1pp/3P2bp/3Q1Bn1/PPPB4/1K2R1NR w - - 40 21
go perft 6
Nodes searched: 6136386434
```

Passed STC:
LLR: 2.94 (-2.94,2.94) {-1.00,0.20}
Total: 16072 W: 1473 L: 1349 D: 13250
Ptnml(0-2): 57, 1047, 5710, 1159, 63
https://tests.stockfishchess.org/tests/view/60629e7ef183b42957b423b1

Passed LTC:
LLR: 2.94 (-2.94,2.94) {-0.70,0.20}
Total: 59064 W: 2214 L: 2177 D: 54673
Ptnml(0-2): 26, 1944, 25556, 1979, 27
https://tests.stockfishchess.org/tests/view/6062dce4f183b42957b423de

closes https://github.com/official-stockfish/Stockfish/pull/3415

No functional change

4 years agoAdd more comments for variant code
Fabian Fichter [Mon, 29 Mar 2021 19:38:00 +0000]
Add more comments for variant code

No functional change.

4 years agoExtend variant configuration checks
Fabian Fichter [Mon, 29 Mar 2021 19:07:37 +0000]
Extend variant configuration checks

4 years agoAdd more variant documentation comments (#248)
Fabian Fichter [Mon, 29 Mar 2021 18:32:07 +0000]
Add more variant documentation comments (#248)

No functional change.

4 years agoImprove test coverage
Fabian Fichter [Mon, 29 Mar 2021 17:56:27 +0000]
Improve test coverage

Add further tests to cover more variant configuration features.

Closes #279.

No functional change.

4 years agoBetter document piece value calculation
Fabian Fichter [Mon, 29 Mar 2021 10:38:49 +0000]
Better document piece value calculation

No functional change.

4 years agoSimplify and unify FRC cornered bishop.
mstembera [Thu, 25 Mar 2021 20:33:05 +0000]
Simplify and unify FRC cornered bishop.

tested locally as fishtest doesn't support FRC:

STC NNUE
9646 - 9647 - 20707 [0.500] 40000 -0.0 +/- 2.4, LOS: 49.7 %, DrawRatio: 51.8 %

STC classical
9678 - 9609 - 20713 [0.501] 40000 0.6 +/- 2.4, LOS: 69.0 %, DrawRatio: 51.8 %

and verified independently:

Score of master vs patch: 6463 - 6580 - 34957 [0.499] 48000

closes https://github.com/official-stockfish/Stockfish/pull/3413

bench: 4321677

4 years agoAllow using Intel SDE for PGO builds.
Tomasz Sobczyk [Sat, 27 Feb 2021 10:52:18 +0000]
Allow using Intel SDE for PGO builds.

The software development emulator (SDE) allows to run binaries compiled
for architectures not supported by the actual CPU. This is useful to
do PGO builds for newer architectures. The SDE can currently be obtained from
https://software.intel.com/content/www/us/en/develop/articles/intel-software-development-emulator.html

This patch introduces a new optional makefile argument SDE_PATH.
If not empty it should contain the path to the sde executable

closes https://github.com/official-stockfish/Stockfish/pull/3373

No functional change.

4 years agoSimplify variant piece value adjustments
Fabian Fichter [Sat, 27 Mar 2021 12:39:05 +0000]
Simplify variant piece value adjustments

No functional change.

4 years agoTweak racing kings middlegame values
Fabian Fichter [Fri, 26 Mar 2021 22:51:40 +0000]
Tweak racing kings middlegame values

racingkings STC
LLR: -2.96 (-2.94,2.94) [0.00,10.00]
Total: 60604 W: 19139 L: 18467 D: 22998
http://www.variantfishtest.org:6543/tests/view/605e65a06e23db21bb8bca5c

racingkings LTC
LLR: 2.97 (-2.94,2.94) [0.00,10.00]
Total: 3390 W: 1026 L: 914 D: 1450
http://www.variantfishtest.org:6543/tests/view/605e78e96e23db21bb8bca5f

4 years agoTweak racing kings endgame piece values
Fabian Fichter [Sat, 27 Mar 2021 10:49:35 +0000]
Tweak racing kings endgame piece values

racingkings STC
LLR: 2.97 (-2.94,2.94) [0.00,10.00]
Total: 1035 W: 370 L: 282 D: 383
http://www.variantfishtest.org:6543/tests/view/605e58276e23db21bb8bca56

racingkings LTC
LLR: 2.97 (-2.94,2.94) [0.00,10.00]
Total: 1886 W: 582 L: 488 D: 816
http://www.variantfishtest.org:6543/tests/view/605e64f66e23db21bb8bca5a

4 years agoSmall cleanups (march 2021)
Stéphane Nicolet [Fri, 19 Mar 2021 18:43:25 +0000]
Small cleanups (march 2021)

With help of @BM123499, @mstembera, @gvreuls, @noobpwnftw and @Fanael
Thanks!

Closes https://github.com/official-stockfish/Stockfish/pull/3405

No functional change

4 years agoSimplify horde breakthrough bonus
Fabian Fichter [Mon, 22 Mar 2021 19:07:04 +0000]
Simplify horde breakthrough bonus

horde STC
LLR: 2.96 (-2.94,2.94) [-10.00,5.00]
Total: 9409 W: 4643 L: 4643 D: 123
http://www.variantfishtest.org:6543/tests/view/6057234c6e23db669974eb19

horde LTC
LLR: 2.98 (-2.94,2.94) [-10.00,5.00]
Total: 3083 W: 1546 L: 1500 D: 37
http://www.variantfishtest.org:6543/tests/view/6058449d6e23db669974eb27

4 years agoTweak atomic pawn endgame value
Fabian Fichter [Mon, 22 Mar 2021 19:06:20 +0000]
Tweak atomic pawn endgame value

atomic STC
LLR: 2.97 (-2.94,2.94) [0.00,10.00]
Total: 2650 W: 951 L: 839 D: 860
http://www.variantfishtest.org:6543/tests/view/60571eb36e23db669974eb07

atomic LTC
LLR: 2.95 (-2.94,2.94) [0.00,10.00]
Total: 5829 W: 1959 L: 1810 D: 2060
http://www.variantfishtest.org:6543/tests/view/60579b976e23db669974eb24

4 years agoFix en passant in FEN for grand in CECP
Fabian Fichter [Fri, 19 Mar 2021 23:44:19 +0000]
Fix en passant in FEN for grand in CECP

Closes #282.

4 years agoMake atomic castling fully compatible with lichess
Fabian Fichter [Fri, 19 Mar 2021 23:38:29 +0000]
Make atomic castling fully compatible with lichess

Closes #281.

4 years agoUse reference instead of pointer for pop_lsb() signature
Guy Vreuls [Sat, 13 Mar 2021 16:40:07 +0000]
Use reference instead of pointer for pop_lsb() signature

This patch changes the pop_lsb() signature from Square pop_lsb(Bitboard*) to
Square pop_lsb(Bitboard&). This is more idomatic for C++ style signatures.

Passed a non-regression STC test:
LLR: 2.93 (-2.94,2.94) {-1.25,0.25}
Total: 21280 W: 1928 L: 1847 D: 17505
Ptnml(0-2): 71, 1427, 7558, 1518, 66
https://tests.stockfishchess.org/tests/view/6053a1e22433018de7a38e2f

We have verified that the generated binary is identical on gcc-10.

Closes https://github.com/official-stockfish/Stockfish/pull/3404

No functional change.

4 years agoExtend perft tests (#279)
Fabian Fichter [Fri, 19 Mar 2021 18:52:37 +0000]
Extend perft tests (#279)

4 years agoDouble atomic PSQT bonus
Fabian Fichter [Fri, 19 Mar 2021 12:10:44 +0000]
Double atomic PSQT bonus

atomic STC
LLR: 2.94 (-2.94,2.94) [0.00,10.00]
Total: 2344 W: 840 L: 733 D: 771
http://www.variantfishtest.org:6543/tests/view/605270806e23db669974eac6

atomic LTC
LLR: 2.97 (-2.94,2.94) [0.00,10.00]
Total: 1767 W: 615 L: 517 D: 635
http://www.variantfishtest.org:6543/tests/view/605300406e23db669974ead2

4 years agoTweak PSQT for pseudo-kings
Fabian Fichter [Fri, 19 Mar 2021 12:09:37 +0000]
Tweak PSQT for pseudo-kings

extinction STC
LLR: 2.96 (-2.94,2.94) [0.00,10.00]
Total: 495 W: 285 L: 181 D: 29
http://www.variantfishtest.org:6543/tests/view/605300756e23db669974ead6

extinction LTC
LLR: 2.97 (-2.94,2.94) [0.00,10.00]
Total: 429 W: 253 L: 150 D: 26
http://www.variantfishtest.org:6543/tests/view/605390b36e23db669974eada

atomic STC
LLR: 2.97 (-2.94,2.94) [0.00,10.00]
Total: 3389 W: 1195 L: 1074 D: 1120
http://www.variantfishtest.org:6543/tests/view/605271e06e23db669974eacc

atomic LTC
LLR: 2.94 (-2.94,2.94) [0.00,10.00]
Total: 1599 W: 546 L: 452 D: 601
http://www.variantfishtest.org:6543/tests/view/605390e86e23db669974eade

4 years agoAdd a specific FRC correction from classical to NNUE
Vizvezdenec [Mon, 15 Mar 2021 18:05:01 +0000]
Add a specific FRC correction from classical to NNUE

our net currently is not trained on FRC games, and so doesn't know about the important pattern of a bishop that is cornered in FRC.
This patch introduces a term we have in the classical evaluation for this case, and adds it to the NNUE eval.

Since fishtest doesn't support FRC right now, the patch was tested locally at STC conditions,
starting from the book of FRC starting positions.

Score of master vs patch: 993 - 2226 - 6781  [0.438] 10000

Which corresponds to approximately 40 Elo

The patch passes non-regression testing for traditional chess (where it adds one branch).

passed STC:
https://tests.stockfishchess.org/tests/view/604fa2532433018de7a38b67
LLR: 2.95 (-2.94,2.94) {-1.25,0.25}
Total: 30560 W: 2701 L: 2636 D: 25223
Ptnml(0-2): 88, 2056, 10921, 2133, 82

passed STC also in an earlier version:
https://tests.stockfishchess.org/tests/view/604f61282433018de7a38b4d

closes https://github.com/official-stockfish/Stockfish/pull/3398

No functional change

4 years agoChange definition of between_bb()
bmc4 [Mon, 15 Mar 2021 19:06:42 +0000]
Change definition of between_bb()

We remark that in current master, most of our use cases for between_bb() can be
optimized if the second parameter of the function is added to the segment. So we
change the definition of between_bb(s1, s2) such that it excludes s1 but includes s2.

We also use a precomputed array for between_bb() for another small speed gain
(see https://tests.stockfishchess.org/tests/view/604d09f72433018de7a389fb).

Passed STC:
LLR: 2.96 (-2.94,2.94) {-0.25,1.25}
Total: 18736 W: 1746 L: 1607 D: 15383
Ptnml(0-2): 61, 1226, 6644, 1387, 50
https://tests.stockfishchess.org/tests/view/60428c84ddcba5f0627bb6e4

Yellow LTC:
LTC:
LLR: -3.00 (-2.94,2.94) {0.25,1.25}
Total: 39144 W: 1431 L: 1413 D: 36300
Ptnml(0-2): 13, 1176, 17184, 1178, 21
https://tests.stockfishchess.org/tests/view/605128702433018de7a38ca1

Closes https://github.com/official-stockfish/Stockfish/pull/3397

---------

Verified for correctness by running perft on the following position:

./stockfish
position fen 4rrk1/1p1nq3/p7/2p1P1pp/3P2bp/3Q1Bn1/PPPB4/1K2R1NR w - - 40 21
go perft 6

Nodes searched: 6136386434

--------

No functional change

4 years agoRemove advanced_pawn_push()
Vizvezdenec [Thu, 11 Mar 2021 11:51:20 +0000]
Remove advanced_pawn_push()

Continuation of work by @topologist: we now do futility pruning and movecount
pruning in qsearch() for pawn pushes up to the 7th rank. So the condition to
avoid the pruning is if the move is a promotion or not. This allows to get rid
of the advanced_pawn_push() function in position.h alltogether.

Passed STC
https://tests.stockfishchess.org/tests/view/6048c5842433018de7a387e6
LLR: 2.93 (-2.94,2.94) {-1.25,0.25}
Total: 34424 W: 3081 L: 3015 D: 28328
Ptnml(0-2): 110, 2442, 12052, 2488, 120

Passed LTC
https://tests.stockfishchess.org/tests/view/6048f7d22433018de7a387f0
LLR: 2.94 (-2.94,2.94) {-0.75,0.25}
Total: 142024 W: 5170 L: 5202 D: 131652
Ptnml(0-2): 50, 4678, 61613, 4596, 75

Closes https://github.com/official-stockfish/Stockfish/pull/3390

Bench: 4339126

4 years agoSimplify move generation (2/2)
bmc4 [Tue, 16 Mar 2021 15:21:24 +0000]
Simplify move generation (2/2)

STC:
LLR: 2.97 (-2.94,2.94) {-1.25,0.25}
Total: 39352 W: 3551 L: 3493 D: 32308
Ptnml(0-2): 143, 2695, 13928, 2781, 129
https://tests.stockfishchess.org/tests/view/6050007a2433018de7a38bbb

LTC:
LLR: 2.96 (-2.94,2.94) {-0.75,0.25}
Total: 44944 W: 1629 L: 1596 D: 41719
Ptnml(0-2): 22, 1319, 19762, 1342, 27
https://tests.stockfishchess.org/tests/view/60500e892433018de7a38bc4

Closes https://github.com/official-stockfish/Stockfish/pull/3399

No functional change

4 years agoSimplify move generation (1/2)
bmc4 [Mon, 15 Mar 2021 22:52:45 +0000]
Simplify move generation (1/2)

STC:
LLR: 2.95 (-2.94,2.94) {-1.25,0.25}
Total: 29792 W: 2611 L: 2545 D: 24636
Ptnml(0-2): 94, 1982, 10659, 2086, 75
https://tests.stockfishchess.org/tests/view/604fe5b62433018de7a38ba8

LTC:
LLR: 2.92 (-2.94,2.94) {-0.75,0.25}
Total: 22040 W: 826 L: 777 D: 20437
Ptnml(0-2): 8, 646, 9664, 693, 9
https://tests.stockfishchess.org/tests/view/604fec892433018de7a38bac

Closes https://github.com/official-stockfish/Stockfish/pull/3399

No functional change

4 years agoIntroduce least_significant_square_bb()
bmc4 [Tue, 16 Mar 2021 19:51:31 +0000]
Introduce least_significant_square_bb()

Introducing least_significant_square_bb(). It is a function that returns a value equal
to square_bb(lsb(bb)), but it uses fewer instruction. It should speed up more on older
processors like armv7-a Clang.

Passed STC:
LLR: 2.93 (-2.94,2.94) {-0.25,1.25}
Total: 213200 W: 19171 L: 18753 D: 175276
Ptnml(0-2): 680, 14513, 75831, 14861, 715
https://tests.stockfishchess.org/tests/view/604bc7632433018de7a38982

Closes https://github.com/official-stockfish/Stockfish/pull/3391

No functional change

4 years agoRestore development version
Fabian Fichter [Sat, 13 Mar 2021 21:58:24 +0000]
Restore development version

4 years agoFairy-Stockfish 13
Fabian Fichter [Sat, 13 Mar 2021 12:41:25 +0000]
Fairy-Stockfish 13

bench: 4257608

4 years agoUpdate readme
Fabian Fichter [Fri, 12 Mar 2021 19:54:33 +0000]
Update readme

4 years agoFurther fix for placement castling
Fabian Fichter [Fri, 12 Mar 2021 18:45:31 +0000]
Further fix for placement castling

Also initialize castling king square on rook drops.

4 years agoFix castling initialization for placement chess
Fabian Fichter [Fri, 12 Mar 2021 18:20:41 +0000]
Fix castling initialization for placement chess

Refactoring of 960 castling broke placement chess castling.
Add initialization of king position to prevent undefined behavior.

4 years agoSuppress variant setup output on protocol change
Fabian Fichter [Fri, 12 Mar 2021 11:44:36 +0000]
Suppress variant setup output on protocol change

Since this might confuse GUIs, skip the variant definition output
whent he variant change is triggered by a protocol change.

4 years agoReset to starting position on protocol change
Fabian Fichter [Fri, 12 Mar 2021 09:48:45 +0000]
Reset to starting position on protocol change

Prevents issues with non-compliant GUIs not sending position command.

4 years agoAdd variant reinitialization on protocol change
Fabian Fichter [Thu, 11 Mar 2021 22:12:46 +0000]
Add variant reinitialization on protocol change

This ensures reinitialization of PSQT when changing the
default variant by switching the protocol via usi/ucci.

No functional change for UCI and CECP usage.

4 years agoSupport red as color in Xiangqi FENs
Fabian Fichter [Thu, 11 Mar 2021 19:52:50 +0000]
Support red as color in Xiangqi FENs

4 years agoChange advanced pawn push threshold
Topologist [Mon, 8 Mar 2021 18:46:41 +0000]
Change advanced pawn push threshold

A pawn push is now considered to be "advanced" if the relative destination
rank is > 6 (previously it was > 5). This affects the search heuristic.

Also remove an assert concerning en passant moves in qsearch().

STC:
LLR: 2.97 (-2.94,2.94) {-0.25,1.25}
Total: 46744 W: 4224 L: 4040 D: 38480
Ptnml(0-2): 165, 3206, 16451, 3380, 170
https://tests.stockfishchess.org/tests/view/604746082433018de7a3872e

LTC:
LLR: 2.94 (-2.94,2.94) {0.25,1.25}
Total: 107840 W: 4198 L: 3892 D: 99750
Ptnml(0-2): 58, 3472, 46557, 3772, 61
https://tests.stockfishchess.org/tests/view/60475eae2433018de7a38737

Closes https://github.com/official-stockfish/Stockfish/pull/3389

Bench: 4796780

4 years agoMerge official-stockfish/master
Fabian Fichter [Tue, 9 Mar 2021 20:58:02 +0000]
Merge official-stockfish/master

No functional change.

4 years agoMerge official-stockfish/master
Fabian Fichter [Tue, 9 Mar 2021 20:41:09 +0000]
Merge official-stockfish/master

bench: 4257608

4 years agoMerge official-stockfish/master
Fabian Fichter [Tue, 9 Mar 2021 20:33:44 +0000]
Merge official-stockfish/master

No functional change.

4 years agoMerge official-stockfish/master
Fabian Fichter [Tue, 9 Mar 2021 20:27:54 +0000]
Merge official-stockfish/master

bench: 4788083

4 years agoMerge official-stockfish/master
Fabian Fichter [Tue, 9 Mar 2021 20:19:56 +0000]
Merge official-stockfish/master

No functional change.

4 years agoAdd Fairy-Stockfish contributor list
Fabian Fichter [Mon, 8 Mar 2021 19:39:41 +0000]
Add Fairy-Stockfish contributor list

4 years agoTweak atomic pawn value
Fabian Fichter [Mon, 8 Mar 2021 18:23:34 +0000]
Tweak atomic pawn value

atomic STC
LLR: 2.98 (-2.94,2.94) [0.00,10.00]
Total: 14941 W: 5277 L: 5004 D: 4660
http://www.variantfishtest.org:6543/tests/view/60432e216e23db669974e9f5

atomic LTC
LLR: 2.98 (-2.94,2.94) [0.00,10.00]
Total: 1915 W: 676 L: 575 D: 664
http://www.variantfishtest.org:6543/tests/view/6045d4176e23db669974ea0b

4 years agoUse Bitboard over Square in movegen
bmc4 [Fri, 5 Mar 2021 11:57:43 +0000]
Use Bitboard over Square in movegen

It uses pos.checkers() on target when movegen is the type of EVASION.
It simplify the code. And it's also expected a slightly speed up,
because Bitboard is more direct when doing bitwise.

Passed STC:
LLR: 2.93 (-2.94,2.94) {-1.25,0.25}
Total: 28176 W: 2506 L: 2437 D: 23233
Ptnml(0-2): 80, 1904, 10063, 1949, 92
https://tests.stockfishchess.org/tests/view/60421d18ddcba5f0627bb6a9

Passed LTC:
LLR: 2.93 (-2.94,2.94) {-0.75,0.25}
Total: 9704 W: 402 L: 341 D: 8961
Ptnml(0-2): 3, 279, 4230, 334, 6
https://tests.stockfishchess.org/tests/view/60422823ddcba5f0627bb6ae

closes https://github.com/official-stockfish/Stockfish/pull/3383

No functional change

4 years agoDeal with commented lines in UCI input
mattginsberg [Sun, 28 Feb 2021 15:59:07 +0000]
Deal with commented lines in UCI input

commands starting with '#' as the first character will be ignored

closes https://github.com/official-stockfish/Stockfish/pull/3378

No functional change

4 years agoDo not try to use large pages on 32 bit Windows.
noobpwnftw [Wed, 3 Mar 2021 14:30:23 +0000]
Do not try to use large pages on 32 bit Windows.

verified to work on windows XP.

fixes  #3379

closes https://github.com/official-stockfish/Stockfish/pull/3380

No functional change.

4 years agoPrecalculate endgameEval flag
Fabian Fichter [Sun, 7 Mar 2021 17:42:06 +0000]
Precalculate endgameEval flag

This supersedes manual definition of the endgame evalution flag.

For a few variants the automatically calculated flag now enables
endgame evaluation where before it was explicitly disabled.

losalamos STC
LLR: 2.97 (-2.94,2.94) [-10.00,5.00]
Total: 1151 W: 292 L: 253 D: 606
http://www.variantfishtest.org:6543/tests/view/6044ea416e23db669974ea05

Closes #271.

4 years agoSupport shogi NNUE evaluation (#199)
Fabian Fichter [Sat, 6 Mar 2021 15:19:33 +0000]
Support shogi NNUE evaluation (#199)

Refactor variant NNUE handling in order to support
shogi NNUE evaluation and to improve maintainability.

Credits to @tttak for the initial implementation of this feature.

4 years agoAdd Stockfish namespace.
Dieter Dobbelaere [Fri, 26 Feb 2021 09:02:13 +0000]
Add Stockfish namespace.

fixes #3350 and is a small cleanup that might make it easier to use SF
in separate projects, like a NNUE trainer or similar.

closes https://github.com/official-stockfish/Stockfish/pull/3370

No functional change.

4 years agoClean functions returning by const values
Antoine Champion [Sat, 30 Jan 2021 08:50:04 +0000]
Clean functions returning by const values

The codebase contains multiple functions returning by const-value.
This patch is a small cleanup making those function returns
by value instead, removing the const specifier.

closes https://github.com/official-stockfish/Stockfish/pull/3328

No functional change

4 years agoFix parsing of player time in USI protocol
Fabian Fichter [Sat, 6 Mar 2021 09:52:50 +0000]
Fix parsing of player time in USI protocol

Since white and black are interchanged in the USI protocol,
the parsing of the player time needs to consider that.

Closes #269.

4 years agoFurther increase atomic PSQT bonus
Fabian Fichter [Sat, 6 Mar 2021 07:18:32 +0000]
Further increase atomic PSQT bonus

atomic STC
LLR: 2.96 (-2.94,2.94) [0.00,10.00]
Total: 5823 W: 2030 L: 1878 D: 1915
http://www.variantfishtest.org:6543/tests/view/6042003b6e23db669974e9da

atomic LTC
LLR: 2.95 (-2.94,2.94) [0.00,10.00]
Total: 2639 W: 878 L: 771 D: 990
http://www.variantfishtest.org:6543/tests/view/6042044d6e23db669974e9e0

4 years agoGeneralize doubled pawn drop setting
Fabian Fichter [Fri, 5 Mar 2021 15:33:16 +0000]
Generalize doubled pawn drop setting

Allow to prohibit dropping a doubled piece of any type,
not only restricted to shogi pawns.

Closes #266.

4 years agoFurther increase atomic PSQT bonus
Fabian Fichter [Fri, 5 Mar 2021 09:54:11 +0000]
Further increase atomic PSQT bonus

atomic STC
LLR: 2.97 (-2.94,2.94) [0.00,10.00]
Total: 3903 W: 1340 L: 1214 D: 1349
http://www.variantfishtest.org:6543/tests/view/60412a806e23db669974e9d0

atomic LTC
LLR: 2.97 (-2.94,2.94) [0.00,10.00]
Total: 3286 W: 1074 L: 959 D: 1253
http://www.variantfishtest.org:6543/tests/view/60413a0b6e23db669974e9d6

4 years agoUpdated ffish.js to 0.6.1
QueensGambit [Thu, 4 Mar 2021 22:55:36 +0000]
Updated ffish.js to 0.6.1
+ added unit test for atomic(960) castling rights from @gbtami

4 years agoOnly download NNUE net if required
Fabian Fichter [Thu, 4 Mar 2021 21:58:47 +0000]
Only download NNUE net if required

Closes #257.

4 years agoTweak atomic PSQT bonus
Fabian Fichter [Thu, 4 Mar 2021 18:40:52 +0000]
Tweak atomic PSQT bonus

atomic STC
LLR: 2.96 (-2.94,2.94) [0.00,10.00]
Total: 17388 W: 5929 L: 5630 D: 5829
http://www.variantfishtest.org:6543/tests/view/603fe73c6e23db669974e9c8

atomic LTC
LLR: 2.98 (-2.94,2.94) [0.00,10.00]
Total: 1756 W: 598 L: 501 D: 657
http://www.variantfishtest.org:6543/tests/view/6040982f6e23db669974e9cc

4 years agoBonus for advanced pieces in atomic
Fabian Fichter [Wed, 3 Mar 2021 19:39:49 +0000]
Bonus for advanced pieces in atomic

atomic STC
LLR: 2.98 (-2.94,2.94) [0.00,10.00]
Total: 2361 W: 814 L: 708 D: 839
http://www.variantfishtest.org:6543/tests/view/603e83ab6e23db669974e9b6

atomic LTC
LLR: 2.95 (-2.94,2.94) [0.00,10.00]
Total: 2720 W: 904 L: 796 D: 1020
http://www.variantfishtest.org:6543/tests/view/603ec8a26e23db669974e9bb

4 years agoBump pyffish version
Fabian Fichter [Sun, 28 Feb 2021 12:43:30 +0000]
Bump pyffish version

4 years agoUpdate readme
Fabian Fichter [Sun, 28 Feb 2021 12:41:44 +0000]
Update readme

4 years agoSupport pseudo-royal pieces
Fabian Fichter [Sun, 28 Feb 2021 11:27:27 +0000]
Support pseudo-royal pieces

This implements support for pseudo-royal pieces,
which allows to now fully support some new variants:
- lichess atomic rules
- coregal chess
- maharajah and the sepoys

Closes #81.

No functional change for other variants.

4 years agoTweak singular extension for horde
Fabian Fichter [Sat, 27 Feb 2021 23:20:07 +0000]
Tweak singular extension for horde

horde STC
LLR: -2.95 (-2.94,2.94) [0.00,10.00]
Total: 4108 W: 2008 L: 2050 D: 50
http://www.variantfishtest.org:6543/tests/view/60291c396e23db669974e90c

horde LTC
LLR: 2.95 (-2.94,2.94) [0.00,10.00]
Total: 9687 W: 4917 L: 4676 D: 94
http://www.variantfishtest.org:6543/tests/view/6029a4e46e23db669974e910

horde VLTC
LLR: 2.97 (-2.94,2.94) [0.00,10.00]
Total: 11750 W: 5938 L: 5667 D: 145
http://www.variantfishtest.org:6543/tests/view/602e57606e23db669974e923

4 years agoBumb pyffish version (#263)
Bajusz Tamás [Sat, 27 Feb 2021 21:06:58 +0000]
Bumb pyffish version (#263)

4 years agoFix assertion for atomic
Fabian Fichter [Sat, 27 Feb 2021 10:26:02 +0000]
Fix assertion for atomic

No functional change.

4 years agoIntroduce DistanceFromPV
Stéphane Nicolet [Fri, 26 Feb 2021 09:13:37 +0000]
Introduce DistanceFromPV

We introduce a metric for each internal node in search, called DistanceFromPV.
This distance indicated how far the current node is from the principal variation.

We then use this distance to search the nodes which are close to the PV a little
deeper (up to 4 plies deeper than the PV): this improves the quality of the search
at these nodes and bring better updates for the PV during search.

STC:
LLR: 2.96 (-2.94,2.94) {-0.25,1.25}
Total: 54936 W: 5047 L: 4850 D: 45039
Ptnml(0-2): 183, 3907, 19075, 4136, 167
https://tests.stockfishchess.org/tests/view/6037b88e7f517a561bc4a392

LTC:
LLR: 2.95 (-2.94,2.94) {0.25,1.25}
Total: 49608 W: 1880 L: 1703 D: 46025
Ptnml(0-2): 22, 1514, 21555, 1691, 22
https://tests.stockfishchess.org/tests/view/6038271b7f517a561bc4a3cb

Closes https://github.com/official-stockfish/Stockfish/pull/3369

Bench: 5037279

4 years agoSupport atomic NNUE
Fabian Fichter [Fri, 26 Feb 2021 17:33:40 +0000]
Support atomic NNUE

4 years agoSpeed up move generation
Fabian Fichter [Fri, 26 Feb 2021 16:52:49 +0000]
Speed up move generation

standard STC
LLR: -2.97 (-2.94,2.94) [0.00,10.00]
Total: 7654 W: 1715 L: 1705 D: 4234
http://www.variantfishtest.org:6543/tests/view/60341b906e23db669974e955

seirawan STC
LLR: -2.97 (-2.94,2.94) [0.00,10.00]
Total: 7286 W: 2480 L: 2464 D: 2342
http://www.variantfishtest.org:6543/tests/view/6034b1936e23db669974e962

minishogi STC
LLR: -2.96 (-2.94,2.94) [0.00,10.00]
Total: 9644 W: 4840 L: 4804 D: 0
http://www.variantfishtest.org:6543/tests/view/6034b1796e23db669974e960

bench: 4387985

4 years agoSpeed up slider blockers
Fabian Fichter [Fri, 26 Feb 2021 16:31:15 +0000]
Speed up slider blockers

standard STC
LLR: 2.96 (-2.94,2.94) [0.00,10.00]
Total: 1370 W: 361 L: 280 D: 729
http://www.variantfishtest.org:6543/tests/view/603423e96e23db669974e95c

seirawan STC
LLR: 2.96 (-2.94,2.94) [0.00,10.00]
Total: 2789 W: 1023 L: 908 D: 858
http://www.variantfishtest.org:6543/tests/view/603423f56e23db669974e95e

4 years agoIntroduce ProbCut for check evasions
Vizvezdenec [Sat, 20 Feb 2021 21:48:08 +0000]
Introduce ProbCut for check evasions

The idea of this patch can be described as follows: if we are in check
and the transposition table move is a capture that returns a value
far above beta, we can assume that the opponent just blundered a piece
by giving check, and we return the transposition table value. This is
similar to the usual probCut logic for quiet moves, but with a different
threshold.

Passed STC
LLR: 2.94 (-2.94,2.94) {-0.25,1.25}
Total: 33440 W: 3056 L: 2891 D: 27493
Ptnml(0-2): 110, 2338, 11672, 2477, 123
https://tests.stockfishchess.org/tests/view/602cd1087f517a561bc49bda

Passed LTC
LLR: 2.98 (-2.94,2.94) {0.25,1.25}
Total: 10072 W: 401 L: 309 D: 9362
Ptnml(0-2): 2, 288, 4365, 378, 3
https://tests.stockfishchess.org/tests/view/602ceea57f517a561bc49bf0

The committed version has an additional fix to never return unproven wins
in the tablebase range or the mate range. This fix passed tests for non-
regression at STC and LTC:

STC:
LLR: 2.93 (-2.94,2.94) {-1.25,0.25}
Total: 26240 W: 2354 L: 2280 D: 21606
Ptnml(0-2): 85, 1763, 9372, 1793, 107
https://tests.stockfishchess.org/tests/view/602d86a87f517a561bc49c7a

LTC:
LLR: 2.95 (-2.94,2.94) {-0.75,0.25}
Total: 35304 W: 1299 L: 1256 D: 32749
Ptnml(0-2): 14, 1095, 15395, 1130, 18
https://tests.stockfishchess.org/tests/view/602d98d17f517a561bc49c83

Closes https://github.com/official-stockfish/Stockfish/pull/3362

Bench: 3830215

4 years agoTune search parameters (with Unai Corzo)
Vizvezdenec [Fri, 19 Feb 2021 07:32:12 +0000]
Tune search parameters (with Unai Corzo)

The values used in this patch are taken from a SPSA parameter tuning session
originated by Unai Corzo (@unaiic), but the final difference of his tune was
multiplied x2 by hand. Most of the credits should go to him :-)

STC:
https://tests.stockfishchess.org/tests/view/602f03d07f517a561bc49d40
LLR: 2.94 (-2.94,2.94) {-0.25,1.25}
Total: 67664 W: 6252 L: 6035 D: 55377
Ptnml(0-2): 256, 4799, 23527, 4972, 278

LTC:
https://tests.stockfishchess.org/tests/view/602f41697f517a561bc49d5a
LLR: 2.96 (-2.94,2.94) {0.25,1.25}
Total: 26256 W: 1034 L: 906 D: 24316
Ptnml(0-2): 10, 804, 11377, 922, 15

Closes https://github.com/official-stockfish/Stockfish/pull/3363

Bench: 3957653

4 years agoRestore development version
Stéphane Nicolet [Sat, 20 Feb 2021 21:19:14 +0000]
Restore development version

No functional change

4 years agoEnable faster attack calculation for more variants
Fabian Fichter [Fri, 19 Feb 2021 19:34:34 +0000]
Enable faster attack calculation for more variants

seirawan STC
LLR: 2.95 (-2.94,2.94) [0.00,10.00]
Total: 311 W: 160 L: 75 D: 76
http://www.variantfishtest.org:6543/tests/view/602ff3976e23db669974e93f

capablanca STC
LLR: 2.96 (-2.94,2.94) [0.00,10.00]
Total: 285 W: 151 L: 66 D: 68
http://www.variantfishtest.org:6543/tests/view/603000616e23db669974e943

shogi STC
LLR: 2.97 (-2.94,2.94) [0.00,10.00]
Total: 635 W: 373 L: 261 D: 1
http://www.variantfishtest.org:6543/tests/view/602ff24e6e23db669974e93c

4 years agoFix attackers calculation for cambodian
Fabian Fichter [Fri, 19 Feb 2021 16:27:40 +0000]
Fix attackers calculation for cambodian

Use slow attackers calculation to consider special moves.

4 years agoSimplify fast attackers_to
Fabian Fichter [Fri, 19 Feb 2021 16:20:04 +0000]
Simplify fast attackers_to

dobutsu STC
LLR: -2.97 (-2.94,2.94) [0.00,10.00]
Total: 12830 W: 4730 L: 4640 D: 3460
http://www.variantfishtest.org:6543/tests/view/602c34f06e23db669974e914

standard STC
LLR: 3.01 (-2.94,2.94) [-10.00,5.00]
Total: 5190 W: 1157 L: 1138 D: 2895
http://www.variantfishtest.org:6543/tests/view/602cc4366e23db669974e91c

makruk STC
LLR: 3.02 (-2.94,2.94) [-10.00,5.00]
Total: 12761 W: 978 L: 966 D: 10817
http://www.variantfishtest.org:6543/tests/view/602d5ab46e23db669974e920

4 years agoStockfish 13
Joost VandeVondele [Mon, 15 Feb 2021 12:58:51 +0000]
Stockfish 13

Official release version of Stockfish 13

Bench: 3766422

-----

It is our pleasure to release Stockfish 13 to chess fans worldwide.
As usual, downloads are freely available at

https://stockfishchess.org

The Stockfish project builds on a thriving community of enthusiasts
who contribute their expertise, time, and resources to build a free
and open-source chess engine that is robust, widely available, and
very strong. We would like to thank them all!

The good news first: from now on, our users can expect more frequent
high-quality releases of Stockfish! Sadly, this decision has been
triggered by the start of sales of the Fat Fritz 2 engine by ChessBase,
which is a copy of a very recent development version of Stockfish
with minor modifications. We refer to our statement on Fat Fritz 2[1]
and a community blog[2] for further information.

This version of Stockfish is significantly stronger than any of its
predecessors. Stockfish 13 outperforms Stockfish 12 by at least
35 Elo[3]. When playing against a one-year-old Stockfish, it wins 60
times more game pairs than it loses[4]. This release features an NNUE
network retrained on billions of positions, much faster network
evaluation code, and significantly improved search heuristics, as
well as additional evaluation tweaks. In the course of its development,
this version has won the superfinals of the TCEC Season 19 and
TCEC Season 20.

Going forward, the Leela Chess Zero and Stockfish teams will join
forces to demonstrate our commitment to open source chess engines and
training tools, and open data. We are convinced that our free and
open-source chess engines serve the chess community very well.

Stay safe and enjoy chess!

The Stockfish team
[1] https://blog.stockfishchess.org/post/643239805544792064/statement-on-fat-fritz-2
[2] https://lichess.org/blog/YCvy7xMAACIA8007/fat-fritz-2-is-a-rip-off
[3] https://tests.stockfishchess.org/tests/view/602bcccf7f517a561bc49b11
[4] https://tests.stockfishchess.org/tests/view/600fbb9c735dd7f0f0352d59

4 years agoUse faster attackers generation for more variants
Fabian Fichter [Tue, 16 Feb 2021 21:09:45 +0000]
Use faster attackers generation for more variants

makruk STC
LLR: 2.98 (-2.94,2.94) [0.00,10.00]
Total: 1175 W: 127 L: 66 D: 982
http://www.variantfishtest.org:6543/tests/view/6026a0606e23db669974e8f7

minishogi STC
LLR: 2.96 (-2.94,2.94) [0.00,10.00]
Total: 3706 W: 1931 L: 1775 D: 0
http://www.variantfishtest.org:6543/tests/view/6026a0516e23db669974e8f4

standard STC
LLR: 3.00 (-2.94,2.94) [-10.00,5.00]
Total: 31166 W: 6843 L: 6953 D: 17370
http://www.variantfishtest.org:6543/tests/view/6026a0686e23db669974e8f9

4 years agoUpdate README.md
Stéphane Nicolet [Tue, 16 Feb 2021 15:19:37 +0000]
Update README.md

• reorder some sections of the README file
• add reference to the AUTHORS file
• rename Syzygybases to Syzygy tablebases
• add pointer to the Discord channel
• more precise info about the GPLv3 licence

No functional change

4 years agoSmall trivial clean-ups, February 2021
Lolligerhans [Sun, 31 Jan 2021 12:46:05 +0000]
Small trivial clean-ups, February 2021

Closes https://github.com/official-stockfish/Stockfish/pull/3329

No functional change

4 years agoUpdate Top CPU Contributors
Stéphane Nicolet [Mon, 15 Feb 2021 22:58:03 +0000]
Update Top CPU Contributors

No functional change

4 years agoAdd atomic hybrid variants to variants.ini
Fabian Fichter [Fri, 12 Feb 2021 08:52:30 +0000]
Add atomic hybrid variants to variants.ini

4 years agoPV-Nodes likely to fail low
Vizvezdenec [Thu, 11 Feb 2021 07:45:16 +0000]
PV-Nodes likely to fail low

Do not decrease reduction at pv-nodes which are likely to fail low.

The idea of this patch can be described as following: during the search, if a node
on the principal variation was re-searched in non-pv search and this re-search got
a value which was much lower than alpha, then we can assume that this pv-node is
likely to fail low again, and we can reduce more aggressively at this node.

Passed STC
https://tests.stockfishchess.org/tests/view/6023a5fa7f517a561bc49638
LLR: 2.95 (-2.94,2.94) {-0.25,1.25}
Total: 70288 W: 6443 L: 6223 D: 57622
Ptnml(0-2): 239, 5022, 24436, 5174, 273

Passed LTC
https://tests.stockfishchess.org/tests/view/6023f2617f517a561bc49661
LLR: 2.94 (-2.94,2.94) {0.25,1.25}
Total: 105656 W: 4048 L: 3748 D: 97860
Ptnml(0-2): 67, 3312, 45761, 3630, 58

Closes https://github.com/official-stockfish/Stockfish/pull/3349

Bench: 3766422

4 years agoMerge official-stockfish/master
Fabian Fichter [Thu, 11 Feb 2021 22:29:23 +0000]
Merge official-stockfish/master

bench: 4387592

4 years agoMerge official-stockfish/master
Fabian Fichter [Thu, 11 Feb 2021 22:20:23 +0000]
Merge official-stockfish/master

No functional change.

4 years agoBetter code for hash table generation
mattginsberg [Thu, 11 Feb 2021 21:29:28 +0000]
Better code for hash table generation

This patch removes some magic numbers in TT bit management and introduce proper
constants in the code, to improve documentation and ease further modifications.

No function change

4 years agoSimplify LMR for losers chess
Fabian Fichter [Wed, 10 Feb 2021 19:42:20 +0000]
Simplify LMR for losers chess

losers STC
LLR: 2.99 (-2.94,2.94) [-10.00,5.00]
Total: 1844 W: 823 L: 775 D: 246
http://www.variantfishtest.org:6543/tests/view/602303586e23db669974e8ca

losers LTC
LLR: 2.99 (-2.94,2.94) [-10.00,5.00]
Total: 1579 W: 711 L: 661 D: 207
http://www.variantfishtest.org:6543/tests/view/602389266e23db669974e8d5

4 years agoEnable New Pass Manager for Clang.
Gian-Carlo Pascutto [Tue, 9 Feb 2021 16:38:58 +0000]
Enable New Pass Manager for Clang.

It's about 1% speedup for Stockfish.

Result of 100 runs
==================
base (...fish_clang12) =    1946851  +/- 3717
test (./stockfish    ) =    1967276  +/- 3408
diff                   =     +20425  +/- 2438

speedup        = +0.0105
P(speedup > 0) =  1.0000

Thanks to David Major for making me aware of this part
of LLVM development.

closes https://github.com/official-stockfish/Stockfish/pull/3346

No functional change

4 years agoDisable ThinLTO when using Clang.
Gian-Carlo Pascutto [Tue, 9 Feb 2021 13:05:35 +0000]
Disable ThinLTO when using Clang.

Benchmarking with current Clang 12 shows that
and ThinLTO is a pessimization, see issue #3341.

closes https://github.com/official-stockfish/Stockfish/pull/3345

No functional change.

4 years agoFixes FreeBSD compilation when using Clang
Andy Pilate [Mon, 8 Feb 2021 21:02:42 +0000]
Fixes FreeBSD compilation when using Clang

closes https://github.com/official-stockfish/Stockfish/pull/3342

No functional change

4 years agoTweak move count pruning for atomic
Fabian Fichter [Tue, 9 Feb 2021 20:52:19 +0000]
Tweak move count pruning for atomic

atomic STC
LLR: 2.96 (-2.94,2.94) [0.00,10.00]
Total: 2093 W: 793 L: 686 D: 614
http://www.variantfishtest.org:6543/tests/view/601c78c06e23db669974e7c5

atomic LTC
LLR: 2.96 (-2.94,2.94) [0.00,10.00]
Total: 4343 W: 1478 L: 1347 D: 1518
http://www.variantfishtest.org:6543/tests/view/601d0bc96e23db669974e7ee

atomic VLTC
LLR: 2.98 (-2.94,2.94) [0.00,10.00]
Total: 1722 W: 594 L: 497 D: 631
http://www.variantfishtest.org:6543/tests/view/601e71686e23db669974e813

4 years agoSpeedup for variants with commoners
Fabian Fichter [Tue, 9 Feb 2021 20:39:59 +0000]
Speedup for variants with commoners

No functional change.

4 years agoSimplify antichess move picking
Fabian Fichter [Tue, 9 Feb 2021 20:13:52 +0000]
Simplify antichess move picking

No functional change.

4 years agoImprove caching of must capture state
Fabian Fichter [Tue, 9 Feb 2021 19:30:22 +0000]
Improve caching of must capture state

giveaway STC
LLR: 2.97 (-2.94,2.94) [-10.00,5.00]
Total: 473 W: 222 L: 167 D: 84
http://www.variantfishtest.org:6543/tests/view/6021b77b6e23db669974e884

giveaway LTC
LLR: 2.98 (-2.94,2.94) [-10.00,5.00]
Total: 361 W: 168 L: 114 D: 79
http://www.variantfishtest.org:6543/tests/view/6021ba026e23db669974e893

losers STC
LLR: 2.99 (-2.94,2.94) [-10.00,5.00]
Total: 387 W: 204 L: 144 D: 39
http://www.variantfishtest.org:6543/tests/view/6021b78d6e23db669974e887

losers LTC
LLR: 2.97 (-2.94,2.94) [-10.00,5.00]
Total: 690 W: 329 L: 273 D: 88
http://www.variantfishtest.org:6543/tests/view/6021b9fb6e23db669974e891

4 years agoRemove redundant legality check
Fabian Fichter [Mon, 8 Feb 2021 21:50:15 +0000]
Remove redundant legality check

No functional change.

4 years agoSimplify shallow depth pruning for anti variants
Fabian Fichter [Mon, 8 Feb 2021 21:35:16 +0000]
Simplify shallow depth pruning for anti variants

giveaway STC
LLR: 2.96 (-2.94,2.94) [-10.00,5.00]
Total: 1787 W: 698 L: 655 D: 434
http://www.variantfishtest.org:6543/tests/view/601e7a7a6e23db669974e823

giveaway LTC
LLR: 2.96 (-2.94,2.94) [-10.00,5.00]
Total: 1501 W: 550 L: 507 D: 444
http://www.variantfishtest.org:6543/tests/view/601ecee96e23db669974e834

losers STC
LLR: 2.99 (-2.94,2.94) [-10.00,5.00]
Total: 14506 W: 6366 L: 6407 D: 1733
http://www.variantfishtest.org:6543/tests/view/601e7a8f6e23db669974e827

losers LTC
LLR: 2.96 (-2.94,2.94) [-10.00,5.00]
Total: 3939 W: 1740 L: 1707 D: 492
http://www.variantfishtest.org:6543/tests/view/601fbb036e23db669974e846

4 years agoSearch Parameters Tuning
bmc4 [Fri, 5 Feb 2021 00:50:22 +0000]
Search Parameters Tuning

A simple tuning on search.cpp.

based SPSA test:
https://tests.stockfishchess.org/tests/view/601f2a787f517a561bc493cd

passed STC:
LLR: 2.94 (-2.94,2.94) {-0.25,1.25}
Total: 117840 W: 10796 L: 10508 D: 96536
Ptnml(0-2): 422, 8381, 41041, 8639, 437
https://tests.stockfishchess.org/tests/view/602144c37f517a561bc494ae

passed LTC:
LLR: 2.96 (-2.94,2.94) {0.25,1.25}
Total: 25024 W: 972 L: 847 D: 23205
Ptnml(0-2): 7, 767, 10847, 876, 15
https://tests.stockfishchess.org/tests/view/602156877f517a561bc494be

closes https://github.com/official-stockfish/Stockfish/pull/3340

Bench: 3974098

4 years agoPawns Tuning
FauziAkram [Sun, 7 Feb 2021 11:44:11 +0000]
Pawns Tuning

A simple tuning of Pawns parameters, and some PSQT changes.

Passed STC:
LLR: 2.94 (-2.94,2.94) {-0.25,1.25}
Total: 219424 W: 43681 L: 43103 D: 132640
Ptnml(0-2): 4014, 25760, 49669, 26172, 4097
https://tests.stockfishchess.org/tests/view/601bce167f517a561bc491eb

Passed LTC:
LLR: 2.94 (-2.94,2.94) {0.25,1.25}
Total: 317312 W: 42525 L: 41579 D: 233208
Ptnml(0-2): 2447, 30157, 92636, 30835, 2581
https://tests.stockfishchess.org/tests/view/601c21557f517a561bc49227

closes https://github.com/official-stockfish/Stockfish/pull/3337

Bench: 4154473

4 years agoSimplify En Passant
bmc4 [Sun, 31 Jan 2021 13:12:32 +0000]
Simplify En Passant

simplifies the handling of en passant during search, needs a little more care in initialization.

Passed STC:
LLR: 2.95 (-2.94,2.94) {-1.25,0.25}
Total: 72608 W: 6569 L: 6559 D: 59480
Ptnml(0-2): 233, 5117, 25629, 5057, 268
https://tests.stockfishchess.org/tests/view/600f1363735dd7f0f0352ce7

Passed LTC:
LLR: 2.92 (-2.94,2.94) {-0.75,0.25}
Total: 24328 W: 913 L: 864 D: 22551
Ptnml(0-2): 10, 731, 10633, 780, 10
https://tests.stockfishchess.org/tests/view/600f2e93735dd7f0f0352cf6

closes https://github.com/official-stockfish/Stockfish/pull/3330

No functional change.