Chess variant tournament: Capablanca and similar

[enhorning]

Been a bit busy lately, but I am hoping to start this tournament towards next weekend.

Game will be Capablanca chess, but with a few different starting setups (so participating program will need to be able to accept a FEN with the starting position from Winboard).

Time control and hardware will be the usual - 40 moves in 20 minutes, single thread on i-970, 512 MB hash.

I've got the following programs that are supposed to support Capablanca (I haven't actually tested any yet - not even sure all are WB compatible, and in particular, some of these programs might not accept different starting positions):

Fairy-Max
Bihasa (Deuterium derivative)
Nebiyu
Sjaak
ChessV
ArcBishop
BigLion
Joker80
Smirf
TJChess

Anybody know of any other 10x8 Capablanca programs that I should include?

Next tournament after this one will be Makruk (Thai Chess) without the endgame counting rules (as that's the way WB does it and the way most (all?) programs handle it).

[hgm]

TSCP-Gothic play Gothic Chess. It dos not implement setting up positions, but I hacked the binary. So I can provide you with versions for arbitrary initial setups. (It contains a FEN string of the initial position, and I made a program to overwrite that with another FEN string. The new FEN must not be longer than the old, but if you only permute back-rank pieces that shouldn't be a problem.)

Later I also obtained the source code from its author, but I never tried to compile it. (It was i the form of an MSVC project, and I use gcc/Cygwin.)

TSCP-Gothic ignores the variant command, just like ChessV. But as it is a WB v1 engine that should not be a problem if WB is already set for the variant at the time it starts up TCSP-G. The also both ignore setting up of positions; in ChessV this is solved by running it with an SGF file as argument, specifying the variant and position (and possibly renaming the pieces).

Spartacus should play variant capablanca and gothic, and supports setboard. (The same version as you used in the Spartan tourney.) The same holds for the Pair-o-Max version you already have.

I have made attempts to upgrade Pair-o-Max to use real move sorting (and convert the source to more verbose variable names). But there must still be something wrong in this version, as so far it did not manage to win a single game against Fairy-Max 4.8S. (although it was close once, where it drew a KRNK end-game by the 50-move rule). Initially it searches 2 ply deeper than Fairy-Max (8 ply vs 6 at 40/min), but in the early end-game Fairy-Max' search depth shoots up to 16 ply, while this new version stays at 8. I will try if I can get this one to work; perhaps it could replace Pair-o-Max in that case.

[enhorning]

TSCP-Gothic play Gothic Chess. It dos not implement setting up positions, but I hacked the binary. So I can provide you with versions for arbitrary initial setups.

Okay, I am leaning towards running from these four setups:
[FEN "rnabqkbcnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq - 0 1"]
(Capablanca chess)

[FEN "rnbackqbnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNBACKQBNR w KQkq - 0 1"]
(Embassy chess, but mirrored to keep the King on the same square)

[FEN "rqnbakbncr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq - 0 1"]
(The starting position for Schoolbook chess, but without the Schoolbook castling rules)

[FEN "rbbnnkcaqr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq - 0 1"]
(Unbalanced position with all the power pieces on one side of the King)

TSCP-Gothic ignores the variant command, just like ChessV. But as it is a WB v1 engine that should not be a problem if WB is already set for the variant at the time it starts up TCSP-G. The also both ignore setting up of positions; in ChessV this is solved by running it with an SGF file as argument, specifying the variant and position (and possibly renaming the pieces).

If you can supply me with a sample SGF file for one of the starting positions, I can use that as a template, and include ChessV.

Spartacus should play variant capablanca and gothic, and supports setboard. (The same version as you used in the Spartan tourney.) The same holds for the Pair-o-Max version you already have.

The Spartacus I used in the tournament (Sparta.exe, you emailed it to me on Feb 9th) only plays Fide and Spartan chess. The version you sent me before that one (SpartaN2.exe, you emailed it to me on Feb 6th) does support Capablanca chess, but has a crashing problem (which was why you sent me another version for the Spartan chess tournament) - I just had it crash in a Capa test game.

I have made attempts to upgrade Pair-o-Max to use real move sorting (and convert the source to more verbose variable names). But there must still be something wrong in this version, as so far it did not manage to win a single game against Fairy-Max 4.8S. (although it was close once, where it drew a KRNK end-game by the 50-move rule). Initially it searches 2 ply deeper than Fairy-Max (8 ply vs 6 at 40/min), but in the early end-game Fairy-Max' search depth shoots up to 16 ply, while this new version stays at 8. I will try if I can get this one to work; perhaps it could replace Pair-o-Max in that case.

If you do get it to work, feel free to send it to me.

[enhorning]

[FEN "rqnbakbncr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq - 0 1"]
(The starting position for Schoolbook chess, but without the Schoolbook castling rules)

[FEN "rbbnnkcaqr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq - 0 1"]
(Unbalanced position with all the power pieces on one side of the King)

Those last two FENs were wrong - the black pieces were correct, but I had forgotten to change the white pieces to match. Here are the correct ones:

[FEN "rqnbakbncr/pppppppppp/10/10/10/10/PPPPPPPPPP/RQNBAKBNCR w KQkq - 0 1"]

[FEN "rbbnnkcaqr/pppppppppp/10/10/10/10/PPPPPPPPPP/RBBNNKCAQR w KQkq - 0 1"]

[enhorning]

Ran a quick blitz test tournament (using rbanbkncqr as the back row) with the 10 programs I have working, and all of them seemed to accept the setup command with no trouble. So, that means we should have at least 10 programs participating, and more if Spartacus / ChessV / TSCP-Gothic can be sorted out.

Code: Select all

Cross table, sorted by score percentage, Buchholz, SB

                              Bi TJ Ne Jo Fa Sm Sj Pa Bi Ar
 1. Bihasa v3.2               ## 01 11 11 11 11 11 11 11 11   94%  17.0 (146.0, 130.5)
 2. TJchess10x8 1.1-x64       10 ## 10 11 =1 11 11 11 11 11   86%  15.5 (149.0, 112.5)
 3. NebiyuChess_1.43          00 01 ## 11 11 11 11 11 11 11   83%  15.0 (150.0, 100.5)
 4. Joker80.np                00 00 00 ## 1= 11 11 01 11 11   58%  10.5 (159.0,  54.5)
 5. Fairy-Max 4.8S            00 =0 00 0= ## 10 01 11 11 11   50%   9.0 (162.0,  44.0)
 6. Smirf BC-173g-X           00 00 00 00 01 ## 01 11 11 11   44%   8.0 (164.0,  32.0)
 7. Sjaak 524                 00 00 00 00 10 10 ## 01 11 11   39%   7.0 (166.0,  28.0)
 8. Pair-o-Max 4.8S           00 00 00 10 00 00 10 ## 10 11   28%   5.0 (170.0,  20.5)
 9. BigLion80 2.23x WB        00 00 00 00 00 00 00 01 ## 11   17%   3.0 (174.0,   5.0)
10. ArcBishop80 1.01 WB       00 00 00 00 00 00 00 00 00 ##    0%   0.0 (180.0,   0.0)

Smirf would cause a popup ("Note: Refused own move") as black, but seemed to play the games fine despite that.

[Daniel Shawul]

I have put up a perft page for the chess variants I have in NebiyuChess towards end of this page https://sites.google.com/site/dshawul/. I have checked some of them against sjaak's results but most are unverified. Authors let me know if you see some errors.
As to the competition, I will try to add some evaulation to try and prevent Pedro's engine from having a field day. It should have had some eval already since it is not an part of alien branch.

[Evert]

I have put up a perft page for the chess variants I have in NebiyuChess towards end of this page https://sites.google.com/site/dshawul/. I have checked some of them against sjaak's results but most are unverified.

The same is true of Sjaak's results, of course - part of the reason I put them up in the first place (and because it annoyed me that something similar wasn't available).
I'll see about updating the webpage with perft results from other variants so we have more of a comparison.

Unfortunately I won't have time to try to tweak Sjaak's evaluation to make it a bit stronger...

[Ajedrecista]

Hello Daniel and Evert:

I have put up a perft page for the chess variants I have in NebiyuChess towards end of this page https://sites.google.com/site/dshawul/. I have checked some of them against sjaak's results but most are unverified. Authors let me know if you see some errors.
As to the competition, I will try to add some evaulation to try and prevent Pedro's engine from having a field day. It should have had some eval already since it is not an part of alien branch.

Bihasa 3.2 by Ferdinand Mosca has the perft command. Bihasa agrees in Perft(6) of both Capablanca and Gothic variations.

Capablanca variation:

Code: Select all

perft 6

8  r n a b q k b c n r
7  o o o o o o o o o o
6  . . . . . . . . . .
5  . . . . . . . . . .
4  . . . . . . . . . .
3  . . . . . . . . . .
2  P P P P P P P P P P
1  R N A B Q K B C N R

   a b c d e f g h i j

rnabqkbcnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNABQKBCNR w KQkq -


 i2i3 - 24802715
 i2i4 - 26996704
 h2h3 - 31717791
 h2h4 - 35870856
 a2a3 - 28701556
 a2a4 - 32556619
 e2e3 - 43107759
 e2e4 - 46818840
 c2c3 - 30363782
 c2c4 - 32155480
 j2j3 - 28588868
 j2j4 - 32465690
 b2b3 - 28126504
 b2b4 - 30369092
 d2d3 - 46397175
 d2d4 - 48325852
 f2f3 - 52623654
 f2f4 - 52678872
 g2g3 - 27723215
 g2g4 - 29988375
 b1a3 - 28723639
 b1c3 - 32058743
 i1h3 - 33345105
 i1j3 - 30250326
 c1b3 - 35660984
 c1d3 - 39217127
 h1g3 - 53499031
 h1i3 - 52668083

 Perft 6 Summary
 Count: 1015802437
 Time: 301.81 sec

------------------------

Gothic variation:

Code: Select all

perft 6

8  r n b q c k a b n r
7  o o o o o o o o o o
6  . . . . . . . . . .
5  . . . . . . . . . .
4  . . . . . . . . . .
3  . . . . . . . . . .
2  P P P P P P P P P P
1  R N B Q C K A B N R

   a b c d e f g h i j

rnbqckabnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNBQCKABNR w KQkq -


 f2f3 - 34872810
 f2f4 - 39637679
 b2b3 - 30383101
 b2b4 - 31008755
 a2a3 - 27166638
 a2a4 - 30975700
 c2c3 - 32854158
 c2c4 - 35099755
 i2i3 - 32395645
 i2i4 - 32792521
 e2e3 - 44387071
 e2e4 - 47620503
 h2h3 - 28638475
 h2h4 - 32192319
 j2j3 - 27192885
 j2j4 - 30994639
 g2g3 - 43282666
 g2g4 - 42731248
 d2d3 - 39464276
 d2d4 - 43835739
 b1a3 - 28942056
 b1c3 - 32136750
 i1h3 - 30322738
 i1j3 - 28933894
 g1f3 - 44620553
 g1h3 - 37809042
 e1d3 - 59098582
 e1f3 - 55839014

 Perft 6 Summary
 Count: 1025229212
 Time: 324.66 sec

------------------------

Ferd has more variant engines in his personal site:

Sangga can play Berolina chess.

------------------------

Tadyak 1.0 (there is a 1.1 version) can play Knightmare chess:

Code: Select all

perft 6

8  r m b q k b m r
7  o o o o o o o o
6  . . . . . . . .
5  . . . . . . . .
4  . . . . . . . .
3  . . . . . . . .
2  P P P P P P P P
1  R M B Q K B M R

   a b c d e f g h


 f2f3 - 2911930
 f2f4 - 3287482
 g2g3 - 4232616
 g2g4 - 4086545
 a2a3 - 3551382
 a2a4 - 3932115
 b2b3 - 3867050
 b2b4 - 3901827
 c2c3 - 4586617
 c2c4 - 4503549
 d2d3 - 4438701
 d2d4 - 5425047
 e2e3 - 5872106
 e2e4 - 5788856
 h2h3 - 3548296
 h2h4 - 3965844
 e1d3 - 3278311
 e1f3 - 3390709

 Perft 6 Summary
 Nodes: 74568983
 Time: 28.45 sec
 Nodes/sec: 2620778

------------------------

Sugod 1.0 can play Berolina chess:

Code: Select all

perft 6

8  r n b q k b n r
7  o o o o o o o o
6  . . . . . . . .
5  . . . . . . . .
4  . . . . . . . .
3  . . . . . . . .
2  P P P P P P P P
1  R N B Q K B N R

   a b c d e f g h


 g2f3 - 25564123
 g2e4 - 28939805
 g2h3 - 25540855
 a2b3 - 36675142
 a2c4 - 38045858
 b2a3 - 25576566
 b2c3 - 25550030
 b2d4 - 28296775
 c2b3 - 24219580
 c2a4 - 27635999
 c2d3 - 27130497
 c2e4 - 29366193
 d2c3 - 42157116
 d2b4 - 47943923
 d2e3 - 36259045
 d2f4 - 39616022
 e2d3 - 33948623
 e2c4 - 37925078
 e2f3 - 33661156
 e2g4 - 41182235
 f2e3 - 23916436
 f2d4 - 24825307
 f2g3 - 24069667
 f2h4 - 25429381
 h2g3 - 36834945
 h2f4 - 38059646
 b1a3 - 30560409
 b1c3 - 29016218
 g1f3 - 29161739
 g1h3 - 30622502

 Perft 6 Summary
 Nodes: 947730871
 Time: 282.97 sec
 Nodes/sec: 3349239

------------------------

Tiyaga 1.0 can play Shatranj:

Code: Select all

perft 6

8  r n b k q b n r
7  o o o o o o o o
6  . . . . . . . .
5  . . . . . . . .
4  . . . . . . . .
3  . . . . . . . .
2  P P P P P P P P
1  R N B K Q B N R

   a b c d e f g h


 g2g3 - 1163857
 a2a3 - 1062225
 b2b3 - 1163900
 c2c3 - 1171567
 d2d3 - 1470254
 e2e3 - 1318410
 f2f3 - 1178010
 h2h3 - 1066655
 b1c3 - 1657184
 b1a3 - 1217310
 g1h3 - 1211844
 g1f3 - 1637880
 c1a3 - 1061532
 c1e3 - 1333680
 f1d3 - 1200263
 f1h3 - 950138

 Perft 6 Summary
 Nodes: 19864709
 Time: 5.72 sec
 Nodes/sec: 3473458

------------------------

Makulit 1.0 can play Xiangqi.

------------------------

Here are other variants that I have not seen in your Perft summary pages.

Imortal 1.0 plays Crazyhouse chess:

Code: Select all

perft 6

8  r n b q k b n r
7  o o o o o o o o
6  . . . . . . . .
5  . . . . . . . .
4  . . . . . . . .
3  . . . . . . . .
2  P P P P P P P P
1  R N B Q K B N R

   a b c d e f g h


 h2h3 - 4508005
 h2h4 - 5464230
 a2a3 - 4508366
 a2a4 - 5419462
 g2g3 - 5371637
 g2g4 - 5374782
 b2b3 - 5335508
 b2b4 - 5426953
 d2d3 - 8179173
 d2d4 - 9106168
 f2f3 - 4425331
 f2f4 - 4981464
 e2e3 - 9858093
 e2e4 - 10024249
 c2c3 - 5442736
 c2c4 - 5967648
 b1a3 - 4914167
 b1c3 - 5776620
 g1f3 - 5794066
 g1h3 - 4934284

 Perft 6 Summary
 Nodes: 120812942
 Time: 46.58 sec
 Nodes/sec: 2593777

------------------------

Sigla 1.0 can play Seirawan chess.

------------------------

Liit 1.0 can play Los Álamos chess.

------------------------

All Perft(6) that I ran agree with your results if I am not wrong. I hope that it will be useful for you.

Good luck with the tournament!

Regards from Spain.

Ajedrecista.

[Daniel Shawul]

Thanks for the verification Jesus!

Bihasa 3.2 by Ferdinand Mosca has the perft command. Bihasa agrees in Perft(6) of both Capablanca and Gothic variations.

Silly me, I meant Ferdinand's engines. I didn't know they have perft though. It is very helpful to have those numbers for authors. TJchess has also some perfts for crazyhouse and shogi, which i verified against. I still haven't put up the variants in NebiyuAlien so will do that later.

Makulit 1.0 can play Xiangqi.

Xiangqi's many rules frustrated me but it seems now fairymax, sjaak and makulit all can play it so I will try again.

Cheers
Daniel

[Daniel Shawul]

The same is true of Sjaak's results, of course - part of the reason I put them up in the first place (and because it annoyed me that something similar wasn't available).

Exactly the reason why I wanted to put those up. I had trouble verifying perft for mini-shogi variants, because even TJchess don't have those. I asked HG for perfts but he doesn't believe they are helpful

I'll see about updating the webpage with perft results from other variants so we have more of a comparison.

Great. I will do the same for some variants i have.