Chess variant tournament: Spartan Chess

[Evert]

But, like you say, it will be very interesting to view the games. If Spartacus would lose to Catalyst I would suspect a bug, and analyze the critical position in the game carefully, to see how it could make the fatal decision. Of course it could just be strategic ignorance of the kind that is common even in strong engines. I remember Joker losing to BremboCE, which is nearly 1000 Elo below it. With its King on a1 it allowed black Pawns on a3 (as it was not a passer; a2 was still there) and c2 (because it could push a passer itself to g7), after it was obviously doomed, although it took BremboCE some 40 more moves to discover that.

Annoying, isn't it, if a program that is clearly much better manoeuvres itself into a lost position and is better at seeing the loss than the opponent is?

That should not help that much, should it? How much can you gain? 20% faster is only 20 Elo.

I admit I haven't done a direct test, but Sjaak loses almost a factor 3 in speed as measured by time-to-depth from the opening position (in Spartan chess, it'll be much worse than that for large variants, but they actually seem to crash at the moment if I run them in 32 bit).

I'm sure I could improve performance of the 32 bit compiles, but it's never been very high on my list of priorities. There are more interesting things to do, after all.

[enhorning]

Standings after two cycles out of six:

Code: Select all

Cross table, sorted by score percentage, Buchholz, SB

                              Sang Nebi Sjaa Spar Leon Pair Cata Fair
 1. Sangga v1.0               #### 1001 =111 0111 1101 1=11 1111 1111   82%  23.0 (356.0, 267.5)
 2. NebiyuChess_1.43          0110 #### 1111 0110 11=1 1011 1111 1111   80%  22.5 (358.0, 262.5)
 3. Sjaak $Rev: 480 $         =000 0000 #### 11=1 0111 1110 1111 1111   64%  18.0 (376.0, 175.8)
 4. Spartacus 0.24            1000 1001 00=0 #### 0110 1111 1111 1111   62%  17.5 (378.0, 175.0)
 5. Leonidas 8.3              0010 00=0 1000 1001 #### 101= 1010 1111   46%  13.0 (396.0, 137.3)
 6. Pair-o-Max 4.8S           0=00 0100 0001 0000 010= #### 1=0= 0==1   29%   8.0 (416.0,  91.5)
 7. Catalyst                  0000 0000 0000 0000 0101 0=1= #### 0010   18%   5.0 (428.0,  47.0)
 8. Fairy-Max 4.8Rt           0000 0000 0000 0000 0000 1==0 1101 ####   18%   5.0 (428.0,  31.0)

Catalyst managed some more wins.

Regarding hash tables - I've set them to 512 MB in Winboard, and where the copy I had of the program came with an .ini file or similar, I've had a look at it - but there are still programs using smaller hash. Looking at the ones running right now, Spartacus has only a 68MB memory footprint, for example (while Leonidas, Fairy-Max and Pair-o-Max seem to allocate what has been set). Oberon was another one not using much memory... I haven't really checked which ones actually use all the hash memory, and which ones doesn't, but that could be causing some difference as well.

Are there any pgn-tools that can cope with working with a variant pgn-file? E.g. if I want to extract all the games of a certain player, or calculate white vs. black winning rates?

[enhorning]

Catalyst's first win against Leonidas:
[Event "Computer Chess Game"]
[Site "ENHORNING"]
[Date "2013.02.10"]
[Round "21"]
[White "Leonidas 8.3"]
[Black "Catalyst"]
[Result "0-1"]
[TimeControl "40/1200"]
[Variant "spartan"]
[Annotator "2. -0.57 1... -0.32"]

1. d4 Hbd5 {-0.32/10 25} 2. Nc3 {-0.57/13 31} Hxd4 {-0.16/9 20} 3. Qxd4
{-0.60/13 18} Lf6 {-0.07/9 19} 4. Qd3 {-0.65/13 28} Lc6 {-0.06/9 30} 5. a4
{-0.68/12 25} Hge5 {+0.03/8 15} 6. Nf3 {-0.74/12 17} Hd5 {+0.09/8 18} 7. b3
{-0.84/11 19} Hed6 {+0.23/7 16} 8. h4 {-0.74/11 32} Ce7 {+0.33/7 23} 9. Bd2
{-0.73/11 17} Hd4 {+0.34/8 23} 10. Qg6 {-0.81/12 18} Hce5 {+0.36/7 15} 11.
Rc1 {-0.89/12 33} Cde8 {+0.41/8 33} 12. e3 {-0.85/12 26} C8e6 {+0.32/9 33}
13. Qf5 {-0.83/13 33} Ce4 {+0.28/8 17} 14. Ng5 {-0.85/12 20} Kg7
{+0.49/9 21} 15. Qf3 {-0.86/12 34} C7e6 {+0.53/9 34} 16. Qg3 {-0.82/12 22}
Hg6 {+0.42/8 13} 17. f3 {-0.60/12 21} Hxg5 {+0.38/10 35} 18. fxe4
{-0.76/12 35} Hxe4 {+0.83/9 13} 19. hxg5 {-1.22/13 35} Le5 {+0.99/10 13}
20. Qh3 {-1.28/13 21} Lxc3+ {+0.99/9 37} 21. Bxc3 {-1.21/14 31} Hxe3
{+0.95/10 37} 22. Qxe3 {-1.13/13 27} Ce4 {+1.04/9 21} 23. Qg3 {-1.38/14 36}
We7 {+1.12/9 38} 24. Bd2 {-1.29/13 18} He5 {+1.19/9 38} 25. c4
{-1.20/12 21} He6 {+1.22/8 13} 26. cxd5 {-0.90/12 26} Wxd5 {+1.28/8 13} 27.
Rc3 {-1.33/12 40} Kb7 {+1.32/8 29} 28. Rc5 {-1.57/12 36} We7 {+1.46/9 19}
29. Rc3 {-1.69/12 22} Ld6 {+1.57/9 45} 30. Bc1 {-2.19/12 22} Gf8
{+1.59/8 36} 31. Rh4 {-2.33/12 33} Hf4 {+1.94/8 17} 32. Rc2 {-3.69/13 45}
Gf5 {+2.10/8 19} 33. Qh2 {-2.82/12 44} Gxg5 {+2.21/8 28} 34. Rh5
{-2.95/13 39} Gg4 {+2.25/8 26} 35. Rc3 {-2.96/13 37} Le5 {+2.85/9 1:01} 36.
Rxe5 {-3.12/14 38} Cxe5 {+2.88/9 47} 37. Bxf4 {-3.47/13 42} Wd5
{+2.88/9 1:04} 38. Bd2 {-4.32/15 49} We4 {+2.85/8 27} 39. Qg1 {-4.60/14 45}
He3 {+3.75/10 1:23} 40. Rc4 {-5.28/15 56} Wg3+ {+3.80/11 51} 41. Kd1
{-5.29/15 20} Gf5 {+4.17/11 31} 42. Bd3 {-5.25/14 16} Gf2 {+4.16/10 16} 43.
Qxf2 {-5.41/15 24} Wxf2+ {+4.14/10 26} 44. Kc2 {-5.42/16 27} Wg3
{+4.14/9 21} 45. Rg4 {-5.33/14 20} Wh5 {+4.04/10 11} 46. Rb4 {-5.35/14 24}
Kc6 {+4.11/9 16} 47. Rh4 {-5.40/14 32} Wf6 {+4.19/9 14} 48. Be1
{-5.53/13 20} Cd5 {+4.40/9 27} 49. Be4 {-5.62/12 31} Wa1+ {+5.53/10 33} 50.
Kd2 {-5.69/15 23} Wxb3+ {+5.52/10 33} 51. Kxe3 {-5.70/15 17} Wxa4
{+5.52/8 12} 52. Kf2 {-5.70/13 18} Kd7 {+5.77/9 23} 53. Kg1 {-5.57/14 25}
Ce5 {+5.58/10 24} 54. Bf3 {-5.52/15 21} Wc2 {+5.61/9 13} 55. Bf2
{-5.52/16 34} Ce3 {+5.59/9 13} 56. Rc4 {-5.46/15 29} Cg3+ {+5.57/10 37} 57.
Kh2 {-5.29/17 26} Wd3 {+5.29/11 27} 58. Rd4 {-5.59/17 35} Wf1+
{+5.36/10 18} 59. Kh1 {-5.59/1 0.1} Kc7 {+5.43/10 37} 60. Be4 {-5.40/15 35}
Hb6 {+5.43/9 23} 61. Ra4 {-5.37/15 29} Ce3 {+5.50/9 17} 62. Bd5
{-5.39/15 37} Ce2 {+5.54/9 16} 63. Rg4 {-5.39/14 37} Kf7 {+5.50/10 17} 64.
Bc5 {-5.49/15 37} Cc2 {+5.61/10 42} 65. Bg1 {-5.46/14 36} We2 {+5.63/9 24}
66. Bf3 {-5.36/15 25} Wd3 {+5.71/9 22} 67. Be4 {-4.86/17 38} We5
{+4.22/11 34} 68. Bf5 {-4.76/16 22} Kf6 {+5.20/10 46} 69. Bh2 {-5.13/17 37}
Kxf5 {+5.01/13 46} 70. Bxe5 {-6.90/17 22} Kxe5 {+4.94/13 46} 71. Rg7
{-5.76/18 28} Kc6 {+5.01/13 27} 72. Rf7 {-5.86/17 28} Ce2 {+5.01/13 16} 73.
Kg1 {-5.51/17 26} Ce4 {+5.01/13 20} 74. Kf2 {-5.73/16 26} Kc5 {+4.99/12 21}
75. Ke1 {-5.50/16 44} Kf4 {+5.04/11 28} 76. Ra7 {-6.47/17 56} Ke3
{+5.06/11 33} 77. Ra2 {-8.43/17 57} Kc4 {+5.47/14 1:13} 78. Kf1
{-8.96/18 26} Kcd3 {+5.54/14 1:01} 79. g4 {-9.85/20 1:13} Ce2
{+6.63/17 1:04} 80. Ra3 {-10.95/20 24} Kde4 {+6.62/17 1:08} 81. Rxe3+
{-13.21/21 26} Cxe3 {+6.64/18 31} 82. g5 {-13.27/23 31} Ce5 {+6.68/18 31}
83. g6 {-15.89/24 40} Cg5 {+12.96/22 11} 84. g7 {-18.83/23 35} Cxg7
{+12.97/22 32} 85. Ke1 {-19.98/24 16} Hc5 {+13.04/21 32} 86. Kd1
{-19.99/23 34} Kd4 {+163.63/19 32} 87. Kd2 {-18.94/20 32} Cg6
{+13.04/19 14} 88. Ke2 {-16.65/19 17} Hb4 {+163.65/16 13} 89. Kf3
{-159.85/18 31} Hc3 {+163.65/15 20} 90. Kf4 {-159.87/14 0.1} Cf6+
{+163.73/16 12} 91. Kf3 {-159.87/14 0.1} Hd2 {+163.75/14 23} 92. Ke2
{-159.89/11 0.2} Hc1=G {+163.73/13 13} 93. Kf2 {-159.91/10 0.1} Gb1
{+163.77/14 20} 94. Kg2 {-159.93/8 0.1} Gc2+ {+163.79/16 17} 95. Kh3
{-159.93/8 0.1} Cf4 {+163.77/15 36} 96. Kg3 {-159.94/1 0.1} Gf2+
{+163.77/16 36} 97. Kh3 {-159.96/1 0.1} Gf3+ {+163.79/15 27} 98. Kh2
{-159.96/1 0.1} Cf2+ {+163.81/16 15} 99. Kh1 {-159.99/2 0.1} Gg2#
{+163.83/17 14}
{Xboard adjudication: Checkmate} 0-1

Catalyst's win against Pair-o-Max:
[Event "Computer Chess Game"]
[Site "ENHORNING"]
[Date "2013.02.10"]
[Round "20"]
[White "Pair-o-Max 4.8S"]
[Black "Catalyst"]
[Result "0-1"]
[TimeControl "40/1200"]
[Variant "spartan"]
[Annotator "2. -1.18 1... -0.32"]

1. d4 Hbd5 {-0.32/10 26} 2. Nf3 {-1.18/10 2:43} Hxd4 {-0.18/9 25} 3. c3
{-1.52/10 2:41} He3 {+0.22/11 19} 4. fxe3 {-1.42/10 35} Lf6 {+0.31/10 30}
5. g3 {-1.27/9 17} Lc6 {+0.41/9 30} 6. Bh3 {-1.13/10 1:02} Hde6
{+0.51/9 30} 7. O-O {-1.02/10 42} Hfd5 {+0.50/9 30} 8. c4 {-1.17/9 30} He4
{+0.57/9 11} 9. Qd3 {-1.29/10 38} Hge5 {+0.55/9 31} 10. Nd4 {-1.19/9 11}
Cd6 {+0.58/8 30} 11. Nxc6 {-0.90/9 15} Cxc6 {+0.85/9 31} 12. Nc3
{-0.93/9 12} Hac5 {+0.74/8 17} 13. b3 {-0.94/10 16} Hxc4 {+0.84/9 31} 14.
bxc4 {-1.27/10 14} Hc5 {+1.32/9 22} 15. Nd5 {-1.32/10 14} Kg7 {+1.78/9 32}
16. Nxf6 {-1.08/10 17} Kxf6 {+1.80/9 12} 17. Bf5 {-1.23/10 17} Kg7
{+1.83/9 33} 18. Bd2 {-1.28/10 17} Cxc4 {+1.85/8 24} 19. Rfb1 {-1.41/10 15}
Ga7 {+2.14/8 20} 20. Rf1 {-1.46/10 15} Cf8 {+2.16/8 12} 21. Bg4
{-1.53/10 14} Cf6 {+2.18/8 14} 22. Bf3 {-1.45/10 18} We7 {+2.36/8 21} 23.
g4 {-1.44/10 30} Cd6 {+2.43/8 19} 24. Bc1 {-1.60/9 12} Cd5 {+2.45/8 26} 25.
Qb3 {-1.82/10 13} Gb8 {+2.98/10 39} 26. Qd1 {-1.99/10 12} Cc3 {+3.09/9 24}
27. Bd2 {-1.79/11 21} Ca3 {+3.07/11 40} 28. Rb1 {-2.08/12 17} Ga7
{+2.95/10 40} 29. Rf2 {-2.08/11 17} Kg8 {+3.17/9 40} 30. Rg2 {-2.07/11 14}
Cxa2 {+3.31/9 15} 31. Rg3 {-2.13/10 13} Ca4 {+3.46/9 34} 32. Ra1
{-2.33/11 37} Ca3 {+3.43/9 38} 33. Rb1 {-2.15/11 27} Ca5 {+3.39/9 44} 34.
Rg2 {-2.15/10 18} Cab5 {+3.46/9 15} 35. Ra1 {-2.10/11 37} Gb7 {+3.50/9 20}
36. Ra3 {-2.26/10 31} Kb8 {+3.54/10 55} 37. Qa1 {-2.16/10 30} Cb3
{+3.76/9 19} 38. Ra5 {-2.59/10 17} Cb1 {+4.60/10 1:06} 39. Qa4
{-2.79/11 11} Wc6 {+4.90/10 1:07} 40. Ra6 {-3.52/11 10} Wxa4 {+4.98/11 23}
41. Rxa4 {-3.95/13 21} Cb2 {+5.46/10 12} 42. Be1 {-4.15/13 20} Gb3
{+5.99/10 13} 43. Ra1 {-4.23/13 28} Cb1 {+6.04/10 23} 44. Ra6 {-4.79/14 51}
Kb7 {+6.32/10 19} 45. Ra5 {-5.37/14 31} Cc1 {+6.77/10 25} 46. Bh4
{-5.65/15 45} Cb5 {+7.62/12 33} 47. Rxb5 {-5.80/14 21} Gxb5 {+7.65/11 23}
48. Kf2 {-5.87/13 17} Cc3 {+7.65/11 33} 49. Bf6 {-6.51/13 35} Cxe3
{+7.91/10 33} 50. g5 {-6.65/13 19} Cxf3+ {+7.92/8 12} 51. Kxf3
{-7.65/14 23} Gb3+ {+8.05/10 34} 52. Kf2 {-7.02/12 25} Hcd4 {+8.39/11 17}
53. g6 {-8.50/13 29} Gh3 {+8.51/11 34} 54. gxh7 {-7.12/14 18} Kxh7
{+9.32/11 21} 55. Kf1 {-11.89/14 18} Hc3 {+12.36/12 35} 56. Rg3
{-13.27/17 35} Gxg3 {+12.42/12 35} 57. hxg3 {-13.35/18 28} Kg6
{+12.59/13 26} 58. Be7 {-12.42/12 20} Hb2 {+12.60/12 16} 59. Bc5
{-12.52/11 17} Hc1=G+ {+16.41/12 36} 60. Kg2 {-21.57/12 55} Gxc5
{+18.18/14 36} 61. Kf3 {-21.92/12 29} Gc3+ {+16.50/12 36} 62. Kg2
{-22.58/12 40} Kg5 {+18.25/12 36} 63. Kf2 {-79.95/13 19} Kg4
{+163.75/13 36} 64. Ke1 {-79.96/23 17} Hf3 {+163.77/11 36} 65. Kd1
{-79.96/22 19} Hg2 {+163.79/10 35} 66. e3 {-79.97/28 1.3} Hf1=G#
{+163.83/10 30}
{Xboard adjudication: Checkmate} 0-1

[enhorning]

Also, after two cycles, there are 43 wins for the Fide side, 9 draws, and 60 wins for the Spartan side, so Spartans are scoring 57.6% so far.

[hgm]

Are there any pgn-tools that can cope with working with a variant pgn-file? E.g. if I want to extract all the games of a certain player, or calculate white vs. black winning rates?

I don't really know. Some PGN tools do not look at the moves at all. (E.g. BayesElo.)

With WinBoard you can load the game file, and select on any text string in the game line. This could be a player name, but also any of the other info. (E.g. by filtering on 0-1 you would select all black wins.) With the View->Game List Tags dialog you can specify which tags should be displayed on the game line.

From the title bar of the game-list window you can see the WDL statistics of what you selected. There is unfortunately no way to write the selection back to a file. You can only do that for individual games. Perhaps I should add a menu item to save all selected games. Usually there is not much reason to do this, however.

As to Spartacus' hash table: Perhaps the version I sent you was from before it supported the WB memory command; I will check if there is some alternative way to set the hash size.


The Spartans vs Persians win rate is of much interest in an asymmetric game like this, where material balance cannot be taken for granted. The setup that was finally chosen was based on Fairy-Max self play, and produced 50-50 scoring there. (In an earlier design the General moved as R+N, which made the Spartans way too strong.)

[enhorning]

The Spartans vs Persians win rate is of much interest in an asymmetric game like this, where material balance cannot be taken for granted. The setup that was finally chosen was based on Fairy-Max self play, and produced 50-50 scoring there. (In an earlier design the General moved as R+N, which made the Spartans way too strong.)

Of course, the assumption is that Fairy-Max handles both sides equally well. It would be interesting to see what kind of numbers using a different program would yield.

This game was just played, in which Fairy-Max has a bishop and three pawns in the ending, and trades down to just the bishop... I wonder if Pair-o-Max could have managed something better from that situation?

[Event "Computer Chess Game"]
[Site "ENHORNING"]
[Date "2013.02.10"]
[Round "32"]
[White "Fairy-Max 4.8Rt"]
[Black "Catalyst"]
[Result "1/2-1/2"]
[TimeControl "40/1200"]
[Variant "spartan"]
[Annotator "2. -0.62 1... -0.28"]

1. Nf3 Lf6 {-0.28/10 10} 2. e3 {-0.62/9 20} Hed6 {-0.14/9 12} 3. Nc3
{-0.61/10 1:12} Lc6 {-0.11/9 27} 4. Bd3 {-0.63/10 45} Ce6 {-0.02/9 24} 5.
O-O {-0.66/10 24} Hhg6 {-0.05/9 23} 6. Bf5 {-0.72/11 46} Ce5 {+0.07/10 15}
7. Nxe5 {-0.63/12 17} Lxe5 {+0.21/11 32} 8. d4 {-0.61/11 27} Lxc3
{+0.35/10 14} 9. bxc3 {-0.67/11 23} Hbd5 {+0.35/9 32} 10. e4 {-0.53/10 17}
Hc4 {+0.44/9 13} 11. Qe1 {-0.66/10 15} We7 {+0.50/9 18} 12. f3
{-0.49/10 25} Hd5 {+0.59/9 34} 13. Be6 {-0.60/9 36} Ce8 {+0.99/10 34} 14.
Bf5 {-0.68/10 36} La4 {+1.26/8 18} 15. Rf2 {-0.73/10 16} Hxc3 {+0.77/8 34}
16. Qxc3 {-0.59/10 23} Wc6 {+0.89/8 15} 17. Be3 {-0.51/10 28} Lb4
{+0.28/8 19} 18. Qd3 {-0.55/10 19} Wb5 {+0.68/8 18} 19. c4 {-0.06/11 17}
Wa6 {-0.41/10 37} 20. exd5 {-0.01/11 56} Hxd5 {-0.43/9 37} 21. Rb1
{+0.06/11 21} Hxd4 {-0.79/9 37} 22. Qxd4 {+0.95/11 17} Cd8 {-0.76/9 37} 23.
a3 {+1.01/11 18} Lxa3 {-1.32/10 26} 24. Rxb8 {+1.31/13 34} Kxb8
{-1.08/10 37} 25. Bxd7 {+1.18/12 47} Cxd7 {-1.32/10 37} 26. Qxd7
{+1.48/12 20} Lc5 {-1.76/9 28} 27. Qd8+ {+2.30/14 17} Kb7 {-2.01/10 25} 28.
Rb2+ {+2.27/15 34} Hab6 {-2.32/10 33} 29. Bf2 {+2.47/15 34} Le7
{-2.12/10 39} 30. Qxf8 {+2.82/17 15} Lxf8 {-2.09/13 21} 31. Rxb6+
{+2.76/18 14} Ka7 {-2.09/11 33} 32. Rxg6+ {+2.75/18 15} Kb7 {-2.13/12 42}
33. Rb6+ {+2.72/19 14} Ka7 {-2.17/12 38} 34. Rc6+ {+2.45/19 2:18} Kb7
{-2.17/13 30} 35. Rxa6 {+2.39/17 10} Kxa6 {-2.21/15 20} 36. Bd4
{+2.39/13 10} Ka5 {-2.17/14 17} 37. Be5 {+1.94/13 34} Kb4 {-2.16/15 25} 38.
Bxc7 {+1.54/14 17} Kxc4 {-2.15/16 24} 39. f4 {+1.68/14 27} Kd4
{-2.09/17 1:29} 40. Kf2 {+1.80/13 8} Ke4 {-2.27/19 1:37} 41. g4
{+2.19/15 52} Lh6 {-2.37/15 15} 42. Kg3 {+2.21/16 40} Lf8 {-2.71/18 19} 43.
h4 {+2.24/15 32} Le7 {-2.74/17 30} 44. g5 {+2.26/15 33} Kf5 {-2.71/18 30}
45. h5 {+2.58/16 21} Lxg5 {-3.15/21 30} 46. fxg5 {+3.03/18 24} Kxg5
{-3.49/24 30} 47. Be5 {+3.02/22 27} Hh6 {-3.54/26 20} 48. Bf4+
{+2.93/22 29} Kxh5 {-3.59/41 13} 49. Kf3 {+2.90/24 26} Kg6 {-3.59/45 31}
50. Kg4 {+2.90/24 32} Kg7 {-3.59/39 11} 51. Kf5 {+2.89/24 21} Kf7
{-3.59/45 13} 52. Bxh6 {+2.94/26 24}
{Xboard adjudication: Insufficient mating material} 1/2-1/2

[enhorning]

Looking at the game with Sangga suggests it was still winnable with 44. h5 instead of 44. g5 : +5.24 evaluation and rising.

Whereas at move 45, Sangga's evaluation is still +4.00, but it is not increasing, and playing out moves, it's just shuffling the Bishop and the Spartan Bishop-like piece back and forth. Perhaps if I gave it more time, it would find a way to advance, but perhaps not.

And at move 46, evaluation drops to close to zero, so looks like a definite draw by then.

[hgm]

Well, I don't know if this game can be won. But Pair-o-Max would definitely reject 45.h5 at sufficient depth, as this is basically an L vs 3P trade, leaving white with no Pawns, and thus no mating potential. At d=4 it is only an L vs 2P trade, leaving B+P vs H, which is not discounted in any way. as normally (in a tactically quiet situation) this would be an easy win. h5 is untenable, though.

I guess this is basically an 'unlike Bishops' endgame, and shows the strength of the Spartan Lieutenant: because of the color-changing sideway move it can pick its color, and two Pawns down it of course switches it to the opposite color as the Bishop to create a drawish ending.

Unfortunately Pair-o-Max is not smart enough to recognize unlike Bishops situation; it only keeps track of the number of pieces per type, not what color they are on. Furthermore the Lieutenant is not really color-bound. This is really something that requires variant-specific knowledge

[Evert]

Looking at the game with Sangga suggests it was still winnable with 44. h5 instead of 44. g5 : +5.24 evaluation and rising.

Whereas at move 45, Sangga's evaluation is still +4.00, but it is not increasing, and playing out moves, it's just shuffling the Bishop and the Spartan Bishop-like piece back and forth. Perhaps if I gave it more time, it would find a way to advance, but perhaps not.

And at move 46, evaluation drops to close to zero, so looks like a definite draw by then.

Sjaak and Leonidas both prefer 44. Be5. They both evaluate the position as "dead draw" at move 46.

[enhorning]

Yeah, I let Sangga run for longer on the position... this time, it also liked Be5, and evaluation was up to +6.90 by the time I stopped it.

Talking about Spartan Chess as a variant, this is what I (from watching games) find the most intriguing about it - very rich endgames. The two sorts of pawns, along with the various possibilities of remaining pieces, give rise to lots of different positions.

In general, I find the pawn interplay between the Hoplites and the normal pawns fascinating.

The two kings, on the other hand, I don't feel contribute much... and some of the rules around them (in particular the duplicheck and the promotion allowances / restrictions), I find very counter-intuitive.

I do wonder about balance between human players as well. I wish some turn-based gaming site would support this variant - does anybody know if the designer has contacted any such sites and tried to get his variant implemented? Sites that support quite a range of chess variants include Gold Token, Brain King, Scheming Minds.