Chess variant tournament: Spartan Chess

[Evert]

Catalyst's first win against Leonidas:

Pretty painful to watch.

The current development version of Leonidas thinks 1. d4 is the first mistake and it goes down hill from there...

By the way, comparing analysis by Sjaak and Leonidas shows an interesting thing: Leonidas is much faster than Sjaak up to ~17 ply or so, but at that point the branching ratio goes through the roof while Sjaak continues more or less the same way. No idea what causes that yet, but it does look as though Leonidas indeed has a bug in its search that hampers it at long time controls.

[enhorning]

[The current development version of Leonidas thinks 1. d4 is the first mistake and it goes down hill from there...

By the way, comparing analysis by Sjaak and Leonidas shows an interesting thing: Leonidas is much faster than Sjaak up to ~17 ply or so, but at that point the branching ratio goes through the roof while Sjaak continues more or less the same way. No idea what causes that yet, but it does look as though Leonidas indeed has a bug in its search that hampers it at long time controls.

Well, it had no choice about 1. d4 - that is one of the six opening moves being enforced for the tournament.

I always have a certain wariness about testing at super-blitz controls... might be okay for smaller evaluation changes, but anything likely to affect the shape of the search tree more, I would want to at least run some testing at slightly slower times.

[Evert]

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.

Interesting. Leonidas prefers Be5 from the beginning, Sjaak considers h5 briefly before switching.

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.

Indeed. It's also a headache to evaluate since they are so very different. Hoplites very easily become passers, which actually somewhat decreases the worth of a passed hoplite compared to a passed pawn in normal chess. A hoplite move always opens up a file, but two hoplites on the same file can block it very effectively (hmm... I actually don't have that evaluation term in, might be worth looking at).

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.

Well, the two spartan kings are part of what sunk the game you posted between Leonidas and Catalyst: white's major pieces were completely powerless to launch a counter attack against the black kings, while the white king was hideously exposed. They add a very interesting dynamic to the game.
Sjaak has no king safety whatsoever for the Spartans (in fact, it has the opposite and actively encourages the two kings to join in the attack), Leonidas calculates safety for both kings and discards the worst result.

I agree that the "duple check" rule is a bit arbitrary though: throughout the game I can ignore attacks on my king while I have another one, but if both are attacked I suddenly have to immediately safe one of them (while logically I could ignore it and worry about it next turn if one of my two kings is taken and the remaining one is still in check). In practice it may be a necessary thing to have because otherwise the Persians have no way at all to force an evasion (but "duple check" can't be that common). It may also not matter too much in the sense that if both my kings are under attack, I probably want to be the one who decides which of them I want to keep.

[Evert]

Well, it had no choice about 1. d4 - that is one of the six opening moves being enforced for the tournament.

Ok, I see.
Personally I don't like moves like 1. e4 and 1. d4 in Spartan. They're actually dangerous because they open up the white position before the white pieces are in play, while the Spartans develop much more quickly (most of their pieces can leap over the hoplites immediately). I'm not saying they can't be played, but I prefer something slower and quieter. It'd be interesting to see some statistics for different opening moves.

I always have a certain wariness about testing at super-blitz controls... might be okay for smaller evaluation changes, but anything likely to affect the shape of the search tree more, I would want to at least run some testing at slightly slower times.

For some things this is obvious: there's no point in testing things that affect the tree from depth 17 (say) if the time-control will only get you to depth 12 or so. Most of the changes I make are evaluation changes though (or move-ordering) and I have no idea at the moment what's causing the problem.

[hgm]

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.

Indeed, this is why it is my favorite variant. I love end-games. And there never can be a dull symmetric situation, because both sides have completely different pieces.

The duple-check rule is indeed a bit odd, but it is a nice quirk. I think the second King is a ice idea. It makes the game go through phases. with one King it becomes a completely different game than with two Kings. (Just like standard Chess becomes a completely different game after the Queens are traded.)

The two Kings are what sets Spartan Chess apart from 'regular' Chess with Different Armies. Fairy-Max plays some of these to, but I never made much work of tuning them. There seemed to be a tendency for designers to make the unconventional armies stronger than the FIDE army, probably because in human play testing people have more difficulty handling the exotic pieces.

[enhorning]

Huh, not sure what happened with the end of this game. Instead of the usual xboard adjudication message, it just says Black mates... and I don't see why white can't move Ke2 in the final position.

[Event "Computer Chess Game"]
[Site "ENHORNING"]
[Date "2013.02.10"]
[Round "35"]
[White "Pair-o-Max 4.8S"]
[Black "Sangga v1.0"]
[Result "0-1"]
[TimeControl "40/1200"]
[Variant "spartan"]
[Annotator "2. -1.12 1... +0.00"]

1. Nf3 Lf6 {+0.00/20 26} 2. c3 {-1.12/9 20} Lc6 {+0.08/17 59} 3. d4
{-1.00/9 29} Hbd5 {+0.32/17 44} 4. Nbd2 {-1.10/9 21} Hxd4 {+0.20/17 30} 5.
Nxd4 {-1.16/11 27} Lxd4 {+0.44/16 6} 6. cxd4 {-1.11/10 21} Hd5
{+0.24/17 20} 7. e3 {-1.13/10 25} Ce6 {+0.28/18 25} 8. Bd3 {-0.99/9 22}
Hxd4 {+0.64/17 23} 9. exd4 {-1.16/10 31} Ld5 {+0.64/17 28} 10. Nf3
{-0.99/10 31} Ced6 {+0.68/19 24} 11. Qc2 {-1.06/11 24} Lxf3 {+1.28/23 33}
12. gxf3 {-1.28/12 17} Wd5 {+1.28/19 35} 13. Qe2 {-1.22/12 54} Wf6
{+1.32/18 25} 14. h4 {-1.25/11 30} Cxd4 {+1.64/16 22} 15. Bc2 {-1.40/11 36}
C8d6 {+1.60/17 35} 16. Bg5 {-1.13/11 20} Wd5 {+1.60/20 30} 17. Bxh7
{-1.18/11 24} Wc4 {+1.52/17 25} 18. Qxc4 {-1.06/13 43} Cxc4 {+1.56/15 3}
19. O-O-O {-1.14/12 1:23} Kf7 {+1.52/16 33} 20. h5 {-1.25/10 19} Cb4
{+0.80/16 34} 21. b3 {-0.95/10 14} Gb5 {+0.64/14 32} 22. Bd2 {-1.17/10 16}
Cdd4 {+0.76/16 30} 23. Be3 {-1.09/10 16} Gc5+ {+1.12/19 30} 24. Bc2
{-1.72/12 19} Gc3 {+1.28/19 30} 25. Bxd4 {-0.16/14 32} Cxd4 {+0.80/20 12}
26. h6 {-0.29/13 1:53} Kg8 {+0.84/22 25} 27. h7 {-0.58/11 26} Kh8
{+1.76/20 22} 28. a4 {-1.05/11 20} Hce5 {+1.84/19 33} 29. Rh5 {-0.85/11 16}
Hf4 {+3.00/19 33} 30. Rf5 {-1.12/12 16} Kxh7 {+3.52/19 33} 31. Rf8+
{-2.30/12 31} Kh6 {+3.76/19 29} 32. Rxc8 {-1.32/13 14} Gxc8 {+3.88/19 23}
33. Rxd4 {-1.63/14 13} Hxf3 {+4.14/20 34} 34. Rh4+ {-2.00/16 39} Kg5
{+2.96/4 0.1} 35. Rh2 {-2.03/15 30} Gc3 {+4.52/18 39} 36. Kd1 {-2.37/15 40}
Hec5 {+5.28/17 39} 37. b4 {-2.42/14 11} Gxb4 {+6.06/18 39} 38. Ke1
{-2.59/14 19} Hd4 {+8.50/17 32} 39. Rg2+ {-3.57/15 15} Kf4 {+10.66/17 36}
40. Rxg7 {-4.97/15 16} Hc3 {+11.30/15 45} 41. Rf7+ {-7.27/16 35} Ke5
{+11.66/15 24} 42. Rf5+ {-8.11/16 43} Ke6 {+12.28/16 22} 43. Bd3
{-8.18/16 22} Hg2 {+15.34/14 25} 44. Rh5 {-9.56/15 51} Hb2 {+17.70/15 22}
45. Rh6+ {-9.97/14 21} Kd5 {+99.81/14 25} 46. Bf1 {-9.40/14 24} Hc1=G+
{+99.97/14 31}
{Black mates} 0-1

[hgm]

It seems there is something fundamentally wrong in Pair-o-Max after under-promotion. Like before the erroneous stalemate claim.

[enhorning]

But Winboard is set to test legality, and verify engine claims... so why does it accept the claim in this case?

[Ferdy]

But Winboard is set to test legality, and verify engine claims... so why does it accept the claim in this case?

Did one of the engines claim the mate? Or no engines have claim mate and winboard just declare it as mate?
From personal test Sangga does not claim mate in this final position since it can see that Ke2 is still legal.

[enhorning]

Did one of the engines claim the mate? Or no engines have claim mate and winboard just declare it as mate?

I don't have a clue. But when Winboard declares it a mate, the comment in the PGN is usually: "{Xboard adjudication: Checkmate} 1-0" - so I assumed it was due to one of the engines. But the PGN doesn't specifically state that one of the engine claimed it either... so I am not sure what was going on.

Cycle 3 (out of 6) has finished, and the cross-table at the half-way point is:

Code: Select all

Cross table, sorted by score percentage, Buchholz, SB

                              San Neb Spa Sja Leo Pai Fai Cat
 1. Sangga v1.0               ### 100 011 =11 110 1=1 111 111
                              ### 111 111 111 1=1 111 111 111   87%  36.5 (789.0, 646.5)

 2. NebiyuChess_1.43          011 ### 011 111 11= 101 111 111
                              000 ### 011 111 111 111 110 111   80%  33.5 (807.0, 577.8)

 3. Spartacus 0.24            100 100 ### 00= 011 111 111 111
                              000 100 ### 001 011 111 111 111   63%  26.5 (849.0, 381.3)

 4. Sjaak $Rev: 480 $         =00 000 11= ### 011 111 111 111
                              000 000 110 ### 1=0 011 1== 111   58%  24.5 (861.0, 348.3)

 5. Leonidas 8.3              001 00= 100 100 ### 101 111 101
                              0=0 000 100 0=1 ### =11 11= 011   49%  20.5 (885.0, 309.8)

 6. Pair-o-Max 4.8S           0=0 010 000 000 010 ### 0== 1=0
                              000 000 000 100 =00 ### 111 =11   29%  12.0 (936.0, 165.0)

 7. Fairy-Max 4.8Rt           000 000 000 000 000 1== ### 110
                              000 001 000 0== 00= 000 ### 1=0   19%   8.0 (960.0, 115.0)

 8. Catalyst                  000 000 000 000 010 0=1 001 ###
                              000 000 000 000 100 =00 0=1 ###   15%   6.5 (969.0,  85.0)

Sangga has pulled ahead a bit at the top, and Pair-o-Max now has 50% more points than Fairy-Max.

Pair-o-Max lost the following game from an advantageous situation:

[Event "Computer Chess Game"]
[Site "ENHORNING"]
[Date "2013.02.11"]
[Round "41"]
[White "Sjaak $Rev: 480 $"]
[Black "Pair-o-Max 4.8S"]
[Result "1-0"]
[TimeControl "40/1200"]
[Variant "spartan"]
[Annotator "2. -0.32 1... +1.35"]

1. Nf3 Wf6 {+1.35/9 53} 2. c4 {-0.32/13 21} Hde6 {+1.37/9 28} 3. Nc3
{-0.20/14 47} Hac5 {+1.35/9 23} 4. e3 {-0.17/14 37} Cd6 {+1.12/10 31} 5. d3
{-0.09/13 22} Lc6 {+1.24/10 19} 6. Nd2 {-0.20/14 39} Hge5 {+0.96/10 40} 7.
d4 {+0.23/14 38} Hxc4 {+0.72/11 20} 8. Nxc4 {+0.58/13 30} Cd7 {+0.71/11 19}
9. Be2 {+0.59/12 20} Hf4 {+0.84/9 20} 10. g3 {+0.90/13 21} Hbd5
{+0.99/10 43} 11. gxf4 {+0.43/14 39} Hxd4 {+0.86/9 15} 12. exd4
{+0.49/14 43} Hfd5 {+0.77/10 28} 13. Rg1 {+0.37/13 32} Hxd4 {+0.90/10 36}
14. Rg6 {+0.52/14 33} Wh4 {+0.85/11 18} 15. Bh5 {+0.66/12 15} Wf5
{+1.08/9 19} 16. Rg3 {+0.39/14 35} Cg8 {+1.00/10 24} 17. Bg4 {+0.39/14 40}
Wg7 {+0.79/11 35} 18. Qe2 {+0.36/13 30} Lf6 {+1.16/9 16} 19. f5
{+0.47/13 29} Hd5 {+0.85/10 22} 20. Bf4 {+0.51/13 26} Hg5 {+0.93/10 36} 21.
Bh3 {+0.62/12 22} Hg6 {+0.77/10 19} 22. Ne5 {+0.86/13 36} He3 {+0.68/12 39}
23. Rxe3 {+0.86/12 25} He4 {+0.56/12 26} 24. Rxe4 {+1.03/13 16} Lxe4
{+0.49/13 33} 25. Nxd7 {+1.03/14 38} Kxd7 {+0.31/12 17} 26. Nxe4
{+1.31/13 20} Ge8 {+0.40/11 21} 27. fxg6 {+1.57/13 19} Kc6 {+0.64/12 42}
28. Rc1 {+1.25/13 27} Cxg6 {+0.75/12 29} 29. Bxc7 {+1.29/14 40} Kd5
{+0.47/12 26} 30. f3 {+1.62/13 18} Ld4 {+0.62/10 14} 31. Rc2 {+1.63/13 29}
Wh5 {+0.66/11 26} 32. Bg2 {+1.24/13 34} Cf6 {+0.55/11 30} 33. Qb5
{+1.59/12 33} Ge6 {+0.60/10 29} 34. Rd2 {+1.77/11 27} We8 {+0.68/11 29} 35.
Qb8 {+1.23/14 47} Wc6 {+0.72/11 26} 36. Qb3 {+1.60/14 44} Lc4 {+0.65/11 29}
37. Qe3 {+1.42/12 39} Cf5 {+0.70/10 34} 38. h4 {+1.40/12 26} Ld4
{+0.73/10 12} 39. Qa3 {+1.14/13 24} Lc5 {+0.77/10 13} 40. Qc3 {-0.87/13 37}
Lb4 {+1.22/11 18} 41. Qh8+ {-2.04/12 20} Lxd2+ {+1.99/11 22} 42. Kxd2
{-2.32/13 27} Wb4+ {+2.21/11 21} 43. Kd1 {-3.35/12 34} Gc6 {+2.03/11 23}
44. Nc3+ {-3.35/15 18} Ke6 {+1.77/11 1:02} 45. Qd4 {-3.35/16 17} Wa6
{+1.77/12 26} 46. Bg3 {-3.09/14 19} Hf4 {+2.12/13 34} 47. Kc2 {-3.35/14 20}
Hxf3 {+2.91/11 24} 48. Bh3 {-3.35/15 22} He2 {+2.96/11 40} 49. Qe4
{-3.35/13 20} Hf1=W {+3.06/12 22} 50. Bxf1 {-1.28/14 21} Wc5 {+1.75/14 38}
51. Qh1 {-1.58/14 22} Cf3 {+1.93/12 41} 52. Bb5 {-3.00/13 19} We3+
{+2.90/13 36} 53. Kb1 {-3.60/16 38} Gxb5 {+2.89/13 27} 54. Nxb5
{-3.60/18 30} Wf5+ {+2.73/20 17} 55. Kc1 {-3.60/19 34} Cf1 {+2.52/19 27:16}
56. Qc6 {-3.60/19 29} Kfe7 {-0.55/1 0.1} 57. b3 {-0.21/16 45} Wxg3
{+0.95/1 0.1} 58. Qc4 {-0.45/15 40} Cf3 {+1.05/1 0.1} 59. Qa4 {-0.41/15 29}
We5 {+1.03/1 0.1} 60. Nd4 {-0.67/16 44} Wf4+ {+0.99/1 0.1} 61. Kc2
{+0.00/16 28} We3+ {+0.99/1 0.1} 62. Kb2 {+0.00/15 33} Wd1+ {+0.94/1 0.1}
63. Ka3 {+0.00/17 33} Ke5 {+0.27/1 0.1} 64. Nc6+ {+0.85/15 23} K7e6
{+0.12/1 0.1} 65. Nxe5 {+1.40/17 29} Kxe5 {+0.12/1 0.1} 66. Qa5+
{+2.57/14 20} Ke4 {+0.12/1 0.1} 67. h5 {+2.68/16 42} We3 {+0.17/1 0.1} 68.
Qb4+ {+5.43/15 27} Ke5 {+0.12/1 0.1} 69. Qe7+ {+5.63/16 26} Kf5
{+0.11/1 0.1} 70. Qf8+ {+5.59/17 36} Ke4 {+0.12/1 0.1} 71. h6 {+8.68/17 36}
Cf4 {-0.38/1 0.1} 72. Qe7+ {+9.76/16 31} Kd4 {-0.41/1 0.1} 73. Qxe3+
{+9.84/16 24} Kxe3 {+0.20/1 0.1} 74. h7 {+10.75/20 25} Ke4 {-0.41/1 0.1}
75. h8=Q {+10.87/19 24} Cf5 {-6.89/1 0.1} 76. b4 {+11.19/18 24} Ce5
{-6.88/1 0.1} 77. b5 {+159.74/17 35} Cf5 {-6.90/1 0.1} 78. b6
{+159.84/11 0.8} Ce5 {-7.52/1 0.1} 79. b7 {+159.88/10 0.5} Cf5
{-8.18/1 0.1} 80. Qh1+ {+159.92/8 0.2} Ke5 {-8.17/1 0.1} 81. b8=Q+
{+159.94/4 0.1} Kd4 {-79.97/28 2.6} 82. Qd6+ {+159.96/3 0.1} Ke3
{-79.98/28 0.1} 83. Qe1+ {+159.98/3 0.1} Kf3 {-79.99/28 0.1} 84. Qdg3#
{+160.00/2 0.1}
{Xboard adjudication: Checkmate} 1-0

At move 55, Pair-o-Max took over 27 minutes for its move, running out of time, and therefore playing 1-ply moves for the rest of the game... I thought the phenomenon where Fairy-Max / Pair-o-Max used way too much time on a move only happened when it was losing, but as this game proves, it can happen when it is ahead as well.