Announcement: Open 10x8 WB engine tournament

[hgm]

I will be running a major 10x8 Chess tournament for WinBoard engines soon, for the title of "General 10x8 Champion". General, because it will not focus on one particular subvariant, such as Capablanca or Gothic, but the engines will have to play from several opening positions.

If you are an engine author, and have written a 10x8-capable WinBoard engine, you are invited to register your engine for this tournament.

For more info, see: http://home.hccnet.nl/h.g.muller/gothad.html

[Matthias Gemuh]

I will surely update BigLion80.

Matthias.

[smrf]

I welcome the "Battle of the Goths General 10x8 Championship 2008"
by releasing a new bonus version SMIRF MS-173d-X. It promises to
become an interesting tournament using different starting arrays.

Best regards, Reinhard.

[TonyJH]

I haven't worked on my 10x8 engine recently, but I'm interested in this tourney. I'll be following it.

[hgm]

I justs started the first cycle of this tournament. The engines will play each other two times with reversed color from the Gothic opening or, if one of the enginese is not licenced to do that from ateh Gothic opening with A and C exchanged.

First match:

Smirf 1.73f-X vs. TJchess10x8

[hgm]

I am happy to say the first round went without a glitch. All engines were properly installed, and I finally figured out how to get the proper scoring in the html viewer for TJchess10x8 when it plays black. (TJchess10x8 is the only participant that gives its score as seen for white, rather than for itself.) It turns out the WinBoard option /firstScoreAbs=true (/secondScoreAbs=true) takes care of that.

One of the surprises this first round was the loss of Smirf 1.68 (the donationware version) against TSCP. Smirf is notorious for poor piece values, especially underestimating the value of the Archbishop. As a consequence it sometimes volutarily trades A for B+N, which is approximately the 10x8 Chess equivalent of trading a pair of Rooks for a pair of Bishops in normal Chess.

Code: Select all

[Event "Computer Chess Game"]
[Site "SCHAAK_PC"]
[Date "2008.04.01"]
[Round "1.3"]
[White "TSCP Gothic"]
[Black "Smirf Donation"]
[Result "1-0"]
[TimeControl "3300+5"]
[Variant "capablanca"]
[FEN "rnbqckabnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNBQCKABNR w KQkq - 0 1"]
[SetUp "1"]
[Number "6"]

{--------------
r n b q c k a b n r
p p p p p p p p p p
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
P P P P P P P P P P
R N B Q C K A B N R
white to play
--------------}
1. Na3 Na6 2. Nb1 Nb8 
{ engines start to play from here: }
3. Nc3 {+0.00/12 1:45} Nh6 {-0.14/11 1:47} 4.
d3 {-0.01/12 1:41} Nc6 {-0.07/12 2:35} 5. Nh3 {+0.08/11 1:38}
g6 {-0.07/10 1:24} 6. Bd2 {+0.18/11 1:34} Bd4 {-0.10/10 1:41} 7.
Rb1 {+0.20/11 1:31} d6 {+0.10/10 1:38} 8. Nd5 {+0.14/11 1:28}
e6 {+0.15/10 1:12} 9. Nc3 {+0.31/11 1:25} Af6 {+0.03/10 1:45} 10.
Bg5 {+0.95/11 1:22} Axg5 {+1.09/11 1:53} 11. Nxg5 {+0.79/11 1:20}
Qxg5 {+0.87/10 29} 12. Ne4 {+0.68/11 1:17} Qd5 {+1.01/11 1:30} 13.
c3 {+0.68/11 1:14} Bb6 {+1.07/10 1:29} 14. Nd2 {+0.61/11 1:12}
Qxa2 {+1.69/9 1:02} 15. Nc4 {+0.91/11 1:09} Qa6 {+1.55/10 1:03} 16.
b4 {+0.76/11 1:07} Qb5 {+1.46/10 1:33} 17. Nxb6 {+0.67/11 1:05}
axb6 {+1.46/11 2:29} 18. c4 {+0.51/11 1:03} Qf5 {+1.22/11 1:25} 19.
Cf3 {+0.43/11 1:01} Qj5 {+0.85/10 1:31} 20. Ah3 {+0.88/11 59}
Ni4 {+1.11/11 2:11} 21. Kg1 {+0.77/11 57} h5 {+1.11/10 1:21} 22.
Ch4 {+0.71/11 55} Qj6 {+1.25/11 1:14} 23. Cj3 {+0.74/11 53}
Qxj3 {+1.21/12 1:16} 24. ixj3 {+0.69/11 51} Nh6 {+1.39/11 1:11} 25.
Ri1 {+0.77/11 49} e5 {+1.39/11 1:24} 26. Ai5 {+0.87/11 48}
Cg7 {+1.33/11 1:08} 27. Qd2 {+0.81/10 46} g5 {+1.27/10 1:21} 28.
Kf1 {+0.91/10 45} Be6 {+1.56/10 1:19} 29. h4 {+1.04/10 43}
f6 {+1.05/10 1:27} 30. hxg5 {+0.90/10 42} fxg5 {+1.39/10 1:08} 31.
Ag3 {+0.95/10 40} Bf5 {+1.26/9 52} 32. Ri5 {+0.88/10 39} i6 {+1.91/10 1:11}
33. Ri1 {+0.85/10 37} Ng4 {+1.63/10 51} 34. Bi2 {+0.87/10 36}
O-O-O {+1.31/10 46} 35. c5 {+1.04/10 35} Rjf8 {+1.29/10 42} 36.
cxb6 {+0.84/10 34} cxb6 {+1.25/10 1:09} 37. e4 {+0.87/10 33}
Be6 {+1.67/10 1:00} 38. Bxg4 {+0.67/10 32} Bxg4 {+1.75/10 39} 39.
Ai5 {+0.62/10 30} Cf7 {+1.62/10 22} 40. Qa2 {+0.58/10 29} Kc7 {+1.62/9 23}
41. Rc1 {+0.58/10 28} Kb8 {+1.86/10 26} 42. Ra1 {+0.55/10 27}
Cf4 {+1.92/9 19} 43. b5 {+0.56/10 27} Na5 {+2.01/9 18} 44.
Ri3 {+0.55/10 26} Rd7 {+2.10/9 16} 45. Re3 {+0.55/10 25} Rc7 {+2.05/9 15}
46. Qa3 {+0.59/10 24} Rc5 {+1.91/9 13} 47. Qb2 {+0.61/10 23}
Rfc8 {+2.30/9 13} 48. Ag3 {+0.00/10 22} Cg6 {+2.54/9 10} 49.
Rb1 {-0.03/10 22} Rc2 {+2.25/9 13} 50. Qb4 {+0.29/10 21} R8c5 {+2.29/8 9}
51. j4 {+0.25/10 20} Cf6 {+2.26/9 9} 52. Ai5 {-0.10/10 19} Bd7 {+2.55/8 5}
53. Qe1 {-0.55/10 19} Cf4 {+2.32/9 7} 54. j3 {-1.29/10 18} j6 {+2.73/8 6}
55. Ag3 {-1.24/10 18} Cxj4 {+4.36/9 3} 56. Ah2 {-1.50/11 17}
Ch4 {+4.04/10 7} 57. Kg1 {-1.49/10 16} Rxb5 {+3.90/9 5} 58.
Rxb5 {-1.22/11 16} Bxb5 {+3.49/10 6} 59. d4 {-1.27/10 15} Nc4 {+3.88/9 4}
60. Rh3 {-1.38/11 15} Cxe4 {+4.31/10 3} 61. Qxe4 {-0.56/12 14}
Rc1+ {+4.12/9 0} 62. Af1 {-0.61/11 14} Rxf1+ {+4.11/10 4} 63.
Kh2 {-0.56/11 13} Bd7 {+3.73/10 3} 64. Rc3 {+0.33/11 13} b5 {+4.01/10 5}
65. f3 {-0.54/11 12} exd4 {+4.38/10 3} 66. Qxd4 {+0.33/11 12}
h4 {+3.83/10 2} 67. Qh8+ {+0.29/11 11} Ka7 {+3.37/11 3} 68.
Qd8 {+1.15/11 11} Be6 {+2.65/10 2} 69. Rc2 {+1.39/11 11} Ra1 {+1.88/11 2}
70. Re2 {+1.94/11 10} Bj1 {+1.28/11 4} 71. Qxg5 {+2.47/11 10}
i5 {+0.64/10 2} 72. Qxb5 {+3.40/11 10} Nb6 {+0.43/11 2} 73.
Qd3 {+3.56/11 9} d5 {+0.38/10 0} 74. Qg6 {+3.66/11 9} Na4 {-0.11/10 4} 75.
Qg7 {+4.88/11 9} Ra3 {-2.51/11 2} 76. Qc7 {+5.39/11 8} Ra1 {-3.46/11 3} 77.
Qa5+ {+7.12/11 8} Kb8 78. Re7 {+6.91/11 8} Bc8 {-5.75/11 1} 79.
Qc7+ {+7.92/11 7} Ka7 80. Qxc8 {+9.00/11 7} Rb1 {-7.68/12 4} 81.
Qc6 {+9.32/11 7} Rb4 {-9.85/12 3} 82. Qxd5 {+9.31/11 7} Rb6 {-11.10/12 1}
83. Qd4 {+13.00/11 6} Nb2 {-11.09/11 2} 84. Re1 {+15.02/11 6}
Na4 {-18.63/13 2} 85. Qxa4+ {+99.85/10 0} Ra6 {-19.29/11} 86.
Qd4+ {+99.85/1 0} Kb8 87. Re8+ {+99.85/1 0} Kc7 88. Re7+ {+99.85/1 0} Kc6
89. Rd7 {+99.85/1 0} Ra8 90. Qd5+ {+99.95/3 0} Kb6 91. Rxb7+ {+99.95/1 0}
Ka6 {-12.82/2} 92. Qb5# {+99.99/1 0}
{Xboard adjudication: Checkmate} 1-0

With 9. Af6 Smirf exposes its Archbishop to a skewer without any worry (score +0.03/10). TSCP-Gothic's score jumps to +0.95/11 as it takes the skewer with 10. ..., Bg5 (TSCP-G doesn't appreciate an Archbishop very much either, but more so than Smirf). Smirf considers this a mistake, as its score jumps to 1.09/11 when it makes the now unavoidable trade of A for B+N.

TSCP-G later sacs a Pawn to destroy Smirf's Bishop pair (a justified practice in 10x8 Chess), Smirf doubles its b-Pawn to get an open a-file, and TSCP-G keeps harrassing Smirf's early-developed Queen until Smirf finally trades it for a Chancellor (23. ..., Qxj3). The latter is a slightly unfavorable trade, which is compensated later when TSCP-G has to give its Bishop to eliminate Smirf's strong Knight (38. Bxg4).

Despite the disastrous material imbalance of A vs. B+N+P, Smirf's score remains around +2, (and TSCP-G around +0.5), and even rises to +4.36 when TSCP-G's appalling positional play allows Smirf to get a Rook on 2nd rank, and cannot hold on to one of its doubled j-Pawn (55. Cxj4). TSCP-G's score is then at -1.5, but as it underestimates the Archbishop value as well, I would say that in practice the chances are about equal at this point.

Smirf grabs another Pawn, letting one of its Rooks be traded in the process (57. Rxb5 {+3.90/9}). Then, with 60. ...,Cxe4 {+4.31/10} Smirf engages in another dubious trade, giving its Chancellor for an Archbishop and a Pawn. This makes TSCP-G's score jump by nearly 2 Pawns (from -1.38 to +0.56), while Smirf's score stays around +4 for 6 more moves. The material imbalance is than Q vs B+N+4P.


After 60. ..., Cxe4 Smirf counts on Rc1+ to gain the Archbishop
if the white Queen leaves the last rank, as after QxC, Rc1+, Af1,
RxA, KxA, it gains the Queen through the Knight fork Nd2+.

Then Smirf's score starts to drop, as its search reveals that B+N are no match for a Q, and the latter starts to gobble up the black Pawns. From there it is downhill all the way for Smirf, and TSCP-G has no problems finishing it of with such a material advantage.

[George Tsavdaris]

What was this strange sequence of moves: 1.Na3 Na6 2.Nb1 Nb8 ??

I don't understand why it has been played.....

As for the game 9...Af6?? is a big blunder that loses the game immediately.
After that TSCP Gothic didn't played ideally so it put itself into troubles, although it finally won.

Smirf's evaluation of A Vs B+N is terrible, but also bad is its endgame evaluation. 65th move +4.38 for black???? It is obvious that white has a clear advantage....

[hgm]

What was this strange sequence of moves: 1.Na3 Na6 2.Nb1 Nb8 ??

They have not been played, they were forced into the engine from a PGN file.

This is a kludge to setup initial positions in TSCP-G, as it neither supports the 'setboard', nor the 'edit' command. And it also does ignore 'variant'. It always plays from the Gothic position. So to have it play different positions, I start each game by forcing moves into both engines (and my html viewer) to create the actual initial posiiton. But unfortunately, these moves end up in the PGN as well.

Of course you might argue that in this case it was not needed, as this cycle we actually play Gothic, but some engines are not allowed to do that, and if TSCP-G plays those engines it will use a PGN file that swaps A and C in the first few moves, in stead of the moves that you see here.

One problem is that the system acts bad to a PGN that has no moves in it.
And I cannot play as /variant=gothic, as some engines do refuse that variant. So I have to play as /variant=capablanca, and then use a PGN or FEN file anyway, to load the other engines with the Gothic position and get them all in the same state as TSCP.

[hgm]

This morning there was the first irregularity of the tournament, in tha game Fairy-Max vs. Joker80. Though a miraculous coincidence, it remained without effect.

Turns out Joker80 contains an under-promotion bug. (Unlike normal Chess, under-promotions are quite important in 10x8 Chess, as Arcbishop and Chancellor are almost worth as much as a Queen.) I implemented under-promotions is Joker80's search, but forgot to implemment them in the MakeMove at game level.

This bug so far went unnoticed; although I did notice that Joker80 seemed to always promote to Queen, this need not be very suspect, as Q is the most valuable piece. And if Joker80 would actually have preferred an under-promotion, by the time the move was executad on the game board (as promotion to Q in stead), it simply adapted to the unexpected situation.

But tonight Joker80 could checkmate Fairy-Max in one through a promotion move, provided it promoted to C. The search of course found this, and printed the mate score. But the move was executed as cxd1=Q, which was not a mate. So there was no result claim, as the mate test in Joker80 is made after the move.

Exactly at this moment Fairy-Max decided to crash, so that it forfeited on time. So Joker80 won after all...

I now corrected the bug and uploaded the new version of Joker80 to my website.

[smrf]

To SMIRF's values derivations see at: http://www.chessbox.de/Compu/schachansatz1_e.html

SMIRF's weaknesses are caused by other things - its average piece values are ok.

Reinhard.