Chess variant tournaments

[Ferdy]

I tried running nebiyu alien under winboard alien selecting the crazyhouse variant, and after 2 moves nebiyu v1.42 crashed. So just ignore this engine.

[hgm]

Hmm, Winboard adjucated this as a false draw claim?


[d]8/8/8/8/6N1/B7/4n3/8 b - - 0 67

Indeed, for the more exotic variants, 'Material draws' adjudication is best switched off.

[hgm]

I guess this wouldnot really help, because it is the 'reverse' poblem. Normally WinBoard sees insufficcient material when it is not really insufficient, in the presence of weak weird pieces. But it has no rules for what constitutes insufficient material if the Kings are not orthodox Kings. So it will never accept an engine insufficient-material draw claim in Knightmate. Engines should not make such claims, or 'Verify Engine Claims' should be switched off.

Note that no GUI will be smart enough to recognize all cases where mate is no longer possible. In standard Ches, for instance, there are positions with locked Pawns (e.g. white: a4, d4, g4; black: a5, d5, g5). So the rule of thumb is that engines should not claim draws if they are not quite obvious.

[enhorning]

Yeah, I had the material checkboxes off, but verify engine claims on.

Problem is, what happens if one engine does make a false claim? These engines are, after all, far from bug free - e.g. your program Joker stopped making moves in this game:

[Event "Computer Chess Game"]
[Site "ENHORNING"]
[Date "2012.12.17"]
[Round "12"]
[White "Dabbaba 652_ja"]
[Black "Joker KM"]
[Result "0-1"]
[TimeControl "40/1200"]
[Variant "knightmate"]
[Annotator "1. +0.26 1... -0.02"]

1. d4 {+0.26/8} d5 {-0.02/14 15} 2. e4 {+0.09/7 14} dxe4 {+0.73/13 11} 3.
Bf4 {+0.07/7 8} e6 {+0.76/13 29} 4. Be2 {+0.13/7 11} Bd6 {+0.74/13 30} 5.
Bxd6 {+0.11/8 14} cxd6 {+0.79/15 28} 6. g3 {-0.12/8 16} f5 {+0.85/15 23} 7.
Mg2 {-0.10/8 11} Mf7 {+0.89/15 52} 8. Kg1 {-0.47/9 30} Kg8 {+0.86/14 32} 9.
c4 {-0.59/9 35} f4 {+1.04/14 49} 10. gxf4 {-0.14/9 28} Mg6 {+1.04/13 19}
11. f3 {-0.14/8 20} Mf5 {+1.11/14 20} 12. fxe4 {-0.08/8 18} Mxe4
{+1.04/14 33} 13. Mc2 {+0.04/8 36} Mxf4 {+1.54/13 33} 14. Qd3 {+0.06/7 15}
g6 {+1.54/13 29} 15. Qc3 {+0.11/7 27} Rf5 {+1.71/12 53} 16. Qh3
{+0.75/7 36} Qe7 {+2.03/13 44} 17. Md2 {-0.21/7 51} e5 {+2.13/13 34} 18.
Qc3 {-0.96/7 38} Qh4 {+3.28/13 34} 19. Bf3 {-1.76/6 15} Be6 {+3.56/13 34}
20. dxe5 {-1.47/8 40} Rg5 {+5.62/13 26} 21. exd6 {-1.10/6 14} Rxg2+
{+5.15/13 38} 22. Bxg2 {-1.52/11 28} Qxh2+ {+5.35/14 22} 23. Ke2
{-1.61/10 34} Qxg2+ {+5.39/14 31} 24. Kd4 {-1.48/8 20} Qe4+ {+5.20/15 50}
25. Kb5 {-1.54/9 19} Qxc4+ {+5.46/16 49} 26. Qxc4 {-2.40/11 23} Bxc4+
{+5.45/17 20} 27. Ka3 {-2.33/9 19} Bxf1 {+5.79/17 30} 28. Rxf1 {-2.48/8 26}
Mf5 {+5.62/16 22} 29. Me3 {-2.59/7 22} Mc8 {+6.48/15 14} 30. Mf4
{-3.06/9 29} Mf6 {+6.66/15 14} 31. Me4 {-3.17/8 19} Md7 {+7.06/14 14} 32.
Md5 {-3.70/8 30} h5 {+7.28/15 19} 33. Rf3 {-3.43/6 20} g5 {+7.61/14 14} 34.
Kc4 {-3.70/7 51} g4 {+8.08/14 17} 35. Rb3 {-3.85/7 48} b6 {+8.44/14 22} 36.
Ra3 {-4.37/7 42} h4 {+8.65/14 17} 37. Rd3 {-4.72/7 51} g3 {+10.40/14 12}
38. Rd1 {-4.85/7 1:15} Mf5 {+11.27/14 33} 39. Ke3 {-4.72/6 33} Mf4+
{+12.97/15 36} 40. Kg2 {-5.55/7 1:02} Mg5 {+13.64/15 9} 41. Ke1
{-5.55/5 22} h3 {+17.39/16 1:08} 42. Rc1 {-5.40/4 14} g2 {+100.09/16 31}
43. Kf3 {-16.15/5 32} h2 {+100.08/15 17} 44. Me6 {-20.72/6 2:49} Mxe6
{+100.07/13 0.1} 45. d7 {-23.68/6 1:35} Rf8+ {+100.06/11 1.6} 46. Kd4
{-299.88/5 18} g1=Q+ {+100.05/9 0.1} 47. Kb5 {-299.88/48 0.5} Qxc1
{+100.04/7 0.1} 48. d8=Q {-299.88/48 0.2}
{User adjudication} 0-1

... which is why I also have auto-flag off, this way, if something odd happens, at least it will still be there when I look, and I can make a manual adjudication.

What happens if one engine claims a mate and the other engine does not think it is one? Does Winboard accept the first engine's claim, or is the second engine queried and need to be in agreement?

Oh... and not to do with Knightmate, but while I am asking you Winboard questions... how do I play Loop Chess (which I as a human player definitely prefer to Crazyhouse - underpromotions are a much more real possibility when your opponent might capture your piece and get whatever you promote to instead of a pawn - there is also ChessGi, which I've never tried, but which is closer to Loop Chess) in Winboard? TJchess supports it (-variant loop), running with Winboard set to crazyhouse, it does play the game... but I couldn't figure out how to drop a captured pawn-promoted-to-queen, as winboard turned it into a pawn again in my holdings (and represented it as a narrow-queen while it was on the board).

[hgm]

Joker KM was a quick-and-dirty conversion of Joker, and I have not worked on Joker since 2008 (apart from adding a FEN reader for the setboard command, which does not reallty count). At the time it could not really be challenged by any other engine. I hope I still remember enough of the code to diagnose this problem and fix it. (If it is reproducible at all.)

Indeed, putting 'Verify engine claims' off would cause any claim, no matter how outrageous, to be accepted. So the best thing is probably to leave it on, and correct unjust adjudication (which will clearly stand out) by hand, just as you did. It would mess up Swiss tourneys, though, unless you made the correction also in the WB tourney file's -result string before the next round started.

As to Loop Chess:

This should be playable by specifying the additional option

/pieceToCharTable="PNBRQKpnbrqk"

(The default for Crazyhouse is "PNBRQ.......~~~~Kpnbrq.......~~~~k" indicating that NBRQ have 'shadow versions' in the second group of 11 pieces, which should be used in promotions. These shadow versions will revert to Pawn on capture, because there is no room in the holdings for these shadow piece types, but only for PNBRQ.)

Unfortunately WB 4.6.2 has a bug that only makes this work when legality testing is off. This is fixed in the WB 4.7 beta.

[enhorning]

JIndeed, putting 'Verify engine claims' off would cause any claim, no matter how outrageous, to be accepted. So the best thing is probably to leave it on, and correct unjust adjudication (which will clearly stand out) by hand, just as you did. It would mess up Swiss tourneys, though

Yeah, I'll leave it as is then. Thanks for the clarification. (... and I don't really get enough engines to need to do a Swiss tourney, so should be fine with regards to that.)

As to Loop Chess:

This should be playable by specifying the additional option

/pieceToCharTable="PNBRQKpnbrqk"

Unfortunately WB 4.6.2 has a bug that only makes this work when legality testing is off. This is fixed in the WB 4.7 beta.

Great, that worked fine (with legality testing off)!

[enhorning]

So, because I was curious, I dropped the bottom four program, and am running the rest again, in another double round-robin, but with twice the thinking time (40 minutes / 40 moves, or 1 min average / move)... and so far, at the halfway point, it is painting a somewhat different picture:

Code: Select all

Cross table, sorted by score percentage, Buchholz, SB

                              L G T J S N S F J
 1. Lime KM 621_ja            # 0 1 1 1 1 1 1 1   87%   7.0 ( 29.0,  24.0)
 2. Gerbil KM 0.2_ja          1 # 0 1 1 0 0 1 1   62%   5.0 ( 31.0,  20.0)
 3. Tadyak 1.1                0 1 # 0 0 1 1 1 1   62%   5.0 ( 31.0,  16.5)
 4. Joker KM                  0 0 1 # 1 0 1 1 0   50%   4.0 ( 32.0,  14.5)
 5. SamChess KM 1.2           0 0 1 0 # 1 = 0 1   44%   3.5 ( 32.5,  12.0)
 6. NebiyuChess_1.43          0 1 0 1 0 # 0 0 1   37%   3.0 ( 33.0,  11.5)
 7. Sjaak r470                0 1 0 0 = 1 # 0 =   37%   3.0 ( 33.0,  11.0)
 8. Faile KM 144_ja           0 0 0 0 1 1 1 # 0   37%   3.0 ( 33.0,   9.5)
 9. Jabba 1.0 KM              0 0 0 1 0 0 = 1 #   31%   2.5 ( 33.5,   8.5)

Tadyak has so far not managed to be as dominating as it previously was (it scored 96% in the previous double round-robin!)... whereas Lime and Gerbil has managed to score higher percentages despite facing on average tougher opposition than before.

[enhorning]

So, because I was curious, I dropped the bottom four program, and am running the rest again, in another double round-robin, but with twice the thinking time (40 minutes / 40 moves, or 1 min average / move)

... and here is the final cross-table for that:

Code: Select all

Cross table, sorted by score percentage, Buchholz, SB

                              Li Ta Jo Ge Sa Ne Fa Ja Sj
 1. Lime KM 621_ja            ## 10 11 00 10 11 11 11 11   75%  12.0 (120.0,  84.0)
 2. Tadyak 1.1                01 ## 00 11 01 11 10 11 11   69%  11.0 (122.0,  74.0)
 3. Joker KM                  00 11 ## 01 11 01 11 0= 11   66%  10.5 (123.0,  77.0)
 4. Gerbil KM 0.2_ja          11 00 10 ## 10 00 11 11 0=   53%   8.5 (127.0,  66.3)
 5. SamChess KM 1.2           01 10 00 01 ## 1= 00 11 =1   50%   8.0 (128.0,  56.0)
 6. NebiyuChess_1.43          00 00 10 11 0= ## 01 11 01   47%   7.5 (129.0,  50.0)
 7. Faile KM 144_ja           00 01 00 00 11 10 ## 01 11   44%   7.0 (130.0,  45.5)
 8. Jabba 1.0 KM              00 00 1= 00 00 00 10 ## =1   25%   4.0 (136.0,  28.0)
 9. Sjaak r470                00 00 00 1= =0 10 00 =0 ##   22%   3.5 (137.0,  26.3)

Next up is the Crazyhouse tournament... but that has its own thread.