SjaakII 1.0 RC1

[Evert]

I won't pretend that its king safety evaluation is in any way clever (or specifically designed with Shogi in mind) but it should be able to do better than that.

Well, in all fairness I should say it usually it does a lot better. But it does have a general tendency to play up a Pawn, and then put the King in the hole that makes on the 3rd rank, in 9x9 Shogi. This seems very bad, due to complete diagonal exposure, plus exposure to forward Knight checks.

It may be under the delusion that it's building a castle for itself. In good news, it does actually see that the king safety is bad:

Code: Select all

#[shogi] 0w (f)>force setboard l2g4l/2rsgk3/pp1p1pnp1/2p1p3p/3K1sp1b/2PPP4/PP2BP1PP/4R4/LNSG1GSNL[PN] w 2 22
 9l     g         l 
 8    r s g k       
 7p p   p   p n p   
 6    p   p       p 
 5      K   s p   b 
 4    P P P         
 3P P     B P   P P 
 2        R         
 1L N S G   G S N L *
  a b c d e f g h i 
White holdings [ P: 01 N: 01 ]
Black holdings [ ]
#[shogi] 0w (f)>eval
Component:       MG:    EG:
Material:         500    500
Piece square:     -58     42
Mobility:          -9     -2
Pawn structure:     0    -64
King safety:      -87      0
Static evaluation: 348
#[shogi] 0w (f)>

It just needs to be ramped up more. Or the material needs to be downscaled a lot.

[hgm]

The following EuroShogi game is remarkable because Sjaak seems to grind Shokidoki completely and methodically to pieces (+16...), and then in a few moves gives it all away and eventually loses. I don't know if this points to some error, or that this is just typical of Shogi.

Code: Select all

[Event "Computer Chess Game"]
[Site "hgm-xboard"]
[Date "2015.02.19"]
[Round "4"]
[White "Shokidoki tori3"]
[Black "Sjaak II 1.0 RC4"]
[Result "1-0"]
[TimeControl "40/300"]
[Variant "euroshogi"]
[VariantMen "P:fW;N:sWffN;B:B;R:R;G:WfF;+P:WfF;+N:WfF;+B:BW;+R:RF;K:K;+p:WfF;+n:WfF;+b:BW;+r:RF"]
[FEN "1nbgkgn1/1r4b1/pppppppp/8/8/PPPPPPPP/1B4R1/1NGKGBN1[-] w 0 1"]
[SetUp "1"]

{--------------
. n b g k g n .
. r . . . . b .
p p p p p p p p
. . . . . . . .
. . . . . . . .
P P P P P P P P
. B . . . . R .
. N G K G B N .
white to play
--------------}
1. Gc2 {+0.32/14} f5 {+0.01/10 7} 2. Kd2 {+1.24/14 7} c5 {+0.30/9 5} 3. Nc1
{+1.32/12 6} Rc7 {+0.35/9 8} 4. g4 {+1.88/12 6} Bf6 {+0.41/8 4} 5. Nd1
{+1.84/12 4} Bb7 {+0.42/8 8} 6. Be2 {+1.72/12 4} Bd5 {+0.71/8 7} 7. Nf1
{+2.08/11 4} f4 {+0.62/7 7} 8. Rf2 {+2.64/12 4} fxf3 {+0.63/7 5} 9. Bxf3
{+2.64/12 10} Bxf3 {+0.82/8 8} 10. Rxf3 {+2.64/10 2.7} B@c6 {+0.45/8 8} 11.
Rf5 {+3.12/9 3} Be5 {+1.59/7 5} 12. g5 {+3.20/9 7} P@f6 {+1.26/7 6} 13. Rf2
{+3.04/11 7} gxg5 {+1.45/8 6} 14. B@g6 {+2.72/10 6} Gf7 {+1.39/8 18} 15.
Bh7+ {+2.12/12 9} Gg7 {+0.69/8 8} 16. e4 {+1.72/11 15} Gxh7 {+3.04/8 6} 17.
exe5 {+1.72/9 1.7} exe5 {+2.61/8 7} 18. Ge2 {+1.36/9 6} B@g2 {+4.97/8 10}
19. Ng1 {+1.68/9 4} Bgd5+ {+4.69/8 6} 20. Nf3 {+0.96/9 14} g4 {+4.93/7 5}
21. Ge3 {+1.24/8 4} +Be6 {+5.07/7 9} 22. Rg2 {+0.68/9 11} Rg7 {+5.06/7 8}
23. Ge2 {-0.24/9 4} g4g3=+ {+7.08/7 4} 24. Rxg3 {-0.84/11 9} Rxg3
{+6.20/8 7} 25. Nxg3 {-0.44/10 14} R@g2 {+6.68/7 6} 26. B@f2 {-1.04/9 7}
Rh2+ {+6.68/6 6} 27. Nf3 {-2.00/9 6} P@g4 {+6.78/6 5} 28. R@a7 {+0.44/8 4}
Gd7 {+7.30/7 9} 29. P@e4 {-2.32/8 7} exe4 {+8.35/7 12} 30. d4 {-4.16/8 8}
Nf8 {+8.42/6 7} 31. d5 {-2.48/7 5} Bxd5 {+8.91/5 4} 32. P@f7 {-2.68/6 4}
Ng8 {+9.79/5 6} 33. h4 {-4.36/7 4} Bc6 {+10.11/5 8} 34. h5 {-4.48/7 5} hxh5
{+11.28/5 5} 35. b4 {-5.96/7 5} h4 {+13.76/7 15} 36. b5 {-7.76/7 4} bxb5
{+14.94/6 7} 37. Bxc5 {-9.40/7 4} h4h3=+ {+15.27/6 9} 38. Ng5 {-8.92/7 5}
Gh6 {+15.58/6 6} 39. P@g7 {-8.52/8 7} Nh8 {+16.84/6 10} 40. f8+
{-8.48/9 12} Kxf8 {+15.32/6 5} 41. Rxd7+ {-8.56/8 1.8} Bxd7 {+17.72/7 11}
42. Bxd6+ {-11.36/9 12} Kxg7 {+13.40/6 5} 43. G@f7 {-11.52/9 35} +Bxf7
{+14.56/6 9} 44. Nxf7+ {-400.00/9 43} Kxf7 {+15.76/5 6} 45. B@c4
{-12.68/8 10} Kg6 {+15.19/5 8} 46. +Bxh2 {-12.92/8 16} +Pxh2 {+15.44/6 6}
47. R@f8 {-12.60/7 13} R@d4 {+14.93/5 8} 48. Gcd3 {-14.52/8 10} Rxc4
{+15.41/6 8} 49. Gxc4 {-15.80/7 8} B@f4 {+15.64/6 14} 50. P@e3 {-16.40/7 6}
G@a2 {+14.70/6 9} 51. Rxh8+ {-9.44/8 7} P@d3 {+10.59/6 7} 52. Gxd3
{-6.44/6 1.9} N@c2 {+8.32/6 7} 53. Kxc2 {-0.48/6 0.8} B@b1 {+4.34/6 6} 54.
Kb3 {+1.04/7 3} b4 {+5.45/5 5} 55. Gxb4 {-1.36/8 7} Be6 {+5.39/6 9} 56.
Gdc4 {+0.24/8 32} e4xe3=+ {+5.24/5 5} 57. N@c1 {+0.32/7 4} Gxb2 {+8.58/6 6}
58. Kxb2 {-3.20/7 9} Be4+ {+8.28/6 8} 59. R@e7 {-1.04/6 7} Bxc4 {+6.49/6 8}
60. Gxc4 {-0.16/6 4} B@c6 {+4.82/5 5} 61. B@f7 {+399.84/5 1.2} Kf5
{-159.94/6 2.5} 62. Be6+ {+399.88/5 0.2} Kg6 {-159.96/4 0.3} 63. Rg7+
{+399.92/3} Kh5 {-159.98/2} 64. +Rxg4=# {+399.96/1}
{White mates} 1-0

[Evert]

The following EuroShogi game is remarkable because Sjaak seems to grind Shokidoki completely and methodically to pieces (+16...), and then in a few moves gives it all away and eventually loses. I don't know if this points to some error, or that this is just typical of Shogi.

Sounds like my impression of Shogi, but I'll have a look. It'd be nice if it turns out to be a bug I can fix.

Down-scaling the piece values I use in Shogi by a factor of 4 seems to improve play considerably, so there may be something to gain there as well (I already had them relatively low compared to other variants so other evaluation terms are relatively more important, but it seems smaller values are even better).

While I was going through this, I realised that I never assigned piece values to pieces in Tori Shogi. Since those pieces are sufficiently weird that I have no clue what they should be worth, I decided to hack in a bit of numerology that tries to guestimate the piece values based on how many squares they attack and what pieces promote to, tuned to get the orthochess rook at 500 and the pawn at 100 evaluation units. It's almost certainly bullshit, but it's probably better than nothing, and it only activates if you leave the piece values blank.

The need to specify piece-values is really annoying, but I don't really see a way around it. I can't exactly go and measure what they should be when a variant is selected and I don't want to rely entirely on dodgy ad-hoc formulas.

[myfish]

I see you've added euro to shokidoki.

Is that just a test windows binary so far?

[hgm]

I see you've added euro to shokidoki.

Is that just a test windows binary so far?

Actually I was testing on Linux. But it is still in the test phase, and also capable of horrendous blunders:

Code: Select all

[Event "Computer Chess Game"]
[Site "hgm-xboard"]
[Date "2015.02.19"]
[Round "8"]
[White "Shokidoki tori3"]
[Black "Sjaak II 1.0 RC4"]
[Result "1-0"]
[TimeControl "40/300"]
[Variant "euroshogi"]
[VariantMen "P:fW;N:sWffN;B:B;R:R;G:WfF;+P:WfF;+N:WfF;+B:BW;+R:RF;K:K;+p:WfF;+n:WfF;+b:BW;+r:RF"]
[FEN "1nbgkgn1/1r4b1/pppppppp/8/8/PPPPPPPP/1B4R1/1NGKGBN1[-] w 0 1"]
[SetUp "1"]

{--------------
. n b g k g n .
. r . . . . b .
p p p p p p p p
. . . . . . . .
. . . . . . . .
P P P P P P P P
. B . . . . R .
. N G K G B N .
white to play
--------------}
1. Kd2 {+0.84/14} f5 {-0.08/10 7} 2. Gc2 {+1.20/13 6} d5 {+0.27/9 5} 3. Nc1
{+1.84/12 9} c5 {+0.29/8 5} 4. g4 {+1.76/11 5} Bf6 {+0.34/9 7} 5. Be2
{+1.96/12 10} Gf7 {+0.26/8 5} 6. Nf1 {+2.52/12 3} e5 {+0.20/9 5} 7. f4
{+2.88/12 7} fxf4 {+0.13/9 5} 8. Rf2 {+2.56/11 3} Re7 {+0.13/8 4} 9. Rxf4
{+2.92/11 4} P@f5 {+0.16/10 8} 10. Rf2 {+3.04/12 11} Ge6 {+0.09/9 7} 11.
Nd1 {+3.08/12 11} g5 {+0.22/10 5} 12. gxg5 {+4.64/11 3} Bxg5 {+0.19/9 7}
13. Rg2 {+4.16/11 6} Rg7 {+0.42/8 5} 14. Bf3 {+3.60/10 5} f4 {+0.56/9 8}
15. Be2 {+3.12/11 3} Gd6 {+0.66/9 12} 16. P@g4 {+3.24/11 15} Bf6
{+0.50/9 5} 17. Rf2 {+3.04/11 18} Bxg4 {+0.22/9 7} 18. Bxg4 {+3.24/11 3}
Rxg4 {+0.88/9 8} 19. P@g3 {+2.96/10 7} Rg7 {+0.60/9 8} 20. Rxf4
{+2.84/10 6} B@g2 {+0.20/7 8} 21. Rg4 {+3.40/9 8} Rxg4 {+1.42/7 6} 22. gxg4
{+3.60/8 2.0} R@h2 {+2.06/6 10} 23. P@f2 {+4.20/8 2.7} d4 {-0.44/5 5} 24.
g5 {+6.52/8 4} Nf8 {+1.92/5 8} 25. g5g6=+ {+7.64/9 4} dxd3 {+2.64/7 9} 26.
Gxd3 {+8.92/10 2.9} Nxg6 {-0.70/7 5} 27. R@g8 {+9.96/9 4} Kf7 {-0.24/7 5}
28. B@h7 {+10.32/10 5} P@g7 {-0.60/8 5} 29. Rh8+ {+10.72/9 4} Rxh3+
{-0.62/7 7} 30. +Rg8= {+18.60/10 4} Ke6 {-0.47/7 7} 31. Bxg6+
{+19.80/10 11} gxg6 {+7.45/7 5} 32. +Rxg6= {-3.00/7 3} P@g7 {+8.91/7 11}
33. N@f4 {+2.88/8 3} Ke7 {+11.34/8 11} 34. Ng3 {+6.80/10 7} gxg6
{+10.62/8 8} 35. Nxh3 {-10.84/8 1.2} R@a2 {+10.12/7 11} 36. Kc1 {-9.60/9 6}
B@a1 {+13.10/7 7} 37. Bxa1 {-10.20/10 10} Rxa1+ {+12.13/6 8} 38. R@b1
{-9.40/10 8} B@b2 {+13.05/7 9} 39. Kc2 {-8.68/10 8} +Rxb1 {+12.91/7 14} 40.
Kxb1 {-7.28/10 6} Bxa3+ {+10.29/7 13} 41. B@c1 {-9.72/10 47} +Bxc1
{+11.46/6 4} 42. Nxc1 {-7.80/10 14} B@h7 {+9.70/6 8} 43. B@f5 {-3.84/7 4}
R@h1 {+8.18/6 8} 44. R@e6 {+7.84/8 4} Kf7 {-6.45/5 7} 45. Rxd6+
{+399.80/9 2.1} Kg7 {-9.10/5 6} 46. +Rxf6 {+399.84/7 0.2} Kxf6
{-159.94/6 2.3} 47. Be6+ {+399.88/5 0.1} Kg7 {-159.96/2 0.1} 48. G@f7
{+399.92/3} Kh8 {-159.98/2} 49. B@g7# {+399.96/1}
{White mates} 1-0

On 31. Bxg6+ it counts itself +19 ahead, because it does apparenly overlook 31... gxg6, and after Sjaak replies with the latter its score drops to -3. When I let Shokidoki think for black in that position it turns out g7g6 is not even in its move list!

[Edit] Ah, it thought that deferral by Pawns on 6th rank was forbidden, because the promotion-flags board was still initialized like 6th rank was the last rank. (I cloned Euro from Judkins)... Of course promotion is also forbidden, as it is not in the black zone, so black Pawns could not move to 6th rank at all.

[Evert]

It's almost certainly bullshit, but it's probably better than nothing,

I guess 10 out of 10 against the previous version (with no piece values at all) is sufficient indication of being superior.

The games are ridiculously funny too: the old version just sits there while the new version gobbles up all available pieces before moving in to mate the lonely king.

[hgm]

Did you really update the source tar ball for RC4? I am trying Sjaak II under Linux again, but the binary I build from the RC4 tar ball gives RC3b as its version number. It also give an error on the following game (I received from Keith, who apparently does have a running RC4): it refuses 62. Bxf4= (now this happens to be illegal in Euro-Shogi, but Sjaak should not know that).

Code: Select all

[Event "Computer Chess Game"]
[Site "kinshoo"]
[Date "2015.02.21"]
[Round "-"]
[White "Sjaak II 1.0 RC4"]
[Black "Shokidoki euro"]
[Result "0-1"]
[TimeControl "40/300"]
[Variant "euroshogi"]
[VariantMen "P:fW;N:sWffN;B:B;R:R;G:WfF;+P:WfF;+N:WfF;+B:BW;+R:RF;K:K;+p:WfF;+n:WfF;+b:BW;+r:RF"]
[FEN "1nbgkgn1/1r4b1/pppppppp/8/8/PPPPPPPP/1B4R1/1NGKGBN1[-] w 0 1"]
[SetUp "1"]

{--------------
. n b g k g n .
. r . . . . b .
p p p p p p p p
. . . . . . . .
. . . . . . . .
P P P P P P P P
. B . . . . R .
. N G K G B N .
white to play
--------------}
1. g4 {+0.05/12} Gf7 {-0.16/14 5} 2. c4 {+0.08/12 4} Ke7 {+0.44/13 7} 3. f4
{+0.48/11 4} Nf8 {+0.84/13 8} 4. Bc3 {+0.47/12 8} c5 {+1.00/11 4} 5. cxc5
{+1.54/12 8} Rc7 {-400.00/12 0.1} 6. Rc2 {+1.54/11 6} Ne8 {+1.00/10 5} 7.
Bg2 {+1.91/10 8} Bd7 {-400.00/12 0.1} 8. Ge2 {+2.61/9 4} Rxc5 {+0.20/11 7}
9. Bb7+ {+3.38/11 11} Gc8 {+0.24/10 6} 10. +Bxb6 {+3.56/11 6} Rc6
{-0.96/11 0.1} 11. +Bd4 {+3.66/10 6} Rc7 {-0.16/10 4} 12. P@c5 {+3.66/10 5}
Bb5 {+0.20/10 5} 13. Bb2 {+4.31/10 8} Gd8 {-0.36/10 7} 14. +Be4
{+4.73/11 9} P@c4 {-0.68/10 7} 15. b4 {+5.88/11 4} Bd7 {-2.60/12 10} 16.
Rxc4 {+5.33/11 7} a5 {-400.00/12 0.1} 17. Nf1 {+5.99/10 8} a4 {-2.76/10 4}
18. Gc2 {+6.35/9 5} e5 {-2.72/10 7} 19. Bxe5 {+5.98/10 8} axa3+
{-3.80/10 11} 20. Nxa3 {+7.06/10 8} Rc8 {-4.40/10 0.1} 21. b5 {+7.06/9 8}
Be6 {-4.52/10 0.4} 22. Ra4 {+8.56/9 5} Bxg4 {-5.12/9 1.6} 23. b6+
{+9.14/8 7} Rxc5 {-5.80/9 0.3} 24. Ra7+ {+9.21/8 8} Gd7 {-5.28/9 8} 25. Bc3
{+10.92/10 27} Bxh3+ {-5.80/8 4} 26. +Rxb8 {+10.45/10 7} +Bxf1 {-4.64/8 7}
27. N@e5 {+11.11/9 15} Rxe5 {-400.00/9} 28. Bxe5 {+12.95/8 4} P@c3
{-8.56/7 7} 29. Bxc3 {+14.01/8 6} N@f2 {-10.52/7 6} 30. Gxf2 {+15.04/9 7}
+Bxf2 {-16.64/8 7} 31. +Pc6 {+16.03/8 4} Gd8 {-14.04/6 6} 32. N@e2
{+15.82/7 7} +Bf1 {-13.24/7 12} 33. R@b7 {+19.90/7 7} Kf8 {-11.36/7 1.0}
34. +Pd7 {+21.61/7 5} Gxd7 {-13.80/8 5} 35. Rxd7+ {+20.98/9 32} G@e7
{-400.00/10 0.1} 36. +Rdc7 {+21.75/7 2.1} P@c4 {-13.36/7 9} 37. Bd4
{+23.06/7 2.7} P@a4 {-11.68/7 6} 38. Nb3 {+23.63/7 3} P@e5 {-13.52/7 6} 39.
+Ba8 {+23.69/7 2.5} a3+ {-13.44/7 7} 40. Nxa3 {+23.87/8 4} c3+
{-15.16/8 19} 41. +Rxc3 {+25.32/9 9} P@a4 {-16.16/8 13} 42. Nb3
{+26.08/9 6} e4 {-17.88/9 35} 43. +Bxe4 {+26.50/9 8} d5 {-19.48/9 13} 44.
+Bxd5 {+27.63/8 4} Gfe6 {-21.96/8 21} 45. +Be4 {+27.62/8 5} a3+
{-21.00/8 10} 46. Nxa3 {+28.66/8 6} Kf7 {-22.48/9 24} 47. P@g5 {+29.22/8 5}
gxg5 {-22.84/9 19} 48. G@g6 {+29.98/9 8} Kf8 {-22.40/9 13} 49. Bb6+
{+30.38/9 8} g4 {-21.96/8 31} 50. Nd4 {+30.51/9 6} G6d6 {-30.20/9 45} 51.
P@g5 {+31.06/9 8} f5 {-25.20/8 22} 52. Gxg7 {+31.41/8 8} Kxg7 {+0.00/1 0.2}
53. +Bxf5 {+31.31/8 9} N@f8 {-24.52/7 13} 54. Nc6+ {+33.07/8 12} Gde6
{-26.08/8 51} 55. +Rb7 {+32.85/8 11} P@d7 {-25.48/8 14} 56. +Bxg4
{+32.98/8 8} Nf6 {-25.60/9 10} 57. +Be2 {+31.06/8 6} +Bxe2 {-23.92/8 6} 58.
Kxe2 {+30.36/9 8} B@h5 {-23.36/8 7} 59. Kf2 {+30.19/7 7} Ng4 {-23.16/7 1.1}
60. Kg3 {+31.47/7 6} Ge5 {-22.60/8 4} 61. B@c7 {+31.30/8 8} Gxf4
{-400.00/7 5} 62. Bxf4= {+31.78/6 4} Nxf4 {-20.80/6 0.9} 63. B@d4
{+31.03/6 6} P@f6 {-19.56/5 0.9} 64. G@g6 {+28.53/6 7} Kg8 {+399.72/7 0.4}
65. Gh7 {+21.40/5 5} Kxh7 {+399.80/9 0.1} 66. g6+ {-159.90/6 5} Nxg6
{+399.84/7 0.4} 67. +Rd2 {-159.94/7 3} B@g1 {+399.88/5 0.1} 68. P@g4
{-159.96/4 0.2} G@h4 {+399.92/3} 69. Kf3 {-159.98/2} Gxg4# {+399.96/1 0.1}
{Black mates} 0-1

[Evert]

Did you really update the source tar ball for RC4?

Yes, I just double-checked: the download link points to the RC4 tarball, and the version inside the tarball is tagged RC4.
Is it possible you accidentally didn't overwrite the old version when you unpacked the tarball?

I am trying Sjaak II under Linux again, but the binary I build from the RC4 tar ball gives RC3b as its version number.

Remind me again: what was the main problem with the Linux binaries on the download page? Is it the lack of an SSE2 version (I can make one easily), or is it a problem with runtime libraries (in which case I guess I should static-link)?

It also give an error on the following game (I received from Keith, who apparently does have a running RC4): it refuses 62. Bxf4= (now this happens to be illegal in Euro-Shogi, but Sjaak should not know that).

I get a different error here, but something does indeed appear to be up with it. I'll try to figure out what the problem is.
Just to clarify: why is Bxf4= illegal and why should Sjaak be unaware of this?
Is the definition for Euro-Shogi still the same as was posted earlier?

[hgm]

Is it possible you accidentally didn't overwrite the old version when you unpacked the tarball?

OK, my mistake. I always do unpack in a fresh directory (I have SjaakIIa ... SjaakIIf now). But I forgot the ./ in XBoard's -fcp argument, and apparently I had installed RC3b (which I normally don't do, but I guess there was a problem with make-install that I wanted to test, and I had forgotten all about it). So this was a false alarm.

Remind me again: what was the main problem with the Linux binaries on the download page? Is it the lack of an SSE2 version (I can make one easily), or is it a problem with runtime libraries (in which case I guess I should static-link)?

Code: Select all

hgm@hgm-xboard:~/SjaakIIf/Build$ ./sjaakii-linux64-sse42
./sjaakii-linux64-sse42: /lib/libc.so.6: version `GLIBC_2.15' not found (required by ./sjaakii-linux64-sse42)
./sjaakii-linux64-sse42: /lib/libc.so.6: version `GLIBC_2.14' not found (required by ./sjaakii-linux64-sse42)

I get a different error here, but something does indeed appear to be up with it. I'll try to figure out what the problem is.
Just to clarify: why is Bxf4= illegal and why should Sjaak be unaware of this?
Is the definition for Euro-Shogi still the same as was posted earlier?

Euro-Shogi has mandatory promotion, so deferral is by defenition illegal. If I am correct then there is still no way in RC4 to specify promotion is mandatory.

Note there originally was a problem with XBoard on loading this file. (Apparently PGN and engine moves use a different high-level move parser, and 'illegal drops' (N on last 2 ranks) were converted like the first two characters were a from-square when loading a PGN.) I pushed a fix for that to hgm.nubati.net.

[Evert]

OK, my mistake. I always do unpack in a fresh directory (I have SjaakIIa ... SjaakIIf now). But I forgot the ./ in XBoard's -fcp argument, and apparently I had installed RC3b (which I normally don't do, but I guess there was a problem with make-install that I wanted to test, and I had forgotten all about it). So this was a false alarm.

Ok, good to know!

Code: Select all

hgm@hgm-xboard:~/SjaakIIf/Build$ ./sjaakii-linux64-sse42
./sjaakii-linux64-sse42: /lib/libc.so.6: version `GLIBC_2.15' not found (required by ./sjaakii-linux64-sse42)
./sjaakii-linux64-sse42: /lib/libc.so.6: version `GLIBC_2.14' not found (required by ./sjaakii-linux64-sse42)

Ok, I'll try to static-link glibc. I think that's possible (but not recommended).

I get a different error here, but something does indeed appear to be up with it. I'll try to figure out what the problem is.
Just to clarify: why is Bxf4= illegal and why should Sjaak be unaware of this?
Is the definition for Euro-Shogi still the same as was posted earlier?

Euro-Shogi has mandatory promotion, so deferral is by defenition illegal. If I am correct then there is still no way in RC4 to specify promotion is mandatory.

Indeed not. In the development version you can set the optional promotion zone explicitly with

Code: Select all

Optional promotion: empty, empty

I'll double-check that this works as intended before pushing out the next release.

[/quote]
Note there originally was a problem with XBoard on loading this file. [/quote]
Yes, I just came to the conclusion that this is the issue I'm running in to.
Either way there is a problem in Sjaak though: if I simply feed it the SAN moves directly from the PGN, it chokes on some of them, but if I then convert them to LAN they are accepted. Not sure what's up with that, but I'll find out.

(Apparently PGN and engine moves use a different high-level move parser, and 'illegal drops' (N on last 2 ranks) were converted like the first two characters were a from-square when loading a PGN.) I pushed a fix for that to hgm.nubati.net.

Ok, I'll grab that.