WCTCC (World Computer Theban Chess Championship)

[Laskos]

Interesting; I'll try this out with SjaakII later. Not that it is any sort of competition for the likes of Stockfish...

I think, if you tune a bit some sort of eval for this game, it may be competitive. Normal chess engines probably play badly this game.

[Evert]

Interesting; I'll try this out with SjaakII later. Not that it is any sort of competition for the likes of Stockfish...

I think, if you tune a bit some sort of eval for this game, it may be competitive. Normal chess engines probably play badly this game.

I'm sure that's true, but I have it as a policy that I don't include variant-specific evaluation in the engine.

Of course I do have evaluation terms that are triggered by particular rules, but since this uses FIDE rules there is no real way to trigger those...

[ZirconiumX]

I tried this out with Dorpsgek, which finds the position to be legal. I don't know what it plans to do with the rank 1/8 pawns though.

[Laskos]

Interesting; I'll try this out with SjaakII later. Not that it is any sort of competition for the likes of Stockfish...

I think, if you tune a bit some sort of eval for this game, it may be competitive. Normal chess engines probably play badly this game.

I'm sure that's true, but I have it as a policy that I don't include variant-specific evaluation in the engine.

Of course I do have evaluation terms that are triggered by particular rules, but since this uses FIDE rules there is no real way to trigger those...

Still plays remarkably well this game. Although Sjaak is weaker than SOS 5.1 by 400 ELO points in normal chess, it beats by 200 ELO points SOS in this game.

Normal Chess:

Code: Select all

Score of SjaakII 1.3.1 vs SOS 5.1: 5 - 89 - 6  [0.080] 100
ELO difference: -424.28 +/- 127.27
Finished match

Theban Chess:

Code: Select all

Score of SjaakII 1.3.1 vs SOS 5.1: 71 - 22 - 7  [0.745] 100
ELO difference: 186.25 +/- 76.40
Finished match

One of the games won by Sjaak:

[pgn]
[Event "?"]
[Site "?"]
[Date "2016.07.06"]
[Round "11"]
[White "Sjaak"]
[Black "SOS"]
[Result "1-0"]
[FEN "1p6/2p3kn/3p2pp/4pppp/5ppp/8/PPPPPPPP/PPPPPPKN w - - 0 1"]
[PlyCount "85"]
[SetUp "1"]
[TimeControl "0.3/move"]

1. a4 {book} e4 {book} 2. f3 {book} d5 {book} 3. b4 {+0.30/10 0.35s}
exf3 {-M10/12 0.32s} 4. exf3 {+0.55/10 0.16s} gxf3 {-6.03/12 0.33s}
5. gxf3 {+0.80/11 0.19s} g4 {-3.41/11 0.31s} 6. fxg4 {+0.50/12 0.19s}
hxg4 {-0.29/12 0.33s} 7. f2 {+0.72/11 0.37s} g3 {-0.19/12 0.33s}
8. fxg3 {+1.51/12 0.19s} fxg3 {-0.09/13 0.30s} 9. hxg3 {+1.44/12 0.19s}
hxg3 {-0.12/13 0.34s} 10. Nxg3 {+1.52/13 0.16s} b7 {-0.15/12 0.33s}
11. b5 {+1.88/10 0.31s} b6 {+0.06/12 0.35s} 12. b2 {+1.92/11 0.16s}
f4 {-0.55/12 0.31s} 13. Ne2 {+3.48/13 0.38s} g5 {-1.04/11 0.31s}
14. b4 {+3.59/12 0.27s} Nf8 {-1.42/12 0.34s} 15. a5 {+3.96/12 0.19s}
bxa5 {-1.50/12 0.33s} 16. bxa5 {+4.09/13 0.29s} c6 {-2.88/11 0.33s}
17. bxc6 {+6.78/12 0.16s} Ne6 {-2.95/13 0.32s} 18. a6 {+7.80/12 0.20s}
Nc7 {-3.59/13 0.33s} 19. a7 {+8.31/11 0.15s} Kf6 {-3.63/13 0.35s}
20. Nc3 {+7.59/12 0.17s} Ke6 {-3.57/13 0.30s} 21. Nb5 {+8.09/13 0.20s}
Na8 {-3.79/13 0.33s} 22. d3 {+7.80/12 0.24s} h5 {-3.54/13 0.31s}
23. a2 {+7.98/12 0.36s} h4 {-3.19/12 0.35s} 24. c4 {+7.64/11 0.24s}
dxc4 {-3.59/13 0.34s} 25. dxc4 {+7.92/12 0.17s} h3 {-4.05/13 0.35s}
26. c7 {+10.26/13 0.16s} Kd7 {-3.81/13 0.32s} 27. c5 {+10.68/15 0.32s}
g4 {-5.08/13 0.33s} 28. c6+ {+11.12/14 0.33s} Kc8 {-7.71/12 0.34s}
29. Kh2 {+13.45/14 0.20s} f3 {-10.09/13 0.32s} 30. a4 {+13.46/15 0.37s}
Nb6 {-14.23/13 0.32s} 31. a5 {+15.25/13 0.20s} Na8 {-20.97/13 0.31s}
32. Kg3 {+17.22/13 0.18s} f2 {-24.52/13 0.33s} 33. exf2 {+17.28/11 0.14s}
Nxc7 {-26.12/12 0.33s} 34. Nxc7 {+18.79/11 0.26s} Kxc7 {-32.50/12 0.33s}
35. a8=Q {+24.22/11 0.29s} Kd6 {-34.19/11 0.32s} 36. Qb7 {+24.22/10 0.35s}
h2 {-35.92/10 0.31s} 37. Kxh2 {+25.60/9 0.23s} Ke6 {-M12/10 0.34s}
38. c7 {+M11/9 0.34s} Ke5 {-M8/12 0.31s} 39. c8=Q {+M9/60 0.091s}
g3+ {-M6/23 0.30s} 40. fxg3 {+M7/60 0.020s} Kf6 {-M6/23 0.32s}
41. Qf3+ {+M5/60 0.040s} Kg7 {-M4/24 0.32s} 42. Qff8+ {+M3/2 0s}
Kg6 {-M2/99 0.014s} 43. Qcf5# {+M1/2 0s, White mates} 1-0
[/pgn]

[Evert]

Still plays remarkably well this game. Although Sjaak is weaker than SOS 5.1 by 400 ELO points in normal chess, it beats by 200 ELO points SOS in this game.

Normal Chess:

Code: Select all

Score of SjaakII 1.3.1 vs SOS 5.1: 5 - 89 - 6  [0.080] 100
ELO difference: -424.28 +/- 127.27
Finished match

Theban Chess:

Code: Select all

Score of SjaakII 1.3.1 vs SOS 5.1: 71 - 22 - 7  [0.745] 100
ELO difference: 186.25 +/- 76.40
Finished match

That's incredible!
I guess it's a case of having the wrong type of knowledge to play a particular variant.
One of the things I don't like about SjaakII's play in this variant is that its king is too passive. I know why it does that (it's an artifact of the king safety), but it's something I should probably go fix anyway.

By the way - Jazz should be able to play this variant too (it's not a variant engine, but it seems to do the right thing if you give it the starting position); it's much better at it than SjaakII too, because it activates its king. It struggles to do so with White, but has a much easier time of it with Black.

Neat variant.

[Evert]

Ok, changing the king safety so SjaakII will actually activate its king makes a considerable difference. Not enough to bridge the gap with Jazz, but still.

(Disclaimer: I know, very small number of games, but I think the trend is quite clear from looking at the games anyway):

Code: Select all

Rank Name                  Elo    +    - games score oppo. draws 
   1 Jazz Orchestra 936M  2392  198  198    10   85%  2157   10% 
   2 Sjaak II 633         2157  161  161    20   58%  2072    5% 
   3 Sjaak II 1.3.1a      1751  275  275    10    0%  2157    0% 

[Laskos]

Ok, changing the king safety so SjaakII will actually activate its king makes a considerable difference. Not enough to bridge the gap with Jazz, but still.

(Disclaimer: I know, very small number of games, but I think the trend is quite clear from looking at the games anyway):

Code: Select all

Rank Name                  Elo    +    - games score oppo. draws 
   1 Jazz Orchestra 936M  2392  198  198    10   85%  2157   10% 
   2 Sjaak II 633         2157  161  161    20   58%  2072    5% 
   3 Sjaak II 1.3.1a      1751  275  275    10    0%  2157    0% 

It's a much improved Jazz? I downloaded Jazz 840, it loses to SjaakII 1.3.1. Could you upload the modified SjaakII and the new Jazz?