Theban Chess

[Laskos]

I propose possibly a viable game, which uses mostly the rules and pieces of standard chess, as much as a chess engine like StockFish 2.1.1 can play it. Basically I change only the starting position and the starting set of pieces, the rest being the same.

The starting postion is:

[D]1p6/2p3kn/3p2pp/4pppp/5ppp/8/PPPPPPPP/PPPPPPKN w[/D]

Spartans are white, in their standard formation, having the King (always on the right wing in Sparta), a Knight, and 14 pawns, a homogenous phalanx.

Thebans are black, in their formation of Pelopidas and Epaminondas against the Spartans. The King and the Knight are on the left flank, the 12 pawns are forming a deep column opposing the Spartan King, and the phalnax in echelon formation.

These two formations are pretty historically accurate, with Thebans in numerical inferiority (12 pawns against 14).

Pieces are all the same as in the standard chess.


Balance:

1) I checked for the balance of the game. I mean, I checked that for the StockFish 2.1.1 engine, one at normal settings, another at a very high Aggressivity 180 instead of 100, with many games.

a) The game is balanced. For 10s/move the result in 210 games is

Games : 210 (finished)

White Wins : 91 (43.3 %)
Black Wins : 75 (35.7 %)
Draws : 44 (21.0 %)
Unfinished : 0

White Perf. : 53.8 %
Black Perf. : 46.2 %


For 1 minute / move, pretty long games, the result is also very balanced

Games : 40 (finished)

White Wins : 16 (40.0 %)
Black Wins : 14 (35.0 %)
Draws : 10 (25.0 %)
Unfinished : 0

White Perf. : 52.5 %
Black Perf. : 47.5 %


b) The very Aggressive StockFish 2.1.1 (Aggressiveness 180 instead of 100) performs better:

Program Score % Av.Op. Elo + - Draws

1 S-Aggressive : 117.0/210 55.7 3180 3220 42 42 21.0 %
2 S-Normal : 93.0/210 44.3 3220 3180 42 42 21.0 %


and respectively at long time control

Program Score % Av.Op. Elo + - Draws

1 S-Aggressive : 23.0/ 40 57.5 3174 3226 97 96 25.0 %
2 S-Normal : 17.0/ 40 42.5 3226 3174 96 97 25.0 %


c) StockFish 2.1.1, having these results in real games (a very small advantage for white), is pretty clueless in the starting position. The analysis is the following:

[D]1p6/2p3kn/3p2pp/4pppp/5ppp/8/PPPPPPPP/PPPPPPKN w[/D]

FEN: 1p6/2p3kn/3p2pp/4pppp/5ppp/8/PPPPPPPP/PPPPPPKN w - - 0 1

Stockfish-211-64-ja:
1/1 00:00 79 0 -4.40 a2-a4 b8-b7 a1-a2 Nh7-f8 d2-d3 Nf8-e6 d1-d2 d6-d5 b2-b4 Kg7-f6 c2-c3 c7-c5 c1-c2 c5xb4 c3xb4
2/2 00:00 140 140,000 -4.84 a2-a4 Nh7-f6
3/3 00:00 252 126,000 -4.80 a2-a4 Nh7-f6 a1-a2 b8-b7 d2-d3 g4-g3 f2xg3 f4xg3 h2xg3 h4xg3 f1-f2 h5-h4 d1-d2 f5-f4 a4-a5 Kg7-f7 c2-c4
4/4 00:00 492 164,000 -4.80 a2-a4 Nh7-f6 a1-a2 b8-b7 d2-d3 g4-g3 f2xg3 f4xg3 h2xg3 h4xg3 f1-f2 h5-h4 d1-d2 f5-f4 a4-a5 Kg7-f7 c2-c4
5/5 00:00 827 206,750 -4.76 a2-a4 Nh7-f6 a1-a2 b8-b7 a4-a5 h4-h3
6/6 00:00 1,676 279,333 -4.32 a2-a4 Nh7-f6 a1-a2 b8-b7 d2-d3 g4-g3 f2xg3 f4xg3 h2xg3 h4xg3 Nh1xg3 h5-h4
7/10 00:00 2,551 318,875 -4.28 a2-a4 Nh7-f6 a1-a2 b8-b7 d2-d3 g4-g3 d1-d2 g3xh2+ Kg1xh2 g5-g4
8/12 00:00 7,654 510,266 -3.63 a2-a4 Kg7-f7 a1-a2 Nh7-f8 c2-c3 b8-b7 c1-c2 Nf8-e6
9/12 00:00 12,522 626,100 -3.63 a2-a4 Nh7-f8 a1-a2 Nf8-e6 b2-b3 b8-b7 b1-b2 Kg7-f6 a4-a5 h4-h3 a2-a4 h5-h4
10/13 00:00 19,308 715,111 -3.39 a2-a4 b8-b7 a1-a2 Nh7-f6 d2-d3 g4-g3 h2xg3 Kg7-f7 d1-d2 f4xg3 f2xg3 Kf7-e6 f1-f2 f5-f4 g3xh4 g5xh4 c2-c4
11/14 00:00 31,646 703,244 -3.19 a2-a4 Nh7-f8 a1-a2 Nf8-e6 d2-d3 Kg7-f7 d1-d2 b8-b7 b2-b4 c7-c5 b1-b2 c5xb4
12/17 00:00 146,733 858,087 -2.74 a2-a4 Nh7-f8 a1-a2 Nf8-d7 d2-d3 Nd7-b6 b2-b3 g4-g3 f2xg3 f4xg3 d1-d2 g3xh2+ Kg1xh2 b8-b7 b1-b2
13/21 00:00 303,584 882,511 -2.74 a2-a4 Nh7-f6 a1-a2 e5-e4 d2-d3 e4xd3 c2xd3 g4-g3 h2xg3 f4xg3 c1-c2 b8-b7 d1-d2
14/21 00:00 625,997 939,935 -2.30 a2-a4 Nh7-f6 a1-a2 b8-b7 d2-d3 Kg7-f8 d1-d2 d6-d5 b2-b4 g4-g3 f2xg3 f4xg3 h2xg3 h4xg3 b1-b2
15/21 00:00 790,182 942,937 -2.38 a2-a4 b8-b7 a1-a2 d6-d5 d2-d3 Kg7-f6 d1-d2 e5-e4 c2-c4 Kf6-e5 c4xd5 Nh7-f6 c1-c2 Nf6xd5 c2-c4 Nd5-f6 a4-a5 Ke5-d4 d3xe4 f5xe4
16/21 00:00 924,304 942,205 -2.38 a2-a4 b8-b7 a1-a2 d6-d5 d2-d3 Kg7-f6 d1-d2 e5-e4 c2-c4 Kf6-e5 c4xd5 Nh7-f6 c1-c2 Nf6xd5 c2-c4 Nd5-f6 a4-a5 Ke5-d4 d3xe4 f5xe4
17/23 00:02 2,699,285 924,096 -2.22 a2-a4 b8-b7 a1-a2 Nh7-f8 d2-d3 Nf8-e6 d1-d2 d6-d5 b2-b4 Kg7-f6 c2-c3 c7-c5 c1-c2
18/26 00:05 4,892,455 908,871 -2.18 a2-a4 b8-b7 a1-a2 e5-e4 c2-c4 Nh7-f8 c1-c2 Nf8-e6 d2-d3 e4xd3 d1-d2 d3xe2 f1xe2 Kg7-f6 b2-b3 h4-h3 g2xh3 g4xh3 b1-b2 Ne6-d4 c2-c3 Nd4xe2+ Kg1-f1
19/29 00:22 19,913,890 891,241 -2.02 a2-a4 b8-b7 a1-a2 Nh7-f8 d2-d3 Nf8-e6 d1-d2 e5-e4 c2-c3 Ne6-c5 b2-b3 e4xd3 c1-c2 d3xc2 b1xc2 Kg7-f6 a4-a5 Nc5-e6 d2-d4
20/29 01:11 61,826,832 865,631 -1.77 a2-a4 b8-b7 a1-a2 Kg7-f6 d2-d3 e5-e4 d1-d2 d6-d5 c2-c4 Kf6-e5 c1-c2 Nh7-f6 b2-b3 e4xd3 c2xd3 g4-g3 h2xg3 f4xg3 b1-b2 d5xc4 b3xc4
21/29 01:55 99,541,617 861,354 -1.65 a2-a4 b8-b7 a1-a2 g4-g3 f2xg3 f4xg3 h2xg3 Nh7-f6 d2-d3 h4xg3 f1-f2 h5-h4 d1-d2 f5-f4 c2-c4 h6-h5 c1-c2 Kg7-f7 a4-a5 Kf7-e6 Kg1-f1
22/30 02:52 148,444,060 860,046 -1.57 a2-a4 b8-b7 a1-a2 Nh7-f6 d2-d3 Nf6-d7 d1-d2 Kg7-f6 b2-b3 e5-e4 c2-c4 g4-g3 f2xg3 f4xg3 h2xg3 e4xd3 c1-c2 d3xe2 f1xe2
23/32 06:13 320,076,434 857,121 -1.49 a2-a4 b8-b7 a1-a2 e5-e4 c2-c4 Nh7-f6 c1-c2 d6-d5 c4xd5 Nf6xd5 d2-d3 Kg7-f6 d1-d2 Kf6-e5 d3xe4 f5xe4 f2-f3 e4xf3 e2xf3
24/33 09:01 463,213,658 855,507 -1.37 a2-a4 b8-b7 a1-a2 Nh7-f6 d2-d3 d6-d5 d1-d2 e5-e4 c2-c4 e4xd3 c4xd5 d3xe2 f1xe2 Nf6xd5 c1-c2 Kg7-f6 b2-b3 Kf6-e5 Kg1-f1 g4-g3 h2xg3

It shows a decreasing with the depth advantage for black. It gave me hope that (also by looking at actual games played by the engine), that it deals wrongly with the position.


Human (me) against StockFish 2.1.1:

1) Me playing as Black (Thebans) against StockFish 2.1.1 (Spartans)

[White "StockFish 2.1.1"]
[Black "Human"]
[Result "0-1"]
[SetUp "1"]
[FEN "1p6/2p3kn/3p2pp/4pppp/5ppp/8/PPPPPPPP/PPPPPPKN w - - 0 1"]

1. a4 g3 2. fxg3 fxg3 3. hxg3 hxg3 4. f2 h4 5. a2 f4 6. b3 Nf6 7. c4 Nd7 8. b2 Kf6 9. c2 Kf5 10. e3 Kg4 11. f3 Kf5 12. e4+ Ke6 13. e2 b7 14. c3 h5 15. d4 g4 16. fxg4 hxg4 17. d2 h3 18. Kf1 Kf6 19. b4 Kg5 20. Kg1 Nf6 21. d3 f3 22. gxf3 Kf4 23. dxe5 gxf3 24. exf3 dxe5 25. Kf1 g2+ 26. Kg1 Kxf3 27. Nf2 h2+ 28. Kxh2 Kxf2 29. Kh3 g1=Q 30. d4 Qg3# 0-1

My plan is to push my strong left wing to kill the Spartan King, plain and simple. If I am not going to kill the King, then at least to get his Knight, trying to obtain a draw, by redirecting my King and my Knight in to stop Spartan strong left wing.

[D]1p6/2pn4/3p2pp/4p1p1/P1P2pkp/1P2P1p1/PPPP1PP1/3PP1KN w[/D]

I am going for his (white's) right wing where the Spartan King is. White (StockFish) seems a bit "undecided" what to do.

[D]8/1ppn4/3p2p1/4p1k1/PPPPPpp1/2P3pp/PP1PP1P1/5K1N w[/D]

Pushing hard there, white left wing advances steadily, but, again, somehow cluelessly.

[D]8/1pp5/5np1/4p3/PPP1Pk2/2PP1P1p/PP4p1/5K1N w[/D]

I don't know what StockFish (white) thinks here, his strong left wing becomes useless.

[D]8/1pp5/5np1/4p3/PPPPP3/2P3qK/PP3k2/8 w[/D]

That's it. It pretty accurately reproduces the battle of Leuctra, where the audacious Theban in this formation killed the Spartan King Cleombrotus and won the battle.

2) Me playing as White (Spartans) against StockFish 2.1.1 (Thebans)

[White "Human"]
[Black "StockFish 2.1.1"]
[Result "1-0"]
[SetUp "1"]
[FEN "1p6/2p3kn/3p2pp/4pppp/5ppp/8/PPPPPPPP/PPPPPPKN w - - 0 1"]

1. a4 b7 2. b4 Nf6 3. c4 d5 4. c5 g3 5. hxg3 d4 6. b2 Nd5 7. b5 Kf6 8. c2 Ke6 9. a5 b6 10. cxb6 cxb6 11. a6 Nc7 12. a2 Nxb5 13. e3 Nc7 14. a7 Kd7 15. exd4 exd4 16. gxf4 gxf4 17. f3 Kc6 18. Nf2 Kb7 19. Nd3 g5 20. Ne5 Nb5 21. Nc6 d3 22. e2 Nxa7 23. Nxa7 Kxa7 24. exd3 Kb7 25. d4 b5 26. d3 b4 27. d2 Kc7 28. c4 Kc6 29. d5+ Kb6 30. d4 Kc7 31. c5 g4 32. a4 bxa3 33. bxa3 Kc8 34. f2 h3 35. gxh3 Kc7 36. a4 Kb7 37. d6 Kc6 38. a5 Kb7 39. d7 Kc7 40. c6 g3 41. a6 Kxc6 42. d8=Q Kb5 43. a7 Kc4 44. a8=Q Kb3 45. Qc6 gxf2+ 46. Kxf2 Ka2 47. Qb8 Ka3 48. Qa6# 1-0

The approach is a bit more of a brute force (Spartan way:)). My plan is to generate a left wing onslaught with my superior forces, trying to defend my right wing King.

[D]8/1pp3k1/6pp/1PPnpppp/P2p1p1p/6P1/1P1PPPP1/P1PPPPKN b[/D]

I am advancing my strong left wing as hard as I can do it safely. Glad that StockFish is clueless that it has to kill my King, he seems to be bothered by my advancing pawns.

[D]8/P1nk4/1p4pp/5ppp/3p1P1p/8/PPPP1PP1/3PPPKN b[/D]

StockFish (Thebans, Black) seems an utter idiot, he completely abandon the idea to kill my King, being very bothered by my advancing pawns. It becomes mostly a material battle by now, StockFish as a Theban is surely not Epaminondas.

[D]8/k7/1p5p/5ppp/5p1p/3P1P2/PPPP2P1/3P1PK1 b[/D]

LOL

StockFish got rid of my first wave of attack, but lost _all_ his attack possibilities. Now my Spartans can roll, a clear brute-force win by now with so many passers.

[D]8/2kP4/2P4p/P4p1p/3P1p2/5PpP/3P1P2/6K1 w[/D]

The stupidity of a 3200 Elo engine pays its dividents. The final is:

[D]1Q6/8/Q6p/5p1p/3P1p2/k4P1P/3P1K2/8 b[/D]



I do not know how well the game is balanced for strong humans. It's entirely possible that a strong human can see a clear win for one side. But the delight of beating a super-engine like StockFish (which generally deals well with pawn endgames) systematically as both Thebans and Spartans is something to me as much as human planning goes. Also, the opposite wing disbalance confuses very much a normal chess engine.

Kai

[hgm]

Pieces are all the same as in the standard chess.

The rules of standard Chess are unfortunately a bit ambiguous as to the allowed moves for 1st-rank Pawns. Are they allowed a double-push? (Or even a triple step?) If they are, can the opponent e.p. capture them? I would think it more natural to abandon the double-push completely. I mean, why should the Theban c7 have a double-push, but not b1 or d3, who are also in their initial position... How about your idea to allow only promotion to Knight?

Actually, I think that it would be more in the spirit of this game to only allow promotion to Ferz, or perhaps to 'omni-directional Pawn' (i.e. capturing diagonally in all directions, like a Ferz, but having only orthogonal non-captures, like a Wazir.

A Ferz or Omni attacking Pawn lines from behind would already be extremely devastating, and have about the same destructive effect as you would expect from a break-through of enemy lines. A Queen seems unnecessary strong. Even promoting to a (non-royal) King, to keep it in the realm of orthodox Chess pieces, seems a bit too strong.

Fairy-Max could easily be configured to do any of that:

Code: Select all

// Theban Chess (legality-testing off!)
Game: fairy/Theban # PNB..Kpnb..k
8x8
1 1 1 1 1 1 3 4
2 2 2 2 2 2 3 4
p:100 -16,6 -15,5 -17,5 
p:100 16,6 15,5 17,5
n:400 14,7 31,7 33,7 18,7 -14,7 -31,7 -33,7 -18,7
k:-1  1,7 -1,7 16,7 15,7 17,7 -16,7 -15,7 -17,7
e:400 15,3 17,3 -15,3 -17,3
R:700 16,3 -16,3 -1,3 1,3
b:200 1,6 16,6 -1,6 -16,6 15,5 17,5 -15,5 -17,5

This in the fmax.ini file would allow you to play Theban Chess as variant fairy, after you selected 'Theban' from the 'Variant fairy selects:' combobox in the Engine Settings dialog. Of course you would have to start from a setup position to specify the Theban phalanx, e.g. put the FEN in a file 'theban.fen', and specify -lpf theban.fen for WinBoard.

[fern]

Man, I see you are a guy with interest and knowledge of arcane historical issues.
Knowledge of Tactic of armies of antiquity is an specialty that very few have.

With my "loriga segmentata" in my chest regards
Fern

[Laskos]

Man, I see you are a guy with interest and knowledge of arcane historical issues.
Knowledge of Tactic of armies of antiquity is an specialty that very few have.

With my "loriga segmentata" in my chest regards
Fern

Lorica segmentata or laminata may well suit the Emperor Titus, but the common mail, lorica hamata made Marius great

Scipio Africanus at Zama regards,

Kai

[Laskos]

Pieces are all the same as in the standard chess.

The rules of standard Chess are unfortunately a bit ambiguous as to the allowed moves for 1st-rank Pawns. Are they allowed a double-push? (Or even a triple step?) If they are, can the opponent e.p. capture them? I would think it more natural to abandon the double-push completely. I mean, why should the Theban c7 have a double-push, but not b1 or d3, who are also in their initial position...

Right, I used Stockfishe's rules: all second rank pawns can double-push with e.p. rules. First rank pawns can only single-push to the second rank, becoming fully enabled second rank. If I modify the rules as not to allow the double-push at all for the second rank, I think White will be heavily disadvantaged, as all its pawns can potentially double-push, and only 2 black pawns. For now, I would stick with the rules for pawns movement I have, I can think of the second rank as an assembly area.

How about your idea to allow only promotion to Knight? Actually, I think that it would be more in the spirit of this game to only allow promotion to Ferz, or perhaps to 'omni-directional Pawn' (i.e. capturing diagonally in all directions, like a Ferz, but having only orthogonal non-captures, like a Wazir.

A Ferz or Omni attacking Pawn lines from behind would already be extremely devastating, and have about the same destructive effect as you would expect from a break-through of enemy lines. A Queen seems unnecessary strong. Even promoting to a (non-royal) King, to keep it in the realm of orthodox Chess pieces, seems a bit too strong.

Right again, Queen promotion seems too strong, but I have to check the balance when changing the rules, Black wins are different in character than White wins, even if the game is balanced. Anyway, a promotion of only to Knight, if possible easily to apply to available strong engines, would be very desirable. Ferz and Omni-Pawn would be even more realistic, but I don't know how many engines would allow this.

Fairy-Max could easily be configured to do any of that:

Code: Select all

// Theban Chess (legality-testing off!)
Game: fairy/Theban # PNB..Kpnb..k
8x8
1 1 1 1 1 1 3 4
2 2 2 2 2 2 3 4
p:100 -16,6 -15,5 -17,5 
p:100 16,6 15,5 17,5
n:400 14,7 31,7 33,7 18,7 -14,7 -31,7 -33,7 -18,7
k:-1  1,7 -1,7 16,7 15,7 17,7 -16,7 -15,7 -17,7
e:400 15,3 17,3 -15,3 -17,3
R:700 16,3 -16,3 -1,3 1,3
b:200 1,6 16,6 -1,6 -16,6 15,5 17,5 -15,5 -17,5

This in the fmax.ini file would allow you to play Theban Chess as variant fairy, after you selected 'Theban' from the 'Variant fairy selects:' combobox in the Engine Settings dialog. Of course you would have to start from a setup position to specify the Theban phalanx, e.g. put the FEN in a file 'theban.fen', and specify -lpf theban.fen for WinBoard.

Thanks, I will check it for balance with Fairy-Max, and try to beat it Will check also the variants with exotic promotions, as they are more realistic.

Kai

[fern]

Aha...an expert like me....so we can together form the "testudo" and attack or defend anything.

With my spata at hand regards
Fern

[hgm]

Will check also the variants with exotic promotions, as they are more realistic.

Note that the game definition above assumes promotion to Omni, which I represented by a Bishop. (To make sure it is the default piece WinBoard proposes when you promote, with the 'Almost always promote to Queen' option on.) WinBoard an Fairy-Max will allow under-promotion to Knight, though; Fairy-Max always allows promotion to any piece type that is defined in the variant, except Pawns, although it never will under-promote itself, and doesn't consider it for either side in its search. When legality testing in WinBoard is off, (which would be needed to allow the 'Bishop' to move as an
Omni), WinBoard will also allow promotion to (royal!) King, and Fairy-Max will not refuse that either. This is an undesired effect of Fairy-Max flaky implementation of under-promotion.

The omni-directional Pawn does have some funny properties, because it can be blocked like a normal Pawn. So if you promote by, say bxc1=B (I used B for Omni, for no good reason), and there are Pawns on b1, c2, d1 but d2 is empty, the Omni is trapped!

If we forget about the Omni completely, but use non-royal King or Knight as promotion pieces, it could be played with legality testing on.

The following game definition would only allow promotion to Knight, and Fairy-Max would only consider that in its search. With legality testing on, WB would not propose promotion to King:

Code: Select all

// Theban Chess (legality-testing off!)
Game: fairy/Theban_N # PN...Kpn...k
8x8
1 1 1 1 1 1 7 4
2 2 2 2 2 2 7 4
p:100 -16,6 -15,5 -17,5
p:100 16,6 15,5 17,5
q:375  1,7 -1,7 16,7 15,7 17,7 -16,7 -15,7 -17,7 
k:-1  1,7 -1,7 16,7 15,7 17,7 -16,7 -15,7 -17,7
e:400 15,3 17,3 -15,3 -17,3
R:700 16,3 -16,3 -1,3 1,3
n:400 14,7 31,7 33,7 18,7 -14,7 -31,7 -33,7 -18,7

To allow promotion to non-royal King, you would need the following definition. Fairy-Max would then only think about promotion to this, but WinBoard would allow you to under-promote to Knight, which Fairy-Max would accept. Actually this would be the default choice proposed by WB, with the 'almost-always queen' option on, so it would take some effort on the user not to use it.

Code: Select all

// Theban Chess (legality-testing off!)
Game: fairy/Theban_K # PN........QKpn........qk
8x8
1 1 1 1 1 1 3 4
2 2 2 2 2 2 3 4
p:100 -16,6 -15,5 -17,5
p:100 16,6 15,5 17,5
n:400 14,7 31,7 33,7 18,7 -14,7 -31,7 -33,7 -18,7
k:-1  1,7 -1,7 16,7 15,7 17,7 -16,7 -15,7 -17,7
e:400 15,3 17,3 -15,3 -17,3
R:700 16,3 -16,3 -1,3 1,3
q:375  1,7 -1,7 16,7 15,7 17,7 -16,7 -15,7 -17,7 

If you want the Pawns to have a double-push (you can define that for white and black Pawns separately, btw), you would have to change their lines to:

Code: Select all

p:100 -16,24 -16,6 -15,5 -17,5
p:100 16,24 16,6 15,5 17,5

[hgm]

Perhaps I should change the WB algorithm for picking the default promotion piece (when the 'Almost always promote to Queen' option is on). Theban Chess is the first case I encounter where it seems to matter.



The current algoritm is that it tries Queen, and if that does not participate, it scans backward through the piece ordering until it does encounter a participating piece. This means it tries R, B, N before wrapping around. (The piece ordering is P, N, B, R, Q, fairy pieces, K.) After having decided on the default piece, it then runs through the pieces in the forward direction to present the user alternatives. For Chess this would immediately wrap to N, which is indeed the most common alternative. For Capablanca the first alternatives it would propose would be A and C, which also makes sense.

I wonder if scanning backwards to find the default makes much sense, however. In variants without Queen there is more often than not a piece amongst the fairies that is stronger than Rook. And in Theban it causes the problem that a Knight would always be proposed first (Q, R and B not participating), even if you introduce a fairy piece especially to satisfy promotions (unless you use B, R or Q to represent that, which would then compromise the ability to play with legality checking).

So it might be better to always scan from Q forward to find the default promotionpiece. Ferz is the first piece after Q, which indeed is the designated promotion piece in many of the Queenless ancient variants. In Theban it would pick a fairy first, (e.g. non-royal King), unless really there was none defined at all, in which case it would settle on the Knight.

[Laskos]

Will check also the variants with exotic promotions, as they are more realistic.

Note that the game definition above assumes promotion to Omni, which I represented by a Bishop. (To make sure it is the default piece WinBoard proposes when you promote, with the 'Almost always promote to Queen' option on.) WinBoard an Fairy-Max will allow under-promotion to Knight, though; Fairy-Max always allows promotion to any piece type that is defined in the variant, except Pawns, although it never will under-promote itself, and doesn't consider it for either side in its search. When legality testing in WinBoard is off, (which would be needed to allow the 'Bishop' to move as an
Omni), WinBoard will also allow promotion to (royal!) King, and Fairy-Max will not refuse that either. This is an undesired effect of Fairy-Max flaky implementation of under-promotion.

The omni-directional Pawn does have some funny properties, because it can be blocked like a normal Pawn. So if you promote by, say bxc1=B (I used B for Omni, for no good reason), and there are Pawns on b1, c2, d1 but d2 is empty, the Omni is trapped!

If we forget about the Omni completely, but use non-royal King or Knight as promotion pieces, it could be played with legality testing on.

The following game definition would only allow promotion to Knight, and Fairy-Max would only consider that in its search. With legality testing on, WB would not propose promotion to King:

Code: Select all

// Theban Chess (legality-testing off!)
Game: fairy/Theban_N # PN...Kpn...k
8x8
1 1 1 1 1 1 7 4
2 2 2 2 2 2 7 4
p:100 -16,6 -15,5 -17,5
p:100 16,6 15,5 17,5
q:375  1,7 -1,7 16,7 15,7 17,7 -16,7 -15,7 -17,7 
k:-1  1,7 -1,7 16,7 15,7 17,7 -16,7 -15,7 -17,7
e:400 15,3 17,3 -15,3 -17,3
R:700 16,3 -16,3 -1,3 1,3
n:400 14,7 31,7 33,7 18,7 -14,7 -31,7 -33,7 -18,7

To allow promotion to non-royal King, you would need the following definition. Fairy-Max would then only think about promotion to this, but WinBoard would allow you to under-promote to Knight, which Fairy-Max would accept. Actually this would be the default choice proposed by WB, with the 'almost-always queen' option on, so it would take some effort on the user not to use it.

Code: Select all

// Theban Chess (legality-testing off!)
Game: fairy/Theban_K # PN........QKpn........qk
8x8
1 1 1 1 1 1 3 4
2 2 2 2 2 2 3 4
p:100 -16,6 -15,5 -17,5
p:100 16,6 15,5 17,5
n:400 14,7 31,7 33,7 18,7 -14,7 -31,7 -33,7 -18,7
k:-1  1,7 -1,7 16,7 15,7 17,7 -16,7 -15,7 -17,7
e:400 15,3 17,3 -15,3 -17,3
R:700 16,3 -16,3 -1,3 1,3
q:375  1,7 -1,7 16,7 15,7 17,7 -16,7 -15,7 -17,7 

If you want the Pawns to have a double-push (you can define that for white and black Pawns separately, btw), you would have to change their lines to:

Code: Select all

p:100 -16,24 -16,6 -15,5 -17,5
p:100 16,24 16,6 15,5 17,5

Thanks, yesterday night I tried your first setting, Omni-Pawn. It was funny, I am not used to these pieces and to these endgames. In short, Omni-Pawn seems too weak and too slow for this game. It cannot catch a passer even if it's low rank, because Omni is orthogonal only. With these dis-balanced wings, usually, if I am not killing the King outright, the other side finally promotes too, and this Omni cannot catch even a close King. The endgames are something like 5 Omni against 3 Omni or similar, 80% of the total games are draws, and I don't know how to deal with these pieces, I myself had difficulties getting more than a draw against Fairy-Max, breaking heavily first (for no use). It's pretty boring, and unrealistic too, breaking the phalanx, exposing many flanks and the rear should be pretty devastating.

It seems more promising today with the Knights, let's see, but again, it looks like the fight is in until the last pawn tries to promote, though the draw ratio is not as high.

On the second thought, if the promotion rules change, an engine like StockFish must modify its whole eval, a thing which I do not wish to happen. Do you change the whole eval of Fairy-Max (using the ini file) to take into account the promotion rules? I see that it's not too stupid dealing with these new pawns.

Probably the adequate piece would be a 3x3 Queen (instead of 8x8 Queen) and the Knight optionally, to have both historical accuracy and playability.

Kai

[Uri Blass]

When did stockfish learn to play non chess games?
I thought that it is a chess engine that knows only to play chess and not another game.

Even if it knows to play another game the 3200 rating points that it has is relevant only for chess so claiming:
"The stupidity of a 3200 Elo engine pays its dividents." is clearly not fair.