Spartan chess, promotion to king

[hgm]

Thanks for all the replies. Changing the topic a little bit:
I was looking at some endgame scenarios for Spartan chess, and these all seem to be winnable for black (in general). I suspect they are forced wins.
kc-K: king and 1 captain vs king (interesting since the captain is just a minor piece)
kll-K: king and 2 lieutenants vs king
kk-K: 2 kings vs 1 king

Indeed. Also:

KR-kl: draw
KR-kc: draw
KR-kk: draw
KQ-kk: draw (!)

Of course all in the case where there is no quick tactical win of one of the pieces, which in the KQ-kk case requires the Spartan Kings to be connected.

Whether a piece with mating potential can ever be considered a minor is debatable. But the tactical (middle-game) value of the Spartan C is definitely not larger than the Persian minors B and N. The rule that in a Pawnless ending being a minor ahead is not enough for a win definitely do not hold when you are a C ahead:

klc-KB: won
klc-KN: won
kcc-KB: won
kcc-KN: won
kcc-KR: draw
kkc-KR: won
kkc-KQ: ?

Other interesting end-games:

kw-KB: won
kw-KN: won
kg-KB: won
kg-KN: won
kll-KN: mostly 50-move draw

kw-KBN: won or fortress draw
kw-KR: draw
kg-KR: draw
KQ-kg: draw

KQN-kw: won
KQB-kw: won

Most of this can be calculated with the WB compatible EGT generator I announced in the other thread, except that you would have to re-compile with DIAGSYM undefined for doing end-games with Spartan L. To prove the general result I usually avoid that by using the pessimistic (no lateral non-capture) or optimistic (1-step non-capture in 4 directions) octosymmetric replacements of the piece, and when it already works with these, it must work with the Lieutenant. kll-KN is very interesting in this cases, as it seems so close to the boundary that you cannot afford such replacements without losing the result you wanted to prove.

I saw that WinBoard 4.5.3 adjudicated kll vs K as a draw (insufficient material), so I had to turn adjudications off to test that one.

As Spartan Chess now is a fully supported variant, this should count as a bug. I guess the adjudication rules are based on the assumption that the Elephant represents a Xiangqi Elephant, and thus is not counted at all as potential mating material. I will fix that.

[Evert]

kc-K: king and 1 captain vs king (interesting since the captain is just a minor piece)

It is, but at close quarters it is equivalent to a rook (or better, because it can jump).
Sjaak uses a heuristic value based on the attack pattern on a 3x3 and a 5x5 board to estimate the danger of a piece in an attack on the king (in Jazz I just use the value of the piece, but since that is arbitrary input in Sjaak I decided not to do that). That means that when it participates in an attack against the king, the captain is weighted as strongly as a rook would be. This didn't seem all that unreasonable.

kll-K: king and 2 lieutenants vs king
kk-K: 2 kings vs 1 king

The latter is actually fairly easy to win as well. I never tried the first though, I wonder if Sjaak can do it...
EDIT: yes, it can. Easily.

[Ferdy]

With white to move in this position, does Qg5 result in checkmate, or can black then play e2d1=K? I thought this would be checkmate, but WinBoard did not consider it checkmate.

8/7h/4NQ2/P6k/2h5/8/4h2P/1K6 w - - 0 46
[d]8/7p/4NQ2/P6k/2p5/8/4p2P/1K6 w - - 0 46

There was a similar position posted here , but in that case it was black to move, so it definitely made sense that black could promote to king to avoid mate.

Received a reply from Steven Streetman, the inventor confirming that Spartan is allowed to promote to another king.