Comments by HGMuller

I did not fully discuss the problem because the solution given for B23 was essentially correct. I only mention it because it failed to point out the essential need for deferring promotion to Lion, and I did not think the discussion in the MSM did that fact justice.

At that time the MSM was my only source for these problems, and it is a copyrighted source (which also contains the FENs). Many of these problems are presented on the website of teh chushogi-renmei, though. The numbers of the B series seem to correspond with the numbers in parentheses of set 1 there.

I took a look at my file with FENs, counting lines. Can this be the one for B23:

4+A1gk2f1/6s2+P1l/1+H1n4r1e1/5N3mp1/5xpO4/11+L/9Q2/12/3+v8/12/12/12 w - 0 1

?

[Edit] Indeed it is. HaChu cannot solve it, though, as it does not consider Kirin deferral. You have to first play the Kirin move; then it finds mated in 8 within a second.

Perhaps I should enhance the 2-vs-1 applet with the possibility to do divergent pieces, like the 3-vs-1 applet already supports.

A trick is to use the 3-vs-1 applet with one of the pieces useless for checkmating. Like mW or mR. I tried that with FAmW as the other piece, and stalemate as a win. This indeed is generally won.

Try to refresh the browser cache. I suspect you still have an old version of betza.js there, which still requires the deflection angle to be fully specified with directional modifiers (fs).

What is a Hexmaster?

If you use parentheses around a repeated group in XBetza, without indicating with a following digit how many repeats you maximally want, it uses the maximum board dimension as a wild guess. For the Hexmaster the repeated group describes 3 steps rather than one. So even on a board of size 10 repeating it 3 times already spans the full board. When there are additional steps outside of the parentheses, you probably only need two repeats.

So you could just insert a 2 behind the closing parenthesis, to suppress all the longer repeats that would never fit the board anyway.

Of course they don't happen at the same time; the game ends the moment the bishop is captured. A pawn doesn't promote until its owner says what it promotes to.

The rules in the article already resolve the dilemma according to that interpretation; the Rules section says:

If a player promotes his last Pawn, he loses (as his Pawns are now extinct), unless he wins by extinction on that very move.

I don't know if that is something that was added later. It is the opposit from what we have in Atomic Chess, where it is forbidden to blow up your own royal, even if that takes out the opponent's one. In Tenjiku Shogi it is also possible to destroy both royals in the same move, by capturing a King that stands next to a Fire Demon with your own King. The historic rules don't mention what would happen in that case (as it of course never happens in practice).

Is that possible?

I am afraid not. The applet serves two purposes: act as an 'electronic chess board', where a single player (or a pair sitting before the same computer display) can see and manipulate the entire board, or to play against an AI. But the AI handles two-player zero-sum games with complete information. For games where (a different) part of the game state is hidden for each player, like in your variant, you would need a completely different type of AI. Conventional chess engines are completely useless for that.