Thanks, Mr. Smith, for some comments that are actually germane to this game. (The comment section for the original version of Scirocco somehow turned into a discussion on the perfectibility of chess variants in general, not something I had in mind when I created a very specific game.)

I'm disinclined to remove the Dervish, since it's one of the pieces that gives this game its particular flavor. It's not a weak piece, considering the company it keeps; it's stronger than a majority of the unpromoted pieces in the game. I have given some thought to making it stronger, though, allowing to relay Firzan power as well as Alfil and Dabbaba, still with the constraint that the destination must be adjacent to the Dervish. I think I'll try this as an experiment, although it might make the Dervish too strong -- I don't want to turn it from *an* important piece into *the* important piece, since that role is intended for the Scirocchi.

The Harpy, on the other hand, I've long felt a little uneasy about. It plays reasonably well in Typhoon, where you are more likely to get one while you have plenty of weak pieces on the board that can benefit from a Knight relay, but in Scirocco, where there are fewer pieces to start with, that happens less often. I'm thinking of redefining it (in both games) in a way that is a compromise between its current power and your suggestion for the Dabbaba promotion: namely, that it moves as a Queen and relays Knight power along Queen lines. Again, I'll try this as an experiment, but reserve the right not to adopt the change if I don't like the results.

As for the starting array, I'm not terribly attached to the current one. My initial idea was to do what I did in Typhoon: that is, to allow you to swap any piece with the one on the same rank that is the same distance from the vertical midline of the board, provided that the overall result had either mirror symmetry or rotational symmetry. However, I decided that people would find it too troublesome to make this adjustment, which (mostly) has a pretty minor effect on gameplay. I settled on the current array, which is serviceable, pretty arbitrarily. Do you think anyone would actually go the trouble of choosing a piece arrangement at the start of each game, or (in Zillions, anyway) accept a randomly chosen arrangement?

The Stork, Goat, and Guard, as well as the current promotions of Alfil and Dabbaba, I'm pretty attached to. They are admittedly a bit quirky and asymmetrical, but after a decade I think I've had enough time to consider them fairly carefully, and they make for a pretty game in practice, if not in theory.

How about having two pieces in the Dervish's place and its mirror:

Ninja - Moves as Alibaba but captures as Commoner. Promotes to Chain, which moves without capturing as a Queen, but relays the move of any piece it 'defends' to another piece it 'defends'.

Lion - Moves as Commoner but captures as Alibaba. Promotes to Flip, which moves without capturing as a Queen, but makes the pieces a Queen's move away's non-capturing moves capturing, and their capturing moves non-capturing.

These pieces are in the spirit of the Dervish, I think.

But then you'd make certain pieces immobile?

The promotion of Goat to Lioness is a given, right? Since the Lioness is a
strict superset of the Goat.

Could you devise a couple more pawns? hehe... It's just one pet peeve of
mine, that you go to the number of pieces of each type and it's 1, 1,
..., 2, 2, ..., and suddenly, 8! hehe. Find it a bit inelegant, if you
know what I mean. Maybe 4 pawns on the center, 4 berolinas on the sides?

I give no ranking because I probably won't get the chance to play this,
though a few strategy guidelines would be nice. In any case, it sounds
pretty interesting!

I've conducted some experiments over the past few days with a stronger Dervish and Harpy.

The Dervish I proposed (relays Firzan power to orthogonally adjacent pieces, as well as Dabbaba plus Alfil, provided the target piece lands next to the Dervish) seems to work well. The Dervish, already important for assisting in the development of the Pawns and weak pieces, now performs this role even better. It's more like itself, I'd say.

On the other hand, The Mad Harpy, which relayed Knight moves along Queen lines instead of Knight lines, was a little too powerful. I was looking for a piece worth about a Queen, but even with the relatively low piece density usually prevailing at the point in the game when you can promote a Dervish to a Harpy, the Mad Harpy side won easily in most of the test games I tried.

I also tried shorter-range versions of the Mad Harpy. The one that moved and relayed as Q2 was a little too weak, but the Q3 version seemed about the right strength, so that's the one I'll adopt.

I haven't experimented with alternative starting arrays yet, but I expect I'll do that soon.

Regarding Mr. Smith's Ninja and Lion, they're fine pieces and I'm sure someone has used them elsewhere to good effect. However, I don't want to add new pieces to a board that is already full enough (an initial piece density of 54%), and I don't want to displace the Dervish, which has a particular role to play in this game. Scirocco is full of slow-moving pieces, and video-game-addled 21st-century types are unlikely to have the patience to play through an opening of this game unless there's some way to expedite it. By improving the mobility of the weakest pieces, the Dervish helps those pieces reach the front lines before (I hope) the snooze factor sets in.

Mr. Smith's Chain sounds like an extremely strong piece, but one that lacks clarity. The Harpy is already a bit of a challenge to keep track of: when you have one, any of your pieces may become able to move as a Knight, which scrambles any ideas you may have formed about your position before getting the Harpy. Still, it's not so bad: the new Harpy's range is limited, and we're used to dealing with Knights. However, a piece that potentially lets any of your pieces move like any other would be very confusing, at least for a poor player like me.

Despite the large size of the board and the novel pieces, Scirocco really is meant to be a rather clear, playable game, not an exercise in complexity and confusion. (Admittedly, my idea of what constitutes playable may be affected by the years I've spent playing computer strategy games, where a game with 1000 pieces on the board that takes ten hours is considered a miniature.)

I don't think Mr. Smith's Flip would be a very exciting addition to this game, because only a few pieces have different capturing and non-capturing moves, so it wouldn't have much effect.

Mr. Solé's suggestion about the Pawns is one I adopted, but in Typhoon rather than this game. The asymmetry between the number of Pawns each player has and the number of other pieces has been around since Shatranj, and I didn't want to make such a sharp break with game aesthetics that have been established for a millennium and a half.

I've revised this game with the new Dervish and Harpy I described in previous posts, and put a link to the new game on the ZRF page (but I haven't figured out how to replace the main ZRF link, if I can do that).

Let me know if you have a chance to try it out.

> why don't you consider revising Typhoon and Jupiter as well as Scirocco?

Consider it done. Because it is done.

The revisions haven't been reviewed yet, but I trust our editors will put them in the index soon. If you absolutely can't wait, check out Typhoon (revised) and Jupiter (revised).

There is still no ZRF for Jupiter, because Jupiter is just too big. The ZRFs for Typhoon and Scirocco (and my attempt to make one for Jupiter) all use a preprocessor I wrote for Zillions to try to make these big games more manageable. If you're interested, the first version of the preprocessor is posted at the Zillions site (I'll send them the current version soon, along with the input files from which the Scirocco/Typhoon/Jupiter ZRFs were generated).

Move-power relaying

The Harpy relays Knight power to every friendly piece in its Q3 vision, and the Dervish relays limited Alfil / Dabbaba / Firzan power to any adjacent friend. It was a challenge to allow the diagram to do this, as how would one describe in Betza notation that, say, N moves are conditional, and what exactly that condition is? The solution I finally arrived at is to consider these relayed N moves as (multi-leg) moves of the Harpy, rather than simple moves of the piece that makes them. (To try that, drag a Dervish (the flag) to the promotion zone, and answer the promotion question above the board with 'Yes' to acquire a Harpy to move around. You can of course also try out the relaying power of the Dervish itself.)

So to make such a relayed move, one first clicks the Harpy (the 'activator' piece), and this highlights the simple moves of the Harpy (in green, as it is a non-capture-only piece), but it highlights all activated pieces in cyan, indicating that going there would just be the first leg of the move, and another leg may follow. Selecting a thus activated piece then shows the relayed N moves of that piece, and clicking one of those then moves the piece, rather than the Harpy.

That the first leg of a multi-leg move acts on friendly pieces to relay power, rather than on enemy pieces to capture them in passing, is indicated by an 'x' modifier for the first-leg modality in the Betza notation understood by the diagram. (This was one of the few characters not yet in use.) So the moves of the Knights in Knight-relay Chess would be written as mNxaN, where the xaN describes the relay of N moves to pieces an N jump away. For the Harpy it was a bit more complex, because the activation step (Q3) was different from the relayed move.

Locust capture

The diagram already did implement locust capture, also as a multi-leg move, where the possible locust victims are highlighted in cyan first. After selecting one, the final destination(s) get highlighted, so one can be picked to land on. In Scirocco the Zig and Zag pieces have Checker-like locust captures, where they have to land on an empty square. This move is indicated a cafmW or cafmF, the first leg (before the 'a') obligatory capturing, and the second leg in the same direction ('f') obligatory landing on an empty square.

Printable versions, a new general diagram feature

About the game

Based on studying the rules, the following characteristics struck me as imperfections, and are the reason that I did not rate it as 'excellent':

• There is an annoying asymmetry in the promotions of Alfil and Dabbaba. If Alfil promotes to Bishop then surely the Dabbaba should promote to Rook!? A Genie seems excessively strong: rifle capture is devastating. It would not be out of line as promoted Knight, however. The other 8-way leapers (Commoner and Camel) also promote to pieces much stronger than a Rook.
• Making the moves that could break severe color binding capture-only seems a bad idea, and makes the Stork a very awkward piece.
• That one can win by forcing repeats through perpetual checking is very unsatisfactory. Even Chu Shogi has always made an exception for that.
• The repetition rule prevents a losing side from halting progress of the future winner by checking, because sooner or later he would run out of new checks. But when the defender has a stong piece like Queen, this can basically take forever. Chu Shogi also suffers from that. A rule against roaming the board giving (ultimately) pointless checks is sorely needed.
• The Dervish seems to start in a wrong location for its move. To really help development I would like to position it in front of the Pawn rank, to give the Pawns behind it forward A and D steps. But due to its high-order color binding the Dervish can never get there. To enhance the Dervish, it would have been better to add some extra non-capture moves to it that would break the color binding (like mW), then to make it activate more moves in its neighbors (as was done in this revision).

In the original version, I rather liked the idea of pawns with multiple-step promotions....however, in a game this big and complicated, it is probably best to simplify to one promotion per piece.

I like the idea of many weak pieces, which makes every game different.  Very nice!

@John Davis: BTW, what is your opinion about the chess variant 'Scirocco'? This also seems heavily Chu-Shogi inspired, except that rather than doing away with all the piece promotions, like you did, it seems to put even more emphasis on those. The piece designs used in the article here are of course awful, hardly any better than writing the name on them in kanji. (I guess this would be enough to discourage anyone from trying it...) But I made an implementation of it in Jocly ( http://hgm.nubati.net/jocly ) using chess-style pieces, and it was possible to shape most of the pieces such that it would be quite easy to recognize / remember how they move, for someone with a bit of knowledge on chess variants (i.e. who knows about Ferz, Wazir, Alfil, Dabbaba, Camel and Zebra).

I think the Dervish is not working properly on the interactive diagram, ata least on mobile.