Comments by HGMuller

It went more quickly with Alfaerie, because it's easier to spot differences in color than differences in orientation. However, while doing it with Alfaerie, I was not taking any time to get any sense of the position, yet I would certainly have to do that if I were playing Shogi. Also, Shogi does not progress from one random position to another. The position changes incrementally, starting from a position where I already know where everything is without even looking at it. So, I would be able to use my knowledge of previous positions to understand what the slightly new position is.

Yeah, sure. Who needs the reminder of pieces? You might as well play blindfold... Well, for GM players that is actually true.  When they think far ahead they intentionally don't look at the board, because it just distracts them with the current position. But for mortals like us playing blindfold is a huge handicap. The sole purpose of having a board and pieces is to constantly remind us of what stands where. And the test shows that Alfaerie des a better job at that.

Recognizing the piece type is of course at least as important as recognizing the color. (Some servers offer an interesting 'semi-blindfold' mode, where you can see the side pieces are on, but not their type, like playing with blank tiles or checkers.) This was never contested. Coloring the pieces does not have any effect on the ability to recognize the type symbol, though. (I hope we agree on that, rather than having to prove it with a 'Spot the Queen' applet...) So it isn't really relevant that you also have to recognize the type; there is no need to degrade the side recognition for improving the type recognition. (And it will probably also not be contested by anyone here that for westerners kanji are doing a poorer job at that than pictograms, as far as type recognitiion is concerned.)

From whatching Chu-Shogi games by experience Chu-Shogi players, I noticed that these frequently blunder in an elementary way (such as overlooking a discovered attack on a high-value piece), and that I, as a non-player, could almost always predict when they were going to blunder. Because I was viewing the game with mnemonic pieces on my bot, while they were using the standard kanji pieces of the client. So it was obvious to me they suffered a iszable handicap by not seeing the board and pieces as I did it. It will of course make a great deal of difference how much time you have available. In correspondence play, where you think perhaps an hour per move, chances are good that after some time it will start to dawn on you that you misidentified a particular piece, and when you do you have every opportunity to think again. In real-time games you often will move before realizing your mistake.

As to the auhenticity... Seems to me you pretty much threw that out of the window when you abandoned the kanji. No Japanese would ever agree that there is any authenticity in these pieces. In fact I noticed they tend to even deny that you are playing Shogi, when you use non-kanji pieces.

Coloring the pieces of each side differently can limit the use of color in distinguishing between types of pieces.

Not really, because there are more than two colors. It is very possible to pick two different dark colors (black and blue), and two different light colors (white and yellow) where for each pair the "Spot the Intruder" game is just as easy, and then use blue and yellow for promoted pieces.

Most Shogi sets I have seen do not have red kanji for the promoted pieces, btw. They just write different kanji on them. (Well, people tell me they are really the same kanji, but in different fonts. But they look different enough.)

It's not for you to say what Japanese people would think, and it is not for Japanese people to judge whether one of my experiences feels more authentic than the other.

I am just relaying what Japanes people said, on the Shogi forums I visited, and at the OTB Shogi tournament in Kanazawa and Yokomama that my Shogi engine participated in. And you surely give a whole new twist to the concept 'authentic'.

I have not had any communication with them on this, but it would be as objectively false as saying that you are not playing Chess if you don't use Staunton pieces.

This has puzzled me too. We have no trouble recognizing a game with, say, Star Wars puppets as Chess. For Shogi that would be unthinkable. The Japanese seem to consider the physical representation just as much part of the game as the rules for how to move the pieces. That explains why the Japanese Shogi Association is so conservative in endorsing any kind of on-line play. They insist that the experience behind the computer screen must be as identical as can be technically expected to playing over the board with the prescribed equipment. Including the possibility to play illegal moves.

I suppose it must be that way, because the over-the-board equipment, being subject to physical limitations, is inferior to what you could do on a computer screen. (Funny story: when I was playing in Kanazawa, my program was of course showing the position as pictograms, even though the official board on which I had to perform the move used of course kanji tiles. And it puzzled many people, both opponents and audience, that I used different color pictograms for both sides. The said ro me: "How can this work? When you capture a piece, it has the wrong color." It did not occur to them that a computer can change colors at will.) So if they would allow people to use arbitrary representations, these would quickly move away from the traditional representation, and use one that doesn't hurt their rating so much. If they would have been confident that kanji tiles were the superior representation, they would not have little reason to forbid anything else. But as it is, they consider it cheating.

This is one of the most important reasons that a good game like Shogi has failed to conquer the world: people that try to popularize it usually have as their main agenda to spread Japanese culture. No to spread a good game.

This is not code; it is HTML text for pasting in the Pieces section of your article about the variant, if you want to present that as an auto-generated table (e.g. such as I used for Makromachy.)

For defining a custom piece set you need what the (old) PTA prints in the section "Custom sets (only needed with non-standard piece set):" after you press the GAME code button. Something like this:

{
"custom": {
  "dir": "/graphics.dir/alfaeriePNG35/",
  "pieces": {
    "P": "wpawn.png", "p": "bpawn.png",
    "N": "wknight.png", "n": "bknight.png",
    "B": "wbishop.png", "b": "bbishop.png",
    "R": "wrook.png", "r": "brook.png",
    "Q": "wqueen.png", "q": "bqueen.png",
    "K": "wking.png", "k": "bking.png"
  }
}
}

But what does this HTML sequence have to do with the Pieces section of my description of the variant?

The Piece Table button in the PTA is an aid for those writing an article about a new variant, to save them the work of writing a description of the pieces by hand. If they use the PTA for creating an Interactive Diagram they can copy-paste to the Setup section (after using the Show HTML button), they can use the Piece Table button to get the Pieces section as a free side effect, as a nice table showing the images of the pieces. The Pieces section in my Makromachy article was made that way.

The highlighting problem in the preset should now be solved too. The preset was using the "accelerated legality test" (by default), and this was testing only for attacks on the first royal type in the wroyals/broyals arrays (i.e. K or k in this case). I now changed that to the first royal type that is actually on board.

In my opinion, however, move diagrams provide a more rapid overview, both for the description of the variation and for the presentation of the rules in Game Courier.

This is apparently a matter of taste, because I hate articles with many move diagrams. Especially of large games. The diagrams dillute the information density by about a factor 20, and you have to scroll like mad to find the information you want. Awful. A simple sentence like "moves 1-4 squares orthogonally" tells me all I need to know, seeing a move diagram with a piece centered on an 8x8 board tells me hardly anything (as the leaps reach the board edge). And for complex pieces a static diagram often cannot really illustrate what the move does (e.g. Chu-Shogi Lion).

For those who want to see move diagrams the I.D. offers plenty of opportunity. You can summon a diagram for every piece without scrolling, either by clicking on the name of the piece (which I usually display to the right of the Diagrams), right-clicking the piece in the Diagram, or clicking in the pieceTable after you opened that.

If I see it correctly, the correct transfer of the piece table when creating a custom piece set in the preset is still missing. Or have I overlooked something. Anyway, I am grateful  for your doing.

I am not sure what you mean here. A 'transfer' is something you do, so how can it be 'missing'?

That is because you still have the HTML text that is supposed to go into an article's Pieces section in the Custom Sets text entry of the preset, rather than the description of a custom set.

Great! That is just what I need to avoid auxiliary intermediate variables and testing those with if-statements.

yafsR should be no problem (yafscabyaifzR). y(a)5K is not valid XBetza (as it includes yK), but I suppose you mean some kind of area move. If there is more than a left-right choice it is indeed not possible to retrace the path using q and z. This could also be solved by a new directional modifier that would do what i does for the range of a continuation leg: mimic a previous deflection.

It doesn't seem obvious to me, but that's probably just me.

Well, you unload something that disappeared elsewhere. And if p' means friendly hopping, it seems natural to make u' mean friendly unload, i.e. that you don't unload the captured piece, but the moving one.

(the "load" spot won't always be the end)

I don't get that. A moving piece would always have to follow the entire trajectory; that is the basis of XBetza  notation. So how could it end up anywhere else than at the end?

This i find a very good idea, 31 new symbols,

Well, I don't know. The success of Betza notation lies in its simplicity. The meaning of most modifiers is obvious (f, b, l, r as directions m, c as move/capture), and for those who are slightly familiar with unorthodox pieces, most atoms as well (standard symbols for orthodox chess pieces, plus W, F, A and D). Perhaps they have to learn a handful of other symbols (v and s, but outside Shogi pieces tend to be fully symmetric, and directional modifiers are only very rarely needed in the first place, p and g.)

Having to know the meaning of '31 new symbols' pretty much destroys that. There is very little benefit in introducing symbols that hardly anyone would know the meaning of.

The q and z modifiers cannot be used in their original Betza meaning of circular / crooked in a multi-leg move. Their meaning in the latter context has been redefined as a directional modifier meaning the same as l or r, but in the same or opposit direction of deflection as the previous one.

The circular and crooked moves were hard to fit into the scheme according to which other moves are handled by the Interactive Diagram. I added a special case of move generation just for those, where a tabulated sequence of legs of a multi-leg move doesn't have to be traversed in its entirety for the move to be valid, but is allowed to stop at any point along the trajectory. This can be combined with (grass)hopping; normally the trajectory would not continue after an occupied square, but would exercise its c or d right there. But in combination with p or g it must continue on the first such encounter, in the latter case exactly one leg. Continuing after capture, or on any other path than the indicated circular/crooked trajectory is not implemented for this type of move generation.

The introduction of parentheses in principle would allow specification of circular and crooked paths in the context of normal multi-leg move generation. But the way parentheses are implemented is very inefficient, which especially hurts the AI. In theory (af)W is the sane as R, but in practice it is expanded to WafWafafWafafafWafafafafW..., which is WnDnHnWXnDX..., a squence of ever larger lame leaps. And for each next leap it will test all squares leapt over, even though the previous lame leaps already tested those. So if the W and D leap in a certain direction are valid non-captures, but the H leap is blocked, it would still try to generate thw WX, DX... leaps, and test the W, D and H squares they pass through for each of those, before coming to the conclusion that these are not possible. The special case of storing only the longest trajectory, and allowing termination at any step along the way, solves this problem. But it is not compatible with straying from the path.

BTW, the I.D. currently does not implement moves with more than one pair of parentheses on an atom. The implementation through pre-processing would really make the length of the XBetza string that results explode if there would be multiple pairs.

I had no idea that the i could be that far separated from the long move it's supposed to imitate.

The official definition says "as many steps as in the previous slider leg". It would never make sense to mimic the length of a leaper leg, as this is known to be 1. So intervening leaper legs are simply ignored.

It would make sense to slightly alter that definition in the case of multiple iso legs, to "the previous slider leg that is not iso itself, and has not yet been mimicked by an intervening iso leg". This is not how the I.D. currently implements it though. Which probably does make the current implementation useless for multiple iso legs, as it does something I cannot imagine you would ever want.