The Shortrange Project

The evolution of Western chess has been toward more powerful, longrange pieces. Shortrange pieces were perceived as weak and not very effective, so were abandoned in favor of longrange pieces. Comparing shatranj and FIDE shows this clearly. The shatranj rooks are the only 2 longrange shatranj pieces compared to the 5 longrange pieces, rooks, bishops and queen, in FIDE. Western chess variants have generally emphasized this trend, giving us more, and more powerful, longrange pieces. These pieces often combine a longrange and a shortrange piece together, making bishop-guard and rook-knight kinds of pieces. But often the intention seems mostly to allow the longrange component easier maneuvering to line up its cross-board attack. And the number of shortrange leapers in modern Western general use has dwindled down to the knight. The quest for "faster, longer, stronger" has left an entire area of piece evolution all but ignored. The ShortRange Project is an attempt to counter this trend by publishing games that feature shortrange pieces. We will begin by looking systematically at shortrange pieces and their utility, development, and classification.

This project started as a joint effort of Christine Bagley-Jones and Joe Joyce.
Sam Trenholme and Jeremy Good next contributed sections.
Currently, Jianying Ji is adding shogi.

Concepts:

It's reasonable to start with some agreements on concepts and working definitions. So, the SRP Guidelines:
1) All games should contain primarily or solely "shortrange pieces". More than half the pieces (not pawns) must be shortrange.
Shortrange pieces are those which may move less than half the length of the board in one turn, maximum. or:
Shortrange pieces are the shortest-ranged pieces of their types; eg: double-dabbaba moves 4 squares, but is shortest-ranged double-jumper. or:
Shortrange pieces are those which have, in absolute terms, a short range; ie: 1, 2…
2) A key component of this project should a method to analyze pieces in some simple and systematic way.
3) To the extent practical, standardized icons and/or icons that somehow represent the piece move will be used/created.
4) A "Shortrange Piecelopaedia" should be developed as part of the project. It should contain at least all the pieces used in any SRP variant.

What games fall inside or outside the guidelines?
Ms Bagley-Jones' recently published Capablanca Shatranj falls just outside the stated guidelines by having equal numbers of long- and short-ranged pieces (which is a pity as it gives an excellent demonstration of the utility of very shortrange pieces "in a hostile environment" - surrounded by longrange pieces). A variant of her game which replaces queen with sliding general will just barely qualify. Both of these games are nice tests of a longrange-shortrange piece mix.
All of Mr Joyce's shatranj variants easily qualify.

Games Released:

please see "Games" subsection

The Piece Builder:

Characteristics of Western pieces

Western-style pieces can generally be categorized by a few simple characteristics: direction, distance, and movement type.
Direction is orthogonal or diagonal.
Distance is the (small, whole) number of squares moved.
Movement is by either slide or leap.
That's it. Call them aspects of a piece. Simple combinations of these aspects provide building blocks, in the form of the simplest pieces, for a gigantic array of pieces that move 1, 2, 3 or 4 squares per turn, maximum. Combinations of these building blocks provide a vast array of pieces (far more than have ever been used), without being close to exhaustive.

Simplest and shortest-moving pieces

The simplest, shortest-range sliders move 1 square.
The Ferz moves 1 square diagonally. This piece is colorbound.
The Wazir moves 1 square orthogonally. This piece is not colorbound.

The simplest, shortest-range leapers move 2 squares.
The Alfil leaps 2 squares diagonally. This piece is colorbound.
The Dabbabah leaps 2 squares orthogonally. This piece is colorbound.

These are the 4 basic building blocks. Call them elements, for want of a better word.

Piece Builder: the Conjunction engine

Combinations

These elements may be combined in a number of very simple ways.
So a piece may move like:
"A OR B",
"A AND B",
"A AND/OR B",
"A AND/OR [B OR C]",
or any other conjunction or string of them you can fit into the concept "shortrange". But be careful, some conjunctions are far more powerful than others; put them in your engine and you can crash.

Simple Combinations: "Or"

The simplest combinations are pieces that combine 2 elements in an either-or manner.
The King moves as either the ferz or the wazir.
The modern Elephant moves as either the ferz or the alfil.
The modern Dabbabah moves as either the wazir or the ancient dabbbah.

Another simple combination uses 3 elements in an either-or.
The FAD moves as either the Ferz, or the Alfil, or the Dabbabah.

Or is the least powerful conjunction. After this, things get more complicated.
The next conjunction is "And". This puts a lot more zip into pieces.

Double Piece Combinations with "And"

An easy idea is to "double-up" on the simpler pieces.
The shortest "double" piece is a 2 square slider.
A ferz and another ferz is a 2-step, 2-square bishop. This seems simple enough.
The shortest double-leaper moves 4 squares.
A modern elephant and another is the Grand Shatranj "Oliphant", a linear 2-step piece that can move 1 to 4 squares per turn, as it can make either move of the elephant, then make either move again.

"And" Combinations: 2-step pieces and direction changes

Combining elements with "and" appears easy, but there are complications. First, let's start with FIDE pieces that seem simple.
The Bishop moves as ferz and ferz and ferz…
The Rook moves as wazir and wazir and…

The first question that becomes obvious is just how many wazir/ferz moves the rook/bishop can make, as it isn't specified. This can easily be dismissed as a quibble by making it "as many as wanted". (This glosses over some geometrical issues that may be addressed later.) A less obvious question is in just what direction(s) these repeated steps are being made. The FIDE rook moves in a straight line, a linear sequence of wazir moves. But a wazir making a number of moves may change directions during the sequence. Any piece whose move is a sequence of steps rather than a unitary whole has the potential to change directions at each step in the sequence.

Thus, 2-step pieces may need 2 directions specified.

"And/Or" Combinations

A similar result occurs with the "and/or" combination.
A Hero (Chieftain Chess) moves as a dabbabah and/or a wazir. It slides 1 or jumps 2 or slides 1 and jumps 2 or jumps 2 and slides 1, in any orthogonal direction. It is linear.
Its diagonal counterpart is the Shaman (Chieftain Chess), an alfil plus ferz. It, too, may move 1, 2 or 3 squares in a turn. Both pieces may attack up to 12 squares per turn, making them stronger than the knight.

Both the hero and the shaman are "2-step" pieces. So both can potentially change direction 1 time, between the steps.
The bent Hero moves like the linear Hero, and may also change directions in such a way that it imitates the knight's move, though it's more restricted than the knight in that it needs a clear landing spot for the first step so it may continue on to the second step. Of course, in other respects, it's much better than the knight. It attacks up to 20 squares and even in a corner attacks 8. It is certainly worth more than a rook. The bent Shaman is a little less valuable, but still easily the equal of the rook, if it isn't stronger.

Null Moves

The Sliding general is a combination of 2 guards. It may be considered a 2-step queen and only a linear mover (HyperModern Shatranj), or it may be a bent mover (Lemurian Shatranj, Chieftain Chess). As a bent mover, it can do something the bent hero and bent shaman cannot do. Unless specifically disallowed, it may move off and back onto its starting square, like any doubled piece may do. Doubled pieces require a null move rule.

Preferred Directions and Precedence in moves

The Knight Family

The standard kNight can be described as a 2-square leaper that combines the moves of wazir and ferz. The standard directions for the knight's move are: "move one square orthogonally, then 1 square diagonally outward, leaping any piece in the path". In other words, first move like a wazir, then like a restricted ferz, one with 2 preferred directions, "outward". Must a "preferred direction" aspect be added to the piece builder system to handle this? No, because the piece builder specifies a distance of 2 squares for the knight, and this forces the "preferred direction" without further specification. If the piece moves like a wazir, then a ferz, and goes 2 squares from its starting square, it must make a knight's move (on a 2D board, at least). This still leaves precedence in moves, doesn't it, to adequately describe the knight? Well, no, because if a piece leaps 2 squares, first moving as a ferz then as a wazir, it makes a knight's move also. So, for the standard knight, neither precedence of movement step nor determination of preferred directions is required. However, the standard knight is not the only member of the family.

The standard knight is a leaper. There are also lamed knights, which must move their 2 squares, but cannot leap. Where the standard knight is a 1-step, 2-square leaper, the lamed knights are 2-step pieces. Here precedence in moves makes a difference. There are 3 lame knights. One moves as a wazir first, then as ferz; and is known as the Mao. The second moves as a ferz and then as a wazir; it's known as the Moa. The third can move either way, and is known variously as Fergus Duniho's Squire and the British Chess Variant Society's Moo. (Some of the names we pick are pretty bad.)
These knights are all "And" knights; first they move as this, and then they move as that, with the first two having a preferred direction of movement which is expressed as the precedence of an element's move over the other element(s).

There are another 3 members of the knight family that are "And/Or" movers, tentatively called ponies, that move as wazir and/or ferz. They are similar to the lamed knights, except that a pony may stop after the first step of its move. A pony doesn't leap over a blocking piece on its first square of movement, but it may capture that piece. So, one pony moves as a wazir and then stops or continues on as a ferz. A second moves as a ferz and then stops or continues on as a wazir. These are the lamed knight-wazir and knight-ferz. The third pony moves as either of the first two, a lame equus rex. This third pony is the only one which doesn't require precedence.
NOTE: These pieces are rhinos, first used by Ralph Betza. See Piecelopedia at CVorg for details.

Utility of this system

The Piece Builder gives one a systematic way to look at shortrange pieces. A real utility of this system is that it allows the deconstruction of pieces into their smallest parts. This may aid in the creation of pieces by changing the reassembly, and make it relatively easy to include some specific qualities. For example, the previous section discusses the "Knight Family". Specifically, it discusses leaping and non-leaping (lame) knights and non-leaping ponies. Clearly, then, a "leaping pony" category can be created that produces not only the "equus rex", but also 2 other pieces, the knight-ferz and the knight-wazir. These latter two pieces are each worth about a rook. Admittedly, these pieces are not new, merely not common, but the piece builder also popped up the non-leaping ponies that preceded them, and those pieces are certainly much less common if not actually new, with a value probably between the knight and the rook.

The catch here is that the piece builder was fully created after all the design work for the pieces discussed herein. It provides a nice after-the-fact structure but implies that design is more structured than it actually is. However, concepts used in the piece builder were used to create and analyze pieces, for one example, the 3 pony pieces. Oddly, this explains why the pony is a default non-leaper. The only part of the system that then existed was the wazir-ferz part, and these pieces don't jump…

On piece icons

Ideally, the icons for a game would all be done in the same style, and would also represent the piece moves. In a series of related games by different designers that use some similar and identical pieces, and some different, consistency of both style and representations of movement across the series is extremely difficult to maintain. At a minimum, consistency of style within a game needs to be an important goal; but consistency of piece icons and moves across a range of games is of major importance for ease of play and preventing confusion. It's fine to have 2 icons for the same piece; it's a (serious) problem to have 2 pieces for the same icon.

Discussion: Making New Pieces

The complexity of chess pieces and variants can defeat any simple classification system. But, for a restricted area, a simple system can give very good results. Examples of pieces that move 1, 2, 3, or 4 squares per turn have been given, generated by simple combinations of basic pieces or by doubling or twinning a simple piece. (And the FIDE pieces were explicitly or implicitly generated as examples of how the system works.) A number of piece types have been looked at, some common, some much less so. We established three basic aspects: direction; distance; and mode of movement. We combined these aspects in simple ways: this OR that, this AND that; and built up pieces that became more and more complex. This brought the addition of three new secondary aspects: the number of possible steps in a move; direction change between steps; and precedence, or which component of a compound piece moves when. The basic aspects we used to make 4 "smallest possible" piece types. With these 4 base pieces as tools, we can analyze more complex pieces in a consistent way, and we can build complex pieces with specific characteristics from basic parts. At least, that's the claim. Therefore, to finish this explanation of the Piece Builder, we will attempt to actually build something new, and see what happens.

We can start by looking at what we've done so far. We've looked at and deconstructed pieces that move up to 4 squares, so let's do something different: grab some basic piece types (elements) and let them add up to 5 squares. The F+A+D (Ferz + Alfil + Dabbabah) is one combination that adds to 5 squares of movement, and it's colorbound. Now, if we haven't overreached enough, let's go a little further and use a more complex compound piece, the D+N+W (Dabbabah + kNight + Wazir) which is not colorbound. (Note all the various FAD compounds must be colorbound because all the components are colorbound. The D component of the DNW is colorbound, and the N and W are restricted to moving to the opposite color. So while the overall piece is not colorbound, the individual components are each restricted to landing on one color per turn.) How do we put them together?
If they are both going to be 3-step pieces, and both also are going to be double-leapers, then simplest way is just to add all three piece components together, as F and A and D, or D and N and W, with no precedence or any attempt at establishing a "preferred direction" by imposing a total distance moved on the resultant piece. The F+A+D winds up 1, 3, or 5 squares from its origin, with some moves doubling back on themselves, and is colorbound. The D+N+W winds up 0, 1, 2, 3, 4, or 5 squares away, again with some twisty moves, and is weakly colorbound. So this version of the pieces is somewhat chaotic and confusing, to say the least. We'll keep looking.
Okay, lets force them to move 5 squares away from their origin square. Now, both pieces are colorbound, with the FAD able to reach up to 16 squares and the DNW only 8, if they're on a large enough board. An 11x11 board (121 cells) provides only the 1 center square where these 2 pieces have enough room to reach their maximum number of squares, so unless we're contemplating a truly large game, on the order of 20x20 or so, these pieces are fairly useless.
To try to reduce chaos but give the pieces a decent attack/defense potential, let's combine the elements with "and/or" instead of merely "and": F and/or A and/or D; D and/or N and/or W. Now they do not have to move as each and every one of their components. But they are still far too strong. The FAD attacks 48 squares in a very odd pattern. This DNW piece attacks 76 squares, almost every square within 4 or 5 of its starting point. This is a bust, too, if you're looking for something that resembles tactics or strategy.
So let's try precedence. This reduces the range of the piece. We make F and [A and/or D]. This piece loses the 8 squares that an initial A or D move would get and it regularizes the move to something like the wizard move in Maxima. But by doubling back on its path, it may move 1 square; it moves 3 or 5 also, but not 2 or 4 squares. It attacks 32 squares in a nice square-with-the-corners-missing pattern, hitting 1/4 of the squares in its range and acting as a sort of giant alibabba. W and [N and/or D] is the corresponding other piece, and it loses no nearby squares of range because the W+D = N, more or less. It does need 2 free squares to move just 1 square total, though, so it's a little harder to go just 1. But it hits 54 squares in its range in an octagonal pattern. Still way too powerful. But there may be hope.
We need to fine tune the conjunction engine for these pieces. We also need to turn down the power. We'll keep precedence. For the FAD, the new piece moves as Either F only Or F And [A Or D]. This gives a piece that can hit 16 squares in a move, all at a range of 1 or 3, in a nice square pattern. This piece is colorbound, and can move to 1/4 of the squares within its range of 3, sort of a large alibaba. As it may move just 1 square diagonally, it is capable of hitting 1/2 of the squares on the board within a few moves. This is a nice piece, and would fit into a "large" variant nicely.
For the DNW, the piece moves as Either W only Or W And [D Or N]. This piece is still very strong, as it attacks the 36 squares closest to it in an octagonal pattern with side 3. It does need to have clear intermediate spaces to land on to get to all 36 squares, but this does not really restrict the power of the piece much. This piece is almost a match for the zigzag general of Atlantean Barroom Shatranj. Useful in large games, this piece is still more powerful than wanted for a pleasant piece for a relaxing game. (Like that's how we play!)
It's time to use all 3 of the aspects identified in the beginning of this essay to fully tame the DNW. We have not yet incorporated distance into this piece. Obviously, if it moves only as a wazir, its distance moved is 1 square. If we require that it must move 3 squares away when it moves as a wazir And [knight or dabbabah], then we have a piece that attacks 16 squares, 8 of each color. Four of the squares are adjacent to the piece, a dozen are 3 squares away, and the move looks rather like an extended Great/Grand Shatranj Minister's move. Finally, we have another nice, simple piece that would fit into a variant readily.

Had I known how difficult this was going to be when I started writing this section, I would probably have have picked much easier-looking pieces to analyze. But I didn't, so got to spend 2 weeks staring at walls, laying out pennies on checkerboards, cursing, and wishing I'd picked 1 much easier example to work out. Eventually, however, the system churned out 2 nice pieces for us, and 1 overpowered piece that may yet show up in some strange mythical shatranj variant. The system works; although we may find it easier to deal with less complex pieces, it did create pieces that are new - to me at least - and it will obviously do the same for other combinations. Victory is declared!
I don't know if you, at this point, are as exhausted reading this as I am writing it, but here is a good place to stop (at least for now). Any questions, comments, or criticisms are welcomed.


Add a New Comment
Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-Share Alike 2.5 License.