Arimaa Forum
        (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi)
      

        Arimaa >> General Discussion >> Short notation for Arimaa games and positions
        
(Message started by: omar on Dec 25th, 2008, 7:10am)
Title: Short notation for Arimaa games and positions
Post by omar on Dec 25th, 2008, 7:10am
For quite some time now I've been thinking on and off about an easier way to convey Arimaa moves and positions to other humans. The current notation is quite verbose and was specially designed to make it easy for computers to process. While writing comments in games or chatting during a game I felt the need to express the moves more concisely and quickly. Then while proof reading the Arimaa books Karl and Jean wrote it really felt cumbersome to read the notation. I am sure it was much worse for the authors to actually write it. I had a lot of ideas floating around about a new notation and last night I finally wrote them down. Here is the document defining the short notation.

http://arimaa.com/arimaa/learn/shortNotation.html

Hope I covered everything because this was a bit tricky to define. Feedback is welcomed.
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Dec 25th, 2008, 7:51am

Quote:
the 3rd gold rabbit moves west and then north. To find this rabbit scan the rows from left to right starting with the closest row until you encounter the 3rd gold rabbit.

This is a bit ambiguous.  Do you scan all of the first nearest row left-to-right before starting on the second row, or globally scan from left-to-right and break ties by which row is closest?
Title: Re: Short notation for Arimaa games and positions
Post by jdb on Dec 25th, 2008, 8:42am
PGN style piece disambiguation
http://www.drpribut.com/sports/standard.txt


Quote:
8.2.3.4: Disambiguation

In the case of ambiguities (multiple pieces of the same type moving to the same
square), the first appropriate disambiguating step of the three following steps
is taken:

First, if the moving pieces can be distinguished by their originating files,
the originating file letter of the moving piece is inserted immediately after
the moving piece letter.

Second (when the first step fails), if the moving pieces can be distinguished
by their originating ranks, the originating rank digit of the moving piece is
inserted immediately after the moving piece letter.

Third (when both the first and the second steps fail), the two character square
coordinate of the originating square of the moving piece is inserted
immediately after the moving piece letter.

Note that the above disambiguation is needed only to distinguish among moves of
the same piece type to the same square; it is not used to distinguish among
attacks of the same piece type to the same square.  An example of this would be
a position with two white knights, one on square c3 and one on square g1 and a
vacant square e2 with White to move.  Both knights attack square e2, and if
both could legally move there, then a file disambiguation is needed; the
(nonchecking) knight moves would be "Nce2" and "Nge2".  However, if the white
king were at square e1 and a black bishop were at square b4 with a vacant
square d2 (thus an absolute pin of the white knight at square c3), then only
one white knight (the one at square g1) could move to square e2: "Ne2".
Title: Re: Short notation for Arimaa games and positions
Post by jdb on Dec 25th, 2008, 8:47am

Quote:
Movements are specified using the letters to indicate the piece to move and one or more special characters to indicate direction of sequential steps. The direction characters are +, -, <, > for up, down, left and right. For example:
g 5g e+>>-
specifies that the board is viewed from the gold side and the gold elephant takes four steps moving north, east, east and then south. The side to move specifies which color pieces the letters represent.


Maybe consider using n,s,e,w for directions. They are already used in the long form notation, which would make it easier for people to switch between reading the two types of notation.
Title: Re: Short notation for Arimaa games and positions
Post by jdb on Dec 25th, 2008, 8:53am

Quote:
The side which the board is being viewed from can be specified before the move or by itself on a single line. The value may be: g, s or n to indicate gold, silver or neutral. In case of neutral the gold player is on the left and the silver player is on the right. If the side to view from is not specified at the start of the move then any previously specified view remains in effect. If no view has previously been specified then the gold side is assumed.


I recommend having only one view, probably gold. Having multiple views adds considerably to the state information required to read through a move list. It is a lot easier to play through a move list if the view is constant. It also makes the notation simpler.
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Dec 25th, 2008, 9:55am

on 12/25/08 at 08:42:53, jdb wrote:
PGN style piece disambiguation

The PGN piece disambiguation is seldom needed because the destination square is given, and it is unusual for two pieces of the same type to be able to move to the same square.  For Arimaa we have the luxury of being able to dispense with the destination square and reduce it to a single character, but that means rabbit moves will constantly need to be disambiguated, in contrast to pawn moves which are almost never ambiguous.  Since it is a different issue I'm not sure we should expect PGN-style to work well for us.

For starters, if we want to indicate that the rabbit on the d-file moves forward, PGN disambiguation would notate that rd+, which Omar has already given the meaning of "dog moves forward, pulling rabbit".
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Dec 25th, 2008, 10:03am

on 12/25/08 at 08:47:14, jdb wrote:
Maybe consider using n,s,e,w for directions. They are already used in the long form notation, which would make it easier for people to switch between reading the two types of notation.

Using cardinal directions was my first reaction as well, but on further consideration I find I am much quicker telling right from left than east from west, and even if I forget my right and left, the symbol itself shows the direction > or <.  A second issue with compass directions is that the letter 'e' would stand for east as well as elephant, making it harder to read a move like deeeenen unless we insert spaces, lengthening the notation, as compared to d>e>++, which is readable without spaces.

I take your point about switching between notations.  I expect to have difficulty using omar's suggested notation at first, but I think it is worth a try.  If it is way easier to use once we get used to it, we may even want to deprecate the long notation.
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Dec 25th, 2008, 10:07am

on 12/25/08 at 08:53:08, jdb wrote:
I recommend having only one view, probably gold. Having multiple views adds considerably to the state information required to read through a move list. It is a lot easier to play through a move list if the view is constant. It also makes the notation simpler.

I quite agree that we should stick to a single perspective.  Even if other perspectives are defined, it would be just too confusing to use them.  I'll wager that if we actually start using this shorthand in the chat room, anyone who uses silver-perspective shorthand will be pilloried.

There's something odd about making Silver reverse right/left and up/down from his perspective, but we have already been forcing that with the old cardinal directions, and gaining advantage of perspective just doesn't seem worth the confusion caused.
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Dec 25th, 2008, 10:13am

Quote:
g 33g d2>dc+c*d+d*
specifies that the 2nd gold dog goes east and pushes the silver cat north; the cat is removed; the dog continues north and is also removed.

As long as we are going for brevity, an asterisk can indicate that a capture was made without specifying the piece captured.  The above move would then become d2>dc+*d+*.

Actually, since we are relying on context to determine lots of other things, the context could also determine whether or not there are captures, making the notation d2>dc+d+ sufficient, but I can see why one might want the redundant information.  In chess notation the plus sign for check is redundant but generally useful enough to be justified.  So I expect my preference will be to indicate a capture with an asterisk, but leave off the letter of the captured piece.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 25th, 2008, 12:38pm

on 12/25/08 at 07:51:26, Fritzlein wrote:
This is a bit ambiguous.  Do you scan all of the first nearest row left-to-right before starting on the second row, or globally scan from left-to-right and break ties by which row is closest?


The first case. I changed the wording as follows now:


Quote:
To find this rabbit scan each row one at a time from left to right starting with the closest row and moving toward further rows until you encounter the 3rd gold rabbit.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 25th, 2008, 1:12pm

on 12/25/08 at 08:47:14, jdb wrote:
Maybe consider using n,s,e,w for directions. They are already used in the long form notation, which would make it easier for people to switch between reading the two types of notation.


The reason for going with symbols was because they are more intuitive. I think there is a bit more processing that humans have to do to map a letter to a direction than to map an arrow symbol to a direction. I was originally going to use ^ for up and v for down, but in some fonts they didn't look good, so I went with + and -.
Title: Re: Short notation for Arimaa games and positions
Post by jdb on Dec 25th, 2008, 1:32pm

on 12/25/08 at 13:12:26, omar wrote:
The reason for going with symbols was because they are more intuitive. I think there is a bit more processing that humans have to do to map a letter to a direction than to map an arrow symbol to a direction. I was originally going to use ^ for up and v for down, but in some fonts they didn't look good, so I went with + and -.


I disagree. The standard directions n,s,e,w are burned into people's brains at a young age.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 25th, 2008, 1:38pm

on 12/25/08 at 08:53:08, jdb wrote:
I recommend having only one view, probably gold. Having multiple views adds considerably to the state information required to read through a move list. It is a lot easier to play through a move list if the view is constant. It also makes the notation simpler.


One of the things I didn't like about the regular notation is that if forced a single view from golds side. Sometimes one would want to show the position diagram from silvers view, but then I have to keep reorienting myself to golds view to follow the notation. Allowing the author to select the view for the notation actually makes it easier to read and write it. The view can be selected so that < means my left, rather than my opponents left.

Giving the ability to change the view at the start of each move may be a bit of an overkill. In actual use I think an author will probably only set the view once and use it for all the moves. However the flexibility is there so I can do something like give golds setup from golds view, give silvers setup from silvers view and then switch to a side neutral view for the rest of the moves.
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Dec 25th, 2008, 1:44pm

on 12/25/08 at 13:12:26, omar wrote:
I was originally going to use ^ for up and v for down, but in some fonts they didn't look good, so I went with + and -.

Ooh, now that you mention it, I like caret and vee better than plus and minus.  I think you should at least list it as an optional replacement.  We should see what people like to use.

In general these theoretical arguments are just a lot of guessing, as you specifically found out.  You created one notation, Omar, and when I objected that it was cumbersome you defended its merits.  Once you actually started to try to use it in practice, you noticed its defects as well and came up with something more compact for a different purpose.

In fact, that's my attitude for giving moves from Silver's perspective as well.  My guess is that people will find the confusion from the same move having two (or three) different notations to be much more of a bother than having directions be inverted for Silver, but that's only a guess, and we will see in practice what people prefer.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 25th, 2008, 1:45pm

on 12/25/08 at 09:55:09, Fritzlein wrote:
For starters, if we want to indicate that the rabbit on the d-file moves forward, PGN disambiguation would notate that rd+, which Omar has already given the meaning of "dog moves forward, pulling rabbit".


Using the file will not always be enough to disambiguate since there could be multiple rabbits on the same file that can move forward. The sequential scan will always work and the number will never be more than 8.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 25th, 2008, 2:21pm

on 12/25/08 at 10:13:23, Fritzlein wrote:
As long as we are going for brevity, an asterisk can indicate that a capture was made without specifying the piece captured.  The above move would then become d2>dc+*d+*.


Good point. I actually like: d2>dc+c*d+*

That way the notation conveys information about which piece was removed in the case of dog pushs cat north (it could be possible that the dog is removed in some cases). Though this could be determined from the context adding the extra letter is useful in conveying some information that can be gleaned just from the notation.

In the case of dog going north and being removed, it works very well to not reference the dog again.

I've changed it now.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 25th, 2008, 2:44pm

on 12/25/08 at 13:44:04, Fritzlein wrote:
Ooh, now that you mention it, I like caret and vee better than plus and minus.  I think you should at least list it as an optional replacement.  We should see what people like to use.


OK I will list it as an optional replacement. Because in some fonts it does look good.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 25th, 2008, 3:33pm

on 12/25/08 at 13:44:04, Fritzlein wrote:
In general these theoretical arguments are just a lot of guessing, as you specifically found out.  You created one notation, Omar, and when I objected that it was cumbersome you defended its merits.  Once you actually started to try to use it in practice, you noticed its defects as well and came up with something more compact for a different purpose.


At the time I gave more importance to the ease of computers being able to parse the notation and being able to even step backwards through the notation. For example d2+* is enough to go forward but can't be used to go backwards. Also I didn't think that humans would be dealing with the notation that much since we would have computer tools to help us. But if there are physical books with notation then humans can't avoid dealing with it directly. The short notation leans way towards the humans without regard to how difficult it would be for computers to deal with. But who know over time we may end up wanting to use the short notation for computers too.

But you are absolutely right that experience is the best teacher of what works in the long run.
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Dec 25th, 2008, 9:42pm
Sure, the notation you came up with first served its purpose well.  And having notation from multiple perspectives serves another purpose.  I really appreciate how flexible you are about considering suggestions and changing your mind.
Title: Re: Short notation for Arimaa games and positions
Post by jdb on Dec 26th, 2008, 6:17am
For push and pull moves, there are two pieces that are moving together. Once one of the pieces is specified, there are only 4 possibilities for the location of the other piece. It's possible to specify the other piece by indicating the direction. (Using 4 new symbols)
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 26th, 2008, 7:51am

on 12/26/08 at 06:17:04, jdb wrote:
For push and pull moves, there are two pieces that are moving together. Once one of the pieces is specified, there are only 4 possibilities for the location of the other piece. It's possible to specify the other piece by indicating the direction. (Using 4 new symbols)


Since we already have four symbols for direction I would prefer to use those rather then introduce four more symbols with the same meanings. So e>^ would mean elephant pushes the piece on the east in the north direction and >e^ would mean the elephant pulls the piece on the east while moving north. But this would not work without introducing spaces so that the leading direction in a push/pull is not viewed as being applied to an earlier piece in the notation. However we could use two new symbols to indicate a push or a pull. Suppose ) stands for push and ( stands for pull. Then e)>^ would mean elephant pushes the piece on the east in the north direction and e(>^ would mean elephant pulls the piece on the east while going north. Or maybe it would be better to have the arrows align with the direction of movement of the two pieces and in the same order as would be done on the board. So e)^> means elephant pushes the piece on the east in the north direction and e(^< would mean elephant pulls the piece on the east while moving north. I think I like the second way better. The good thing about this is that you never have to reference the opponents pieces using letters. One thing you lose though is that if the same piece is being pulled twice or pushed twice it takes more letters to specify. It would e)>>)>> instead of ed2>>. But it is the same for a flip; it would be e(^<)<v instad of d2e^ed<.
Title: Re: Short notation for Arimaa games and positions
Post by jdb on Dec 26th, 2008, 8:03am

on 12/26/08 at 07:51:42, omar wrote:
Since we already have four symbols for direction I would prefer to use those rather then introduce four more symbols with the same meanings.


Maybe just a different mental perspective. The four arrow symbols refer to actually moving the piece(s). The four linking symbols (w,x,y,z or whatever) refer to linking the two pieces together. Then the following arrow symbol refers to both pieces moving together, as in a push/pull.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 26th, 2008, 8:09am
Thinking about this some more I think I like the notation of e)>^ meaning elephant pushes the piece on the east in the north direction and e(^< meaning elephant pulls the piece on the east while moving north. This way the directions symbols still indicate the direction of movement of the pieces and the direction of movement of the stronger piece is always comes first.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 26th, 2008, 8:11am

on 12/26/08 at 08:03:22, jdb wrote:
Maybe just a different mental perspective. The four arrow symbols refer to actually moving the piece(s). The four linking symbols (w,x,y,z or whatever) refer to linking the two pieces together. Then the following arrow symbol refers to both pieces moving together, as in a push/pull.


Can you give some examples.
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Dec 26th, 2008, 9:24am
Although it is one character longer to type e)^> instead of ed>, there is something attractive about not having to use the letters for opposing pieces.  For starters, if the elephant was positioned to push two different dogs, one would no longer have to disambiguate which dog.

On the other hand, ed> is closer to natural language, and probably closer to how I think.  The expression "elephant pushes dog right" is closer to my thought process than "elephant pushes up, displacing piece right.  Although that could be a problem with my thought process that a better notation will fix.  :)

This conversation is really making me eager to get into the chat room with other spectators of a tournament game so that  I can try to talk about the game and see what is easiest to type and most intelligible to read.  Definitely 2g e^^^h2^ is going to be a big improvement, but some of the finer points are non-obvious.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 26th, 2008, 10:20am

on 12/26/08 at 09:24:58, Fritzlein wrote:
On the other hand, ed> is closer to natural language, and probably closer to how I think.  The expression "elephant pushes dog right" is closer to my thought process than "elephant pushes up, displacing piece right.  Although that could be a problem with my thought process that a better notation will fix.  :)


I also think of push/pulls that way.

I was just talking to Aamir about this and he also prefers notation where the opponents piece is mentioned because just by looking at the notation one can tell what pieces got pushed/pulled.

Also I was thinking that in most cases one letter will be enough to identify the piece. So 3 characters will usually be enough to define a push or pull.

Maybe I will keep it as it is for now and after we use it in game comments and chat we will get a better feel for it.

While watching live games I tend to watch from the side of my favored player to win, but I think it might be better now to watch from a side neutral view to make chatting about the game easier with others.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Dec 26th, 2008, 11:08am
I changed the notation now to make ^ and v the characters for up and down and took out reference to + and -. Also mentioned that it is OK to use arrow like symbols if they are available.
Title: Re: Short notation for Arimaa games and positions
Post by clauchau on Jan 5th, 2009, 5:29am

on 12/26/08 at 08:09:02, omar wrote:
Thinking about this some more I think I like the notation of e)>^ meaning elephant pushes the piece on the east in the north direction and e(^< meaning elephant pulls the piece on the east while moving north. This way the directions symbols still indicate the direction of movement of the pieces and the direction of movement of the stronger piece is always comes first.

Then it might be easier to read and remember with the bracket put after the first direction : e>)^ and e^(<

That way it merely looks like specifying another piece and its moving direction.


Quote:
e)>>)>> ... e(^<)<v
would now be e>)>>)> and e^(<<)v
Title: Re: Short notation for Arimaa games and positions
Post by clauchau on Jan 5th, 2009, 5:45am

on 12/26/08 at 09:24:58, Fritzlein wrote:
it is one character longer to type e)^> instead of ed>

Unfortunately, () are also graphically too close to <> and often used to bracket some part of text of formula. It would be confusing if a move was specified along a text inside a paragraph, where () can also be used for remarks - at least in French, where they are used instead of dashes.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Jan 15th, 2009, 1:22pm
Updated the short notation page.

^, v, <, > are the perferred characters for directions.

+, - can be used in place of ^, v

(, ) can be used in place of <, >

Author can use capital letter for gold pieces and small letters for silver pieces, but not required.

Author can use spaces between movement of multiple pieces but not required.

A "piece square" form of specifying placement and movement was also added.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Jan 26th, 2009, 7:43am
Updated the short notation page to get rid of the piece scan and replace it with a square label such as b4. This is possible without confusion if the pieces are specified using capital letters. For example <tt>g 12g Dc2^</tt> mean the gold dog on square c2 moves north. Also in many cases it is possible to specify just the row or column to uniquely identify the piece. For example <tt>g 12g Dc^</tt> may be sufficient if there is only one gold dog on the c column.

Thanks to Karl and Jean for independently suggesting this via email after trying to use the short notation in their books.
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Jan 26th, 2009, 11:56am
I must expand on this latest change.  Omar has been pressing me to use the short notation in my book, which I was reluctant to do with an untested system where there might be kinks we haven't uncovered.  Indeed, as I began to try to use it, I discovered that the disambiguation scheme was a major hassle to use.  No matter how you define the board scan, it is a pain to figure out which is the sixth rabbit.  If you miscount, you are looking at the fifth or seventh rabbit and thinking something totally wrong about the position.

It makes sense to disambiguate by the column of the piece, as they do in chess on the rare cases where disambiguation is required, but for Arimaa the letters of the columns were colliding with the letters for the silver pieces.  To avoid the collision, I actually tried referring to the squares with two number coordinates, e.g. the elephant on square 52 steps north.  Unfortunately, the square numbers are easy to confuse with move numbers, and anyway this gives away the advantage of using square names that are familiar to chess players.

The idea of using capital letters for all pieces, both gold and silver, in order to free up lower-case letters for disambiguation, turns out to be the best solution by far.  Yes, it takes a second to figure out that in 25g RE<Re<^ the first R is a silver rabbit being pulled, while the second R is a gold rabbit moving voluntarily from the e-file.  I found, however, that I got used to the paradigm very quickly.

Once I adjusted to small letters always being disambiguation by column and capital letters always being pieces, I started to love the new notation.  It is not just shorter, it is easier to read and easier to write.  This is not speculation on my part: I have just finished converting my entire book (by hand) to the new notation.  If it wasn't clearly a superior notation, I would never have had the stamina to finish the conversion.

Omar, thank you very much for dragging me into this change.  Although I was right, and the new notation was actually flawed in an unexpected way, that was something we would never have found out except by trying to use it.  The change to capital letters for both colors of pieces has, I think, pushed the short notation over the hump where it can become the new standard.
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Jan 26th, 2009, 12:10pm

on 01/26/09 at 07:43:28, omar wrote:
g 12g Dc^ may be sufficient if there is only one gold dog on the c column.

I still believe that the flexibility of being able to specify a perspective is outweighed by the confusion caused.  For my book I have done everything from Gold's perspective, so I omit the initial 'g', and save another character.  If using other perspectives catches on, I will obviously conform to standard usage, but I think it is unlikely to catch on unless it is forced on people.  Just out of curiosity, Omar, do you intend to rename the squares too, so that a1 for Gold is h8 for Silver?  Or do you keep the square names the same and just reverse the arrows?  This wasn't an issue before, because we didn't use square names for disambiguation, but now it will matter.

A on a completely different point, I found myself tempted to write 12g D^ instead of 12g Dc^ if there was only one dog that could move north.  Is the further abbreviation is a help or a hindrance?  I only use it when the step is unambiguous, so specifying a column is a good idea if and only if it makes the notation easier to read.  I can imagine that it is easier to read without the 'c', because there is one less thing to pay attention to, and it will be clear from the position which dog is meant.  On the other hand, I can imagine that reading the 'c' will be less fuss than checking which dog can move.  Time will tell.
Title: Re: Short notation for Arimaa games and positions
Post by chessandgo on Jan 27th, 2009, 9:34am
"The number used to uniquely identify the piece can also be omitted if it can be determined from the direction of the steps. For example if only one horse can move east then H> is sufficient."

As I have told Omar in our email conversation, I find the new notation(s) terribly hard to read for a human being. I think the length of the notation is much less important than the speed at which a human can understand the move denoted. When gaining lentgh thanks to an unambiguous step like R> if only one rabbit can go eastwards, one also makes things harder to read for a human, who has to check which of the rabbits is the one that can go eastwards. Oh it's the d3 rabbit, so the move means that the d3 rabbits goes east? Why isn't it better to write rd3w, which basically contains all information that we have to recover ourselves by reading R>??

I'm sorry to be stubborn, but for the moment I would be absolutely sad to have to change the notation in my "book", if it ever gets published. My personal feeling, which is possibly not shared, is that the short notation makes things harder for the reader.

Jean
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Jan 27th, 2009, 12:26pm
I'm ambivalent about the case of omitting the disambiguation when there is only one piece of a type that can move a certain direction.  My proofreader (Elmo) is urging me to specify the piece location in that case too, so I probably will.

In my book, there wasn't a single case of a rabbit that didn't need to be disambiguated.  I always was forced to write Rc^ or R4^, and sometimes even Rc4^ because the rabbit shared a row with one rabbit and shared a column with another.  You may ask what the new notation gains if I have to use the same four characters as before; the fact is that it is extremely common for the same piece to move more than once.  In one diagram I could write Re5vvvv instead of having to write Re5s Re4s Re3s Re2s.  In the old notation, you have to spend a split-second on every step to verify that, yes, it is the same rabbit moving again.

Also a large percentage of moves are done by the elephant and camel, which don't ever need to be disambiguated, so omitting the origin square is easier to read without causing any confusion.

I was even surprised how often there was a double-pull or double-push.  Once you get used to the new notation, EH^^* is not only more compact, it is much easier to decipher than Hc4n ed4w Hc5n Hc6x ec4n.

Putting all the arguments aside, though, I strongly believe that nobody should be "forced" to use any particular notation.  Everyone should be allowed to use the notation that is most convenient.  The weight of popular opinion will enforce a standard well enough.  For chess books nobody can force you to write 1.c4 instead of 1. PQB4.  In fact, a lot of authors resisted the transition because they thought the descriptive notation was better.  Eventually the transition happened because most people liked algebraic notation better, and the people who didn't like it better gave in so that they would be understood by the majority.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Jan 27th, 2009, 12:58pm

on 01/27/09 at 09:34:35, chessandgo wrote:
"The number used to uniquely identify the piece can also be omitted if it can be determined from the direction of the steps. For example if only one horse can move east then H> is sufficient."


You can always specify the piece location if you want. I am just saying that in such situations the author can chose to leave it out if they want. If I didn't say that then authors would be required to specify the piece location even in these situations. I don't know enough right now to know if it should be required or not. So I am leaving this aspects of the notation up to the authors. I think over time we will determine it one way or the other based on usage. We may even decide that if there are many rabbits to specify the location and not specify it if there is just a few.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Jan 27th, 2009, 3:11pm

on 01/26/09 at 12:10:23, Fritzlein wrote:
Omar, do you intend to rename the squares too, so that a1 for Gold is h8 for Silver?  Or do you keep the square names the same and just reverse the arrows?  This wasn't an issue before, because we didn't use square names for disambiguation, but now it will matter.


Yes, once the view is specified, then the square a1 is at the bottom left corner. If the view is not specified then it is assumed to be from Gold's side.

I hope authors will use the view that is most natural for the situation. For example if you are presenting "silver to move and win in x" then the preferred view would be from silvers side. If you are providing comments about a game without a bias towards one side or the other (such as the event reports of the WC games) then a side neutral view would be preferred.
Title: Re: Short notation for Arimaa games and positions
Post by chessandgo on Jan 28th, 2009, 3:07am
true, Karl, the new notation is better when the same piece moves again, and for M and E. It might be confusing though to write E for both e and E as is the case in at least one of the proposal.

Ok Omar, I just thought you'd try to standardize the notation. Well then, we'll see what turns out to be preferred by most :)

Jean
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Jan 28th, 2009, 8:33am
Brainstorm:  If we label the columns on the board stuvwxyz instead of abcdefgh, then the square letters don't conflict with with any of the piece letters.  Then we could still use 'E' for the Gold elephant and 'e' for the Silver elephant.

But in my experience the big problem to solve was the disambiguation problem.  The confusion caused by using capital letters for pieces on both sides is much less.
Title: Re: Short notation for Arimaa games and positions
Post by 99of9 on Jan 28th, 2009, 5:44pm
I don't like the silver pieces being written as capitals.  It is much easier for the reader to get the gist quickly when the capitalization helps identify the piece (and it uses no extra characters).

I also don't like changing the names of the columns (I think a-h are standard enough that we should resist changing them).

To avoid confusion between pieces and columns, I suggest not dropping the number in the square identification.  That way, anything with a number after it is a column.

As you say Karl, the major advantage of this notation is when one piece moves more than once.  Therefore it doesn't matter too much if you spend a couple of extra characters giving information about the piece and it's location in the first place:

Ed2^^^^ is nice and short, and is instantly recognizeable even if you don't know the current board state.

I personally am unlikely to ever use the silver perspective, and I think relabelling the squares from that perspective will only confuse people.
Title: Re: Short notation for Arimaa games and positions
Post by camelback on Jan 28th, 2009, 7:09pm
I was thinking the same as 99of9 and I completely agree with him.
Title: Re: Short notation for Arimaa games and positions
Post by chessandgo on Jan 29th, 2009, 1:00am
agreed with 99of9 and camelback :)
Title: Re: Short notation for Arimaa games and positions
Post by Adanac on Jan 29th, 2009, 4:42am

on 01/29/09 at 01:00:25, chessandgo wrote:
agreed with 99of9 and camelback :)


I agree with 99of9, camelback and chessandgo.   :)

I like a system that's convenient and intuitive, but the rules should also be consistent and unambiguous (i.e. always from gold's perspective, no ambiguity between column d and dog, etc.)
Title: Re: Short notation for Arimaa games and positions
Post by jdb on Jan 29th, 2009, 6:39am

on 01/29/09 at 04:42:24, Adanac wrote:
I agree with 99of9, camelback and chessandgo.   :)

I like a system that's convenient and intuitive, but the rules should also be consistent and unambiguous (i.e. always from gold's perspective, no ambiguity between column d and dog, etc.)


I agree.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Jan 29th, 2009, 5:28pm
I like the idea of using capital letters for gold and lower case for silver. Although it should be noted that it is not required and can be determined from context in the notation. However in discussions we will still want to use upper and lower case when referring to the pieces quickly, so it would be nice to preserve that in the notation.

But I also like the notation the way it is because it allows me to uniquely identify the piece by using just the column or row. In practice there are lots of times when only a column or row is sufficient to uniquely identify the piece (always for horse, dog, and cat; and usually for rabbits). This is nice not only because it saves a letter, but because it is less effort; both when recording and replay. Also I can easily tell the piece characters apart from the location characters.

I really would suggest people to record the moves on paper next time they play or review a game; then open a plan window and make the moves by looking at only the notation on paper. It's one thing to talk about it in theory and another thing to do it in practice.

I am still undecided about Toby's suggestion and I am going to try it both ways before deciding.

If only we had another ordered set of common characters :-) Karl's suggestion does try to provide this, but I think I will have a harder time finding the columns with this set then I already do with the a-h set. Also the character 'v' will cause confusion with the down direction character; for example Hvv. It is most intuitive for me to find rows and columns using  numbers for both. I can much more easily find column 6 than I can column 'f'. However, Karl mentioned that in discussions the squares would look a bit like move numbers; as in "on 23 the E moved to 23". The first is the move number and the second is the square location. It gets harder for something like "I should have captured the H on 23" :-)
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Jan 30th, 2009, 6:46am

on 01/29/09 at 17:28:26, omar wrote:
I really would suggest people to record the moves on paper next time they play or review a game; then open a plan window and make the moves by looking at only the notation on paper. It's one thing to talk about it in theory and another thing to do it in practice.

I second this sentiment.  This is not to say that when everyone has tried out the various proposals we all will agree on what is best; I just mean that I had strong intuitions about what short notation would be best, but when I actually tried to use it for my book, what worked best was different than I expected.  
Title: Re: Short notation for Arimaa games and positions
Post by clauchau on Feb 3rd, 2009, 5:02am
After 15 years of playing with computer chess and a lot of variants and similar games by emails, and logging the moves of several game engines of mine, I'm still mentally uncomfortable with identifying columns by their traditional letters on a 8x8 board, except by keeping an eye on a board with labels on every side. Columns f and g are the most difficult to identify to me. I always need a couple of seconds. Columns a and b are the only ones I never mentally hesitate upon.

Row numbers aren't totally easy either, though they are easier than letters.

More compelling, even after 6 years of playing Arimaa, I don't remember having ever gone through the hurdle of thoroughly reading and fully understanding any noted move in the old fashioned way. It has always looked senseless enough to keep me from trying. I don't like how ee6e doesn't strike me at first glance.

Ideally, I'll go for small animal pictures, some symbolic graphics denoting the square and fat arrows. Meanwhile, on limited computer files, screens and keyboard, I vote to make the pieces look as distinct as possible from the squares. I rather agree with using capital letters for pieces, both Gold and Silver.

Maybe we could relabel the columns  ab jk   st yz ?  As soon as I would see t, I would easily see it belongs to the right half of the board, left half of that half, right half of that half of half.

Maybe if one insist to show Gold from Silver, one could prefix any move of Silver's with a minus.
Title: Re: Short notation for Arimaa games and positions
Post by RonWeasley on Feb 3rd, 2009, 8:56am
We learned the letter order with a kid's song that goes:  Albino Bats Climb Down Exits From Glass Houses.  The tune just follows all the notes on the scale.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Feb 3rd, 2009, 12:09pm

on 02/03/09 at 05:02:17, clauchau wrote:
After 15 years of playing with computer chess and a lot of variants and similar games by emails, and logging the moves of several game engines of mine, I'm still mentally uncomfortable with identifying columns by their traditional letters on a 8x8 board, except by keeping an eye on a board with labels on every side. Columns f and g are the most difficult to identify to me. I always need a couple of seconds. Columns a and b are the only ones I never mentally hesitate upon.

Wow, I also have difficulty using letter for the columns. I am OK with a-c, but have to think twice for the rest. I feel like I'm confessing at an AA meeting :-)

Your suggestion for the labeling might be better, but still would require a little getting used to. Using numbers I think is still the most natural and intuitive. I have a feeling that the chess folks used letters for one of the dimension just so there would not be any mix up with which number goes with which dimension.

Yesterday I tried recording the Adanac-Fritzlein game from round 3 using the notation suggested by 99of9, the notation currently described on the short notation page and the original short notation which used the scans. I did this from the point of view that I am playing an OTB game and have to record the moves myself fairly quickly; although I recorded the moves in a computer file rather than on paper. I also timed myself to see how much time I spent on each.

99of9 - using case sensitive piece letters and disambiguating by specifying the square location with letters for columns and numbers for rows took about 35 minutes.
http://arimaa.com/arimaa/learn/shortNotationTest/99

Fritzlein - using upper case for all pieces and disambiguating by specifying the row or column if possible; otherwise specifying the square with letter and number took about 29 minutes.
http://arimaa.com/arimaa/learn/shortNotationTest/kj

original short notation - using lower case for all pieces and disambiguating by scan number took about 26 minutes.
http://arimaa.com/arimaa/learn/shortNotationTest/orig

Though I was trying to do it briskly I was replaying the moves a few times to make sure I got it right. Some things I noticed were:

* using case sensitive letters for the pieces adds a little delay to make sure you have the right case; though this is good for readability it makes recording a bit slower; I wouldn't mind doing it in a write up of a game, but wouldn't want to do it while also playing the game.

* using square location to disambiguate adds a lot of redundancy; most of the time the column is sufficient to disambiguate and having to specify the row just slows you down; I definitely wouldn't want to do this while also playing the game.

* when using columns or rows to disambiguate there were only two times during the whole game that is was not enough and both column and row had to be specified.

* the scan number works really well for horse, dog and cat; even faster than using rows or columns, but rabbits slow you down and you have to carefully count.

* I could feel that the columns being letters was slowing me down; wished the columns were numbers.

* it felt a bit unnatural to write the weaker piece first in a pull; there seems to be a tendency to want to write the stronger piece first.

I then tried somewhat of a hybrid approach using scan number for the horse, dog and cat and columns for the rabbits and specifying column and row only if needed. I also used numbers for the columns so that only one number after a rabbit is the column number and two numbers is column and row. I also used a local scan to specify the weaker piece if needed in a push/pull. Even for a pull I wrote the stronger piece first and used a '/' character to specify that the next direction character is for a pull. I used lower case for all pieces to making the recording faster, but it could be done case sensitive also.
http://arimaa.com/arimaa/learn/shortNotationTest/test

Recording the moves this way felt very easy. I was able to record the game in about 20 minutes. I don't think it's possible to make the recording any faster. Also it is easier to read when the columns are not letters; thus letters are only for pieces; numbers for disambiguation and special characters (except for v) for directions and captures. Of course case sensitive letters could be used if one is not in a hurry.

Now the only thing bothering me was what if someone in a game write up wanted to refer to the square location using numbers for both the column and row; how could you distinguish that from a move number. For example "I should have captured the rabbit on 23". Is that move 23 or the square 23. Usually we specify the moves by including a g or s after them to indicate the side to move so this could help distinguish it. But what I am going to suggest is that a '#' character be used in front of a square location to make it absolutely clear. Thus if you ment the rabbit on square 23 you would write it as: "I should have captured the rabbit on #23".

The other consideration is what if someone wants to use the standard 'a1' type of square locations in the write up of a game. I am going to suggest that you can also use that. However, where as #11 would be at the lower left corner of the specified view the square a1 will always be at the lower left corner of the gold player. I think people are more accustomed to it being that way.

I also considered using a column number and if needed a row for the horse, dog and cat. It would eliminate a bit of exception in the notation rules, but I know from practical experience that a scan between two pieces is much faster than what column they are on; even when the columns are numbers. If I was recording while playing I would definitely prefer to use the scan for the horse, dog and cat.

I know we've been experimenting a lot with making the short notation better and perhaps you are getting tired of me changing it so much. But I am starting to feel as if this version is perhaps the fastest for recording and visually clearest for reading. I would suggest others to really try it out and see how it feels to you.
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Feb 3rd, 2009, 2:47pm
I'm very reluctant to change the notation in my book again, given how close I am to publishing and given how much time it takes to make the change.  If, however, we are not discussing what goes in my book, and we are instead discussing the future evolution of Arimaa notation, I think there is still room for improvement.  We need to keep comparing systems and see what works the best for the most people.


on 02/03/09 at 12:09:27, omar wrote:
Wow, I also have difficulty using letter for the columns. I am OK with a-c, but have to think twice for the rest.

Hmm, I don't have any trouble with using letters for columns.  If you tell me "f4" I can point to the square without hesitating or hitting e4 or g4 by accident.  Surely this is a product of my chess upbringing.

On the other hand, because I am not used to it, I would have to hesitate between square 65 and square 56.


Quote:
I then tried somewhat of a hybrid approach using scan number for the horse, dog and cat and columns for the rabbits and specifying column and row only if needed.

You don't think it will be confusing that h2 means the second horse but r2 means the rabbit on the second column, unless it is a push when er2 means the second rabbit the elephant can push?  That's a triple-meaning for the disambiguating '2'!  One advantage of the previous iteration was that the disambiguation was consistent, so that the 'f' in Hf and Rf and ERf always meant the f-column.


Quote:
I also used numbers for the columns so that only one number after a rabbit is the column number and two numbers is column and row.

There is a slight loss of efficiency there.  In the short notation currently in my book I can write R7 for my one and only rabbit on the seventh row (and did a couple of times), because we know the 7 isn't the 7th column.  In your latest suggestion even if there is only one rabbit on the 7th row, you have to write the column too before disambiguating.


Quote:
Even for a pull I wrote the stronger piece first and used a '/' character to specify that the next direction character is for a pull.

I confess that the notation currently in my book is confusing on the score of pulls.  It is more intuitive to always write the stronger piece first.  The problem is exacerbated because the pieces are all capital, so you don't know which piece is yours and which is the opponent's.  If the stronger piece always came first, that would also mean that your own piece always comes first, which is how we think about it intuitively.

Unfortunately, not only does your suggestion add another character to indicate pulls, it also moves the direction far from the piece moving in that direction.  In 'md/^' it is the camel moving north, not the dog or the slash, so you have to get used to commuting the ^ across two characters to the piece that is actually moving ^.  Still, that may be less confusing than trying to sort which of two pieces is the stronger.  I would have to try it out to know.

An alternative to consider is 'm^/d' which reads as "camel moves up, pulls dog"


Quote:
Now the only thing bothering me was what if someone in a game write up wanted to refer to the square location using numbers for both the column and row; how could you distinguish that from a move number. For example "I should have captured the rabbit on 23". Is that move 23 or the square 23. Usually we specify the moves by including a g or s after them to indicate the side to move so this could help distinguish it. But what I am going to suggest is that a '#' character be used in front of a square location to make it absolutely clear. Thus if you ment the rabbit on square 23 you would write it as: "I should have captured the rabbit on #23".

I note that this is something that you haven't tried out.  You can't really try it out until you try to write text about a position as well as trying to notate moves.  I refer to squares all the time in my book, and I am not persuaded that #23 would read nearly as smoothly as b3.  I believe that we would be trading away ease of reference to squares in order to get the benefit of being able to use lower-case letters for pieces.


Quote:
I would suggest others to really try it out and see how it feels to you.

Yes, I'll try it, but not for this book!  Maybe by the time I come out with a second edition of my book, Arimaa notation will have evolved further, and I will want to replace the notation then.
Title: Re: Short notation for Arimaa games and positions
Post by clauchau on Feb 4th, 2009, 8:46am
I like Omar's last scheme best :P

Have you noticed that about half the disambiguating characters could be further omitted if you added that in case of ambiguity, the choice closest to the last moved piece would prevail? Compare the beginning of Adanac-Fritzlein game as Omar last wrote it

2w e^^^h1^, evvd2vh1v
3w h1^^<m<, ev<vmv
4w e<h2^c2<r6^, e>ec/<h2v
5w r5^e<eh/>, ec^d1vv
6w eh/>m^h1^, emvdvc1v
7w d2<^m<d1^, mvvmh/^
8w ed>v, e>>er/<*
9w m^^h>r3^, r3vr3vh1>>
10w r8^^h1vr5>, r8vvc2vv

with what it becomes with that additional disambiguating rule :

2g e^^^h^, evvdvh1v
3g h^^<m<, ev<vmv
4g e<h^c<r6^, e>ec/<hv
5g r5^e<eh/>, ec^dvv
6g eh/>m^h^, emvdvcv
7g d<^m<d1^, mvvmh/^
8g ed>v, e>>er/<*
9g m^^h>r^, rvrvh1>>
10g r8^^hvr5>, r8vvcvv
Title: Re: Short notation for Arimaa games and positions
Post by clauchau on Feb 4th, 2009, 8:58am

on 02/03/09 at 14:47:13, Fritzlein wrote:
You don't think it will be confusing that h2 means the second horse but r2 means the rabbit on the second column, unless it is a push when er2 means the second rabbit the elephant can push?  That's a triple-meaning for the disambiguating '2'!


Here is a nice fix : Let's not use the digits 1 or 2 when disambiguating by scan. Instead of 1, use nothing, and instead of 2 use "

In other words, digits are only for rabbits !

Now what Omar wrote would become :

2w e^^^h^, evvd"vhv
3w h^^<m<, ev<vmv
4w e<h"^c"<r6^, e>ec/<h"v
5w r5^e<eh/>, ec^dvv
6w eh/>m^h^, emvdvcv
7w d"<^m<d^, mvvmh/^
8w ed>v, e>>er/<*
9w m^^h>r3^, r3vr3vh>>
10w r8^^hvr5>, r8vvc"vv

and if we add disambiguating by shortest distance :

2g e^^^h^, evvdvhv
3g h^^<m<, ev<vmv
4g e<h^c<r6^, e>ec/<hv
5g r5^e<eh/>, ec^dvv
6g eh/>m^h^, emvdvcv
7g d<^m<d^, mvvmh/^
8g ed>v, e>>er/<*
9g m^^h>r^, rvrvh>>
10g r8^^hvr5>, r8vvcvv
Title: Re: Short notation for Arimaa games and positions
Post by Janzert on Feb 4th, 2009, 9:14am
I think there is some inherent tension in the goals between making a format that is easy to write and one that is easy to read. It seems that a lot of the effort here is in making something easy to write. I would argue that for most situations the ease of reading is much more important.

Janzert
Title: Re: Short notation for Arimaa games and positions
Post by Fritzlein on Feb 4th, 2009, 10:59am
I totally agree with Janzert about the tradeoff between ease of reading and ease of writing.  I could record games even faster than Omar's fastest if I didn't disambiguate pieces at all.  Unfortunately then it would be unreadable.

A good test of readability is to ask someone else to recreate the game you have recorded, so that the actual move isn't still lingering in your mind from when you recorded it.  It's not fair to be the one to test decoding if you did the encoding yourself.  Omar, perhaps you can enlist Aamir for this task; each of you record a different game, and then each of you play back the game the other person has recorded.

It was instructive for me to have Elmo reading the text of my book.  She considered the old notation unwieldy, and prefers using arrows.  However, her reactions convinced me that I can't be satisfied just because the notation is unambiguous.  At one point I wrote C^, and didn't put a letter to show which cat was going north.  It was unambiguous because only one cat could go north.  Unfortunately the time it takes for a reader to determine that the other cat can't move that way is greater than the time it takes them to read an extra character in Cd^ and locate the cat on the d-file.  So I have already given up some compression in favor of faster reading.

I think clauchau's disambiguation by proximity to the last piece moved is very clever, since we usually move pieces in the same quadrant.  I want to try this scheme out, because it may be fast to read.  Expectations are important to fast comprehension, and the reason I will not be disambiguating elephants ever.  In the move Ed4^, I was expecting the elephant to be on d4, so the redundancy just slows me down, and E^ is faster.  Similarly when I see E^C> I expect that the cat move will be related to the elephant move, and if my expectations are true, it can be faster to read than E^Cd>.  

But the metric we use for judging should definitely not be as simple as, "I used fewer characters," or, "I recorded the game in less time."  We also have to consider speed of replaying the game, ease of using the notation in the middle of sentences, speed of learning the notation, frequency of mistakes recording, frequency of mistakes replaying, etc.

For example, it might be that clauchau's scheme of disambiguation by proximity is both clear and fast, but error-prone because it is so easy for people to overlook the near cat and accidentally move the far cat when replaying.  Or it may happen too often that one cat is three squares away while the other is four squares away, or that they are equal distance away, or that one is nearer where the last piece to move started, while the other is nearer to where the last piece to move ended.

I love all the ideas being generated.  Let's keep at it and keep trying things out.  The way we will know what notation is best is surely not by well-defined theoretical metrics, but rather by using that notation and seeing what goes wrong with it.
Title: Re: Short notation for Arimaa games and positions
Post by chessandgo on Feb 5th, 2009, 12:39am

on 02/04/09 at 09:14:09, Janzert wrote:
I think there is some inherent tension in the goals between making a format that is easy to write and one that is easy to read. It seems that a lot of the effort here is in making something easy to write. I would argue that for most situations the ease of reading is much more important.

Janzert


This. :)
Title: Re: Short notation for Arimaa games and positions
Post by clauchau on Feb 5th, 2009, 12:42am
Well, good point, since in my eagerness to get as short a notation as possible, I was wrong suggesting using no disambiguating character to mean both the first scanned and the closest. It could mean two different choices without any way to tell which.

I now suggest c' and c" (or c0 and c9?) to mean the first and last cats in scan order.

I also admit seeing which piece is closest to another is easily prone to mistakes over long distances or along slanted paths. It also makes it dependant on the previous move, resulting in potential mistakes when quoting a step out of context. I agree we'd better stay careful on this one and use it only sparingly if ever.


on 02/03/09 at 14:47:13, Fritzlein wrote:
In the short notation currently in my book I can write R7 for my one and only rabbit on the seventh row (and did a couple of times), because we know the 7 isn't the 7th column.  In your latest suggestion even if there is only one rabbit on the 7th row, you have to write the column too before disambiguating.

It could be noted R" since it also happens to be the last rabbit in scan mode.

We could also write R07 to specify a row without specifying a column. Longer but almost as fast.
Title: Re: Short notation for Arimaa games and positions
Post by omar on Feb 9th, 2009, 5:37pm

on 02/03/09 at 14:47:13, Fritzlein wrote:
I'm very reluctant to change the notation in my book again, given how close I am to publishing and given how much time it takes to make the change.  If, however, we are not discussing what goes in my book, and we are instead discussing the future evolution of Arimaa notation, I think there is still room for improvement.  We need to keep comparing systems and see what works the best for the most people.

I think what you have in the book now is much easier for humans than the computer friendly notation. I think we've taken a major leap already towards human friendly notation and the finer details usually take more time, effort and experience to settle down.


Quote:
Hmm, I don't have any trouble with using letters for columns.  If you tell me "f4" I can point to the square without hesitating or hitting e4 or g4 by accident.  Surely this is a product of my chess upbringing.

On the other hand, because I am not used to it, I would have to hesitate between square 65 and square 56.


This will vary from person to person and for people who are not heavily into chess notation the numbers would probably be easier. Also not using letters to specify the squares allows both upper and lower case to be used for pieces. Also it is nice that most of the time you never have to specify the square and columns are usually sufficient; at least for the g and s views.


Quote:
You don't think it will be confusing that h2 means the second horse but r2 means the rabbit on the second column, unless it is a push when er2 means the second rabbit the elephant can push?  That's a triple-meaning for the disambiguating '2'!  One advantage of the previous iteration was that the disambiguation was consistent, so that the 'f' in Hf and Rf and ERf always meant the f-column.

It would seem that this would be confusing, but when I actually did it wasn't confusing at all.


Quote:
There is a slight loss of efficiency there.  In the short notation currently in my book I can write R7 for my one and only rabbit on the seventh row (and did a couple of times), because we know the 7 isn't the 7th column.  In your latest suggestion even if there is only one rabbit on the 7th row, you have to write the column too before disambiguating.

Yes, there is some loss of efficiency, but it's not much since specifying a square is usually not needed and most rabbit disambiguation can be done by columns. Using letters to specify columns does make things like Rc> look a bit confusing (cat pulls rabbit east or rabbit on column c moves east).


Quote:
I confess that the notation currently in my book is confusing on the score of pulls.  It is more intuitive to always write the stronger piece first.  The problem is exacerbated because the pieces are all capital, so you don't know which piece is yours and which is the opponent's.  If the stronger piece always came first, that would also mean that your own piece always comes first, which is how we think about it intuitively.

Unfortunately, not only does your suggestion add another character to indicate pulls, it also moves the direction far from the piece moving in that direction.  In 'md/^' it is the camel moving north, not the dog or the slash, so you have to get used to commuting the ^ across two characters to the piece that is actually moving ^.  Still, that may be less confusing than trying to sort which of two pieces is the stronger.  I would have to try it out to know.

Give it a try, it didn't feel confusing to me. The way I think of it is first linking the two pieces together and then moving them based on the direction characters that follow. One thing that can be done when you do it this way is to easily mix push/pull combinations. For example Eh/^> means elephant goes north pulling the horse and then pushs the horse east (the / applies only to the direction character immediately after it)


Quote:
An alternative to consider is 'm^/d' which reads as "camel moves up, pulls dog"

But could look a bit confusing if the next step was a cat movement: m^/dc>


Quote:
I note that this is something that you haven't tried out.  You can't really try it out until you try to write text about a position as well as trying to notate moves.  I refer to squares all the time in my book, and I am not persuaded that #23 would read nearly as smoothly as b3.  I believe that we would be trading away ease of reference to squares in order to get the benefit of being able to use lower-case letters for pieces.

I would reccomend continuing to use the a1 type of notation in the text. In the move notation squares are rarely specified.


Quote:
Yes, I'll try it, but not for this book!  Maybe by the time I come out with a second edition of my book, Arimaa notation will have evolved further, and I will want to replace the notation then.

I think what would really help get the notation hashed out is if we had an OTB tournament and had to record our moves by hand. It's hard to force yourself to do that when you don't have to :-)
Title: Re: Short notation for Arimaa games and positions
Post by omar on Feb 9th, 2009, 5:49pm

on 02/05/09 at 00:42:24, clauchau wrote:
Well, good point, since in my eagerness to get as short a notation as possible, I was wrong suggesting using no disambiguating character to mean both the first scanned and the closest. It could mean two different choices without any way to tell which.

I now suggest c' and c" (or c0 and c9?) to mean the first and last cats in scan order.

I also admit seeing which piece is closest to another is easily prone to mistakes over long distances or along slanted paths. It also makes it dependant on the previous move, resulting in potential mistakes when quoting a step out of context. I agree we'd better stay careful on this one and use it only sparingly if ever.

It could be noted R" since it also happens to be the last rabbit in scan mode.

Interesting; I'll try out these suggestions.


Quote:
We could also write R07 to specify a row without specifying a column. Longer but almost as fast.

Good suggestion; this would be useful for the 'n' view. Yes, less precision but still unambiguous is what seems to makes the notation faster and less error prone. Perhaps rabbit on row 7 could be written more visually as R=7 (the = kind of looks like rows).
Title: Re: Short notation for Arimaa games and positions
Post by omar on Feb 9th, 2009, 6:01pm

on 02/04/09 at 09:14:09, Janzert wrote:
I think there is some inherent tension in the goals between making a format that is easy to write and one that is easy to read. It seems that a lot of the effort here is in making something easy to write. I would argue that for most situations the ease of reading is much more important.


I am finding that its not so much the number of characters that make the notation easier to write, but rather having to specify less is what makes it easier; and less error prone.

As for easier to read; I am finding that using different types of characters for different things (piece type, direction, disambiguation) makes it easier for me to read.
Title: Re: Short notation for Arimaa games and positions
Post by mattj256 on May 12th, 2015, 10:33am
Hi all.

There's a community volunteer effort under way to create a new Arimaa website.  (front end and back end.)

As part of the discussion, some members of the community have mentioned that they don't like the standard Arimaa notation.  I think it's extremely important that the new site support standard Arimaa notation, and I'm open to also supporting other notations as well.

I wasn't sure if I should start a new thread or resurrect an old one.
If people have specific recommendations and if there's consensus then we can talk about implementing other notations as options.  
(This is assuming that Omar gives his stamp of approval.)  

Some questions to think about:

1. What are the most important features that a good notation should have?
2. Who is the primary audience?  (humans or bots)
3. What features do we like in other notations?
4. What features do we not like in other notations?

Thanks in advance for your help.
Title: Re: Short notation for Arimaa games and positions
Post by half_integer on May 12th, 2015, 7:06pm
I agree that current notation should be kept and any new notations should be provided as options.

Given the large number of variations that can arise in choosing a notation, I suggest that when the community is polled, they should be asked about each attribute separately rather than voting on a whole notation scheme.  Some sample choices to be made:

- origin squares or destination squares
- keep columns as letters or switch to numbers
- capitalization for color or to represent something else
- can multiple steps of a single piece be shown as one move
- how are pushes, pulls, and capture represented

I do suggest that, since we already have a computer-readable standard, any alternate notation be aimed at ease of comprehension for humans - and beginner humans at that.

I also suggest that since this is already being converted to and from some internal representation, some of the choices can be left as options for the user (either for entry or for display).  For instance, there is no reason that both of (nsew) and (^v<>) can't be valid (along with graphic arrows when available).

I have my own preferences; I look forward to the discussions.
Title: Re: Short notation for Arimaa games and positions
Post by CraggyCornmeal on May 12th, 2015, 7:34pm
I believe I'm in the minority, but I actually like the standard notation. It's not as condensed at it could be, but it disambiguates everything (and does so in a consistent way!), leaving no space for misinterpretation.

There is, however, one element of standard notation that can be easily condensed without creating any ambiguity. When a piece takes multiple steps in a turn, we only need to specify the piece and its location once:

Dg4n Dg5n Dg6e would become

Dg4nne

An exception would be made when the order of steps is important:

Dg4n rh4w Dg5n Dg6e would become

Dg4n rh4w Dg5ne

I also think we should keep using nsew for directions rather than switching to ^v><. Because if you're viewing a game from silver's perspective, ^v>< is ambiguous, while nsew is clear.
Title: Re: Short notation for Arimaa games and positions
Post by clyring on May 12th, 2015, 8:00pm
CraggyCornmeal's mentioned unique compressed version of the standard notation is, incidentally, as far as I know, the only implemented (http://pastebin.com/8cQn2fjF) compressed notation for interacting textually with the Arimaa gameroom. I personally like omitting more characters when giving moves by hand in chat, but find that the one optimization of compressing consecutive steps by the same piece already makes the standard notation conveniently usable the majority of the time.

I also agree with his complaint about iconographic direction ^>v<. From the silver perspective, we must accept either lying symbols or the need to disambiguate by perspective, both of which seem ugly. Moreover, <> and especially ^ are awkward to type while nsew flow naturally with the use patterns and key locations on most English keyboards.
Title: Re: Short notation for Arimaa games and positions
Post by lightvector on May 12th, 2015, 10:35pm
I have an idea for a notation that I haven't implemented, but I feel fairly certain I could do so. I think I've described it before in a chatroom message. The notation targets ease of human use and is based in empirically observing the ways in which most people in the chatroom tend to try to specify moves when commenting on games. It heavily depends on knowing the current board position. However, it may be a poor choice for a "standard" notation because in general moves will have a large number of ways to be written in this notation, with no canonical format. It does however include the current standard notation as a special case.

Gritty details:
(skip below for examples if you just want to get the idea):

The notation consists of a sequence of tokens that specify constraints on the move or the position resulting from the move.
Tokens are of the form <piece specifier><property>.
Such a token indicates that <property> is true about the move regarding a piece satisfying <piece specifier>.

A <piece specifier> is one of the following
* [RCDHMErcdhme]: a piece of the specified type
* [a-h][1-8]: a piece in the specified location
* [RCDHMErcdhme][a-h][1-8]: a piece of the specified type in the specified location

A <property> is one of the following:
* [a-h][1-8]: A piece satisfying <piece specifier> at some point during the move ended at the specified destination.
* [news]+: ("steps") A piece satisfying <piece specifier> at some point during the move stepped in these directions in order and the steps began at a time when it did satisfy <piece specifier>.
* [x]: A piece satisfying <piece specifier> at some point during the move was captured.

A move satisfies a sequence of tokens if it satisfies all of the constraints represented by each token AND if that the steps of the move satisfying any of the "steps" tokens occur in the order that they were written in the move. (This is so that "c3n c4n b4e c4e" and "c3n c4e c4n b4e" are distinguishable despite containing the same set of tokens). Aside from the relative ordering of the steps tokens, the ordering of the tokens is otherwise irrelevant.

To find the move (up to equivalence) specified by a sequence of tokens:
* Partition the set of all pseudo-legal moves (legal moves, except not modded-out by equivalence, and not checking 3x repetition legality) into a ordered sequence of buckets: "contains 1 step", "contains 2 steps and moves 1 piece", "contains 2 steps and moves 2 pieces", "contains 3 steps and moves 1 piece", "contains 3 steps and moves 2 pieces", ...
* Find the first bucket that contains any moves that satisfy the sequence of tokens.
* If all such moves in that bucket are equivalent - lead to the same board position - then the sequence of tokens is considered to uniquely specify this move. Otherwise, the sequence of tokens is ambiguous and does not uniquely specify a move.

Examples:

http://arimaa.com/arimaa/notconv/old/boardimg.php?orient=n&size=300&imgtype=png&ranks=2r/2rcrcrr/1d2E1Hr/R5rR/dDm1M2h/hH1e/RrR2CR/RR2DC1R

Gold move notation examples:
"cx b4n" = "Ee6e Ef6w cf7s cf6x Db4n"
"Ra7" = "Db4n Ra5n Db5w Ra6n"
"Ed4 Me" = "Ee6w Ed6s Ed5s Me4e"
"Dww Cf2ww" = "De1w Dd1w Cf2w Ce2w"

Silver move notation examples:
"f3x" = "ed3e ee3s Cf2n Cf3x ee2e"
"f2x" = "ed3e ee3s Cf2n Cf3x ee2e"
"mc2 Re" = "Mc4s Rc2e Mc3s pass"
"Rc3 da6" = "ed3w ec3e Rc2n db6w"
"Dc5" = "mc4e Db4e Dc4n md4w"
"b4c5" = "mc4e Db4e Dc4n md4w"
"hb3" = "ed3w ec3e Hb3e ha3e"

I think this notation shines for most sorts of "combo" moves where the steps are related, and also for common patterns like flips or pull-and-replace.
Title: Re: Short notation for Arimaa games and positions
Post by mattj256 on May 13th, 2015, 1:10am

on 05/12/15 at 19:06:56, half_integer wrote:
I do suggest that, since we already have a computer-readable standard, any alternate notation be aimed at ease of comprehension for humans - and beginner humans at that.
That's an interesting point.  I agree that the standard notation is already fine for computers, so any other options can focus completely on what's best for humans.


on 05/12/15 at 19:06:56, half_integer wrote:
I suggest that when the community is polled, they should be asked about each attribute separately rather than voting on a whole notation scheme.
Some other factors:
- terse vs. verbose
- easy or hard for a beginner to disambiguate
- easy or hard for a beginner to determine if this is a legal move
- reversible or irreversible.  (Can you easily undo a move if given the move and the resulting board position?)


on 05/12/15 at 22:35:54, lightvector wrote:
"Ra7" = "Db4n Ra5n Db5w Ra6n"
There seems to be a tradeoff between terseness and beginner-friendliness.  Saying "I want to move the rabbit to a7" and not specifying that the dog has to come over to unfreeze the rabbit is not beginner-friendly.  This doesn't mean it's a bad notation, but it's optimized in a way that will be hard for beginners to read or write.  If you give this notation to a few beginners I am confident that at least one will not realize the rabbit is frozen and parse the move incorrectly.

This is analogous to the difference between WAV and MP3 music files.  WAV is uncompressed so it can be played on very inexpensive hardware.  MP3 is compressed so it requires significant processing power to decode the file before it can be played.  The first is verbose and easy to decode; the second is terse and much harder to decode.

The two alternatives proposed so far are optimized for different purposes so I wouldn't expect that either one would "win" over the other.

If it helps the discussion, maybe we should be thinking about:
- best notatation for computers (standard notation is fine)
- best notation for human beginners
- best notation for human intermediate/advanced players


on 05/12/15 at 19:34:39, CraggyCornmeal wrote:
Dg4n Dg5n Dg6e would become

Dg4nne
This is the only nonstandard notation that I personally use, because it saves a little typing and it never introduces any ambiguity.

Some other feedback on how I would adapt lightvector's notation to make it more beginner-friendly:
- I think every piece that is moved should be mentioned explicitly, including enemy pieces.  If a piece makes no net movement, mention just the name of the piece with a space before and a space after.  The order that the pieces are mentioned doesn't have to always match the order that they move so that pull-and-replace and swings can be communicated efficiently.
- I think every piece should either include the start square or the end square, unless it didn't make any net movement.
- A capture should include the enemy piece, the square that the enemy piece started on, and some information about the piece doing the capturing.
- You didn't specify how to record every type of capture.  (suicidal step, suicidal push/pull, friendly abandonment, push/pull enemy piece onto a trip, push/pull enemy defender away from the trap.)
Title: Re: Short notation for Arimaa games and positions
Post by Boo on May 13th, 2015, 2:07am

Quote:
There's a community volunteer effort under way to create a new Arimaa website.  (front end and back end.)

As part of the discussion, some members of the community have mentioned that they don't like the standard Arimaa notation.  I think it's extremely important that the new site support standard Arimaa notation, and I'm open to also supporting other notations as well.


It is strange to see a big discussion emerging on a "nice to have" feature. I think first of all you should better spend your energy to get the basic stuff working...
It seems you have troubles with setting up the prioritized issue list :)
Title: Re: Short notation for Arimaa games and positions
Post by mattj256 on May 13th, 2015, 9:12am

on 05/13/15 at 02:07:15, Boo wrote:
It is strange to see a big discussion emerging on a "nice to have" feature. I think first of all you should better spend your energy to get the basic stuff working...
It seems you have troubles with setting up the prioritized issue list :)
This is true.  My reasons for starting this discussion:
1. I don't need community input for the low-level tasks that I'm working on.
2. I want to give the community plenty of time to discuss this.
3. Even though this isn't a high priority feature, there are (at least) two community members who are passionate about it.  From my point of view it's more efficient to acknowledge their passion and carve out a space for it than to ignore them when they bring it up over and over again.
4. If I'm going to do this, I might as well start thinking up front about how the architecture will support this, even though it won't be implemented right away.
5. Unlike software development, this is something the whole community can participate in and feel ownership of.


Arimaa Forum » Powered by YaBB 1 Gold - SP 1.3.1!
YaBB © 2000-2003. All Rights Reserved.