DipTechGeekGiggles

Blind copying the developer of the screen shot I posted. I found Kris
Craig while I was trolling SourceForge for development help. The
explanations of the test cases failed by the JS adjudicator provided
for his benefit as well as for discussion.

On Thu, 9 Jul 2009 08:29:07 -0700 (PDT)
snesin wrote:

> I have a JavaScript adjudicator here:
>
> http://GamesByEmail.com/Games/Politics/Judge

I'm pretty impressed that a JS adjudicator actually works. If nothing
else, it means that I might be able to build a native adjudicator in
the scripting language of my choice (CRM114) someday, although short
term I'm pretty focused on User eXperience and, since form follows
function, architecture.

As such, I'm interested in how you generated the "Permalinks". I'm
working on a REST architecture for a different Diplomacy server. I
think that a fast and reliable adjudicator could be built by caching
adjudication results for dependency chains. I need a computationally
cheap 2-way method for compressing partial order sets without using a
lookup.

> Below the map are a bunch of test cases. Click on each link to see the
> map setup and the results. I would expect any adjudicator to generate
> the same resolution for each.
>
> You can also modify the orders (or enter any new ones) and see the
> results instantly.


The variants from the DATC (Kruijswijk [6...]) explained:

http://web.inter.nl.net/users/L.B.Kruijswijk/#6.D.31
http://GamesByEmail.com/Games/Politics/Judge#3702c0306062S3703

Austria:
A Rumania - Armenia (Fails, no convoy)

Turkey:
F Black Sea Supports A Rumania - Armenia (Succeeds, should be illegal
order)

Because the convoy order and the support are mutually excusive, this
support order should be ruled an illegal order. In that way, if two
orders were entered for this unit - both the convoy and the support -
then the convoy would take effect regardless of the precedence rules of
the adjudicator.


http://web.inter.nl.net/users/L.B.Kruijswijk/#6.D.32
http://GamesByEmail.com/Games/Politics/Judge#1430S21552132m5502320s555542c190

England:
F Edinburgh Supports A Liverpool - Yorkshire
A Liverpool - Yorkshire (Succeeds, should fail)

France:
F London Supports A Yorkshire

Germany:
A Yorkshire - Holland (Dislodged, should hold - illegal order)

There is a difference between "no convoy" and "no convoy possible".


http://web.inter.nl.net/users/L.B.Kruijswijk/#6.G.8
http://GamesByEmail.com/Games/Politics/Judge#0422c1901932m2007032m650

France:
A Belgium - Holland via Convoy (Fails, no convoy. Should succeed.)

England:
F North Sea - Helgoland Bight
A Holland - Kiel

The intent of "via convoy" is to permit the convoy, not require it.

Chris

The programmer who created the screen shot had some questions about the
test cases. Here are the answers I had for him...

-----------------

> See below for some initial clarification questions on your examples.
> I'll need to make sure I'm clear on these before I can perform any
> tests.

It's necessary to take each of the test cases as a unit...

> > The variants from the DATC (Kruijswijk [6...]) explained:
> >
> > http://web.inter.nl.net/users/L.B.Kruijswijk/#6.D.31
> > http://GamesByEmail.com/Games/Politics/Judge#3702c0306062S3703
> >
> > Austria:
> > A Rumania - Armenia (Fails, no convoy)
>
>
> So with this one there's just an attempted move with no convoying
> fleet at all? If so, then I can already tell you from previous tests
> that this move would fail.

That's a correct adjudication, given the orders. The problem is that
the next order should not be given to the adjudicator.

> > Turkey:
> > F Black Sea Supports A Rumania - Armenia (Succeeds, should be
> > illegal order)
> >
> > Because the convoy order and the support are mutually excusive, this
> > support order should be ruled an illegal order. In that way, if two
> > orders were entered for this unit - both the convoy and the support
> > - then the convoy would take effect regardless of the precedence
> > rules of the adjudicator.
>
>
> This one confuses me. Is the move A Rumania - Armenia being convoyed
> by Black Sea? Or is Black Sea trying to do a "support move" on the
> order without a convoy?

The only unit that *could* convoy RUM - ARM is BLA (unless a coastal
convoys variant rule is in play), therefore there is no way for BLA to
provide a meaningful support for RUM - ARM. You would most likely to
see this as a player mistake in an order list like:

F BLK C RUM - ARM
....(other orders)...
F BLK S RUM - ARM

When the adjudicator attempts to validate the F BLK S RUM - ARM order,
it should not consider the unit ordering support as being available to
supply the convoy. As such, the support order would be an illegal
order. If both the convoy and the support were in the same order set
then declaring the support order to be illegal means that the
impossible support does not displace the convoy in the order scheduled
for adjudication.

> >
> > http://web.inter.nl.net/users/L.B.Kruijswijk/#6.D.32
> >
> > http://GamesByEmail.com/Games/Politics/Judge#1430S21552132m5502320s555542c190
> >
> > England:
> > F Edinburgh Supports A Liverpool - Yorkshire
> > A Liverpool - Yorkshire (Succeeds, should fail)
>
>
> Why should this move fail? Is Yorkshire occupied already?

Yes. The issue in this scenario is whether the support for Army
Yorkshire to hold matches the order for G Army Yorkshire. The scenario
is continued below...

> I tried
> visiting that link but it's not entirely clear. The map link leads
> to an image that's just too small for me to make out (I have terrible
> vision, partial blindness).

>% snip of unhelpful comment %<

I'm going to work on some SVG maps soon. I'm researching XSLT to see if
I can produce an SVG from an XML dataset using transforms, but if that
fails I'll use a templating approach.

> I was under the impression that a support move is legal so long as the
> *destination* territory is an adjacent territory that would be a
> valid move for the supporting unit type. Are you saying that the
> *move from* territory must also be adjacent? That doesn't sound
> right.

There's nothing wrong with England's support. It's whether the French
support is void...

> > France:
> > F London Supports A Yorkshire
> >
> > Germany:
> > A Yorkshire - Holland (Dislodged, should hold - illegal order)

.... Germany ordered YOR - HOL. Since that is impossible without any
fleet in the North Sea, that is an illegal order. As an illegal order,
(in the absence of any legal orders) it is rendered as a hold order for
the adjudicator, which would mean the support matches and G A YOR Holds
rather than being dislodged.

This is a classic web programming error. Data needs to test valid
before it can be used. The programmer here voided the support without
checking that the move order was legal.

> > There is a difference between "no convoy" and "no convoy possible".

If there had been a convoy route, the French support would not match
the German order regardless of whether there was an actual convoy. The
absence of a convoy *route* makes the German move illegal and renders it
as a hold order for the adjudicator.

Still almost done?

Chris

> http://web.inter.nl.net/users/L.B.Kruijswijk/#6.G.8
> http://GamesByEmail.com/Games/Politics/Judge#0422c1901932m2007032m650
>
> France:
> A Belgium - Holland via Convoy (Fails, no convoy. Should succeed.)
>
> England:
> F North Sea - Helgoland Bight
> A Holland - Kiel
>
> The intent of "via convoy" is to permit the convoy, not require it.
>
> Chris
>

Is that the intent?

The rules state:

"If none of the convoying Fleets belongs to the player whocontrols the Army,
then the land route is used. However, the player controlling the Army can
use the convoy route if he
or she indicated "via convoy" on the Army move order in question."

The rules are not well written here, and it is a very plausible
interpretation to adjudicate as an attempt to move by convoy *only*.

I'm blind copying Edi Birsan to see if he can be dragged into this
discussion.

My reason for using BCC here and elsewhere in this thread is a courtesy
to people who may not want to have their email addresses published
where they can be harvest by spambots and stalkers.

> > The intent of "via convoy" is to permit the convoy, not require it.
> >
> > Chris
> >
>
> Is that the intent?
>
> The rules state:
>
> "If none of the convoying Fleets belongs to the player whocontrols
> the Army, then the land route is used. However, the player
> controlling the Army can use the convoy route if he
> or she indicated "via convoy" on the Army move order in question."
>
> The rules are not well written here, and it is a very plausible
> interpretation to adjudicate as an attempt to move by convoy *only*.

Since you're playing Devil's Advocate here...

Give me a plausible game situation where a player would want his own
move to an adjacent province to fail because the convoy is unavailable.

My particular ax to grind is not this particular rule, but that
adjudicators (both human and automated) behave predictably. Beyond
adherence to the rules, that means explicit definition of as many of
the corner cases and ambiguities as possible. Ideally, those
definitions should be aligned with the expectations of a clear majority
of players. In the absence of a clear majority opinion on a rule - due
to fractitiousness, ambiguity or complexity - then an interpretation
should be made according to the perceived intent of the author of the
orders as communicated in the order set with the understanding that
order are *usually* written with the intent that they succeed.

There's no case to be made that "the Army can use the convoy route if
he or she indicated 'via convoy'" does not mean "must" because the
1999/2000 rules were not written to be read that precisely. They were
written to make the game more accessible to new players. Accordingly,
an interpretation of the rules cannot be made any more successful by a
deeper analysis, but by a more ready acceptance of how a casual reader
would interpret it... or put another way, "What interpretation of the
rules is least likely to cause a new player to say, 'Now that's &@^%ed
up,' and leave the table?"

Chris

"Chris Babcock" wrote in message

> I'm blind copying Edi Birsan to see if he can be dragged into this
> discussion.
>
> My reason for using BCC here and elsewhere in this thread is a courtesy
> to people who may not want to have their email addresses published
> where they can be harvest by spambots and stalkers.
>
>> > The intent of "via convoy" is to permit the convoy, not require it.
>> >
>> > Chris
>> >
>>
>> Is that the intent?
>>
>> The rules state:
>>
>> "If none of the convoying Fleets belongs to the player whocontrols
>> the Army, then the land route is used. However, the player
>> controlling the Army can use the convoy route if he
>> or she indicated "via convoy" on the Army move order in question."
>>
>> The rules are not well written here, and it is a very plausible
>> interpretation to adjudicate as an attempt to move by convoy *only*.
>
> Since you're playing Devil's Advocate here...
>
> Give me a plausible game situation where a player would want his own
> move to an adjacent province to fail because the convoy is unavailable.

That's an easy one. And E/F/G alliance is about to break down. Let's say
you are France, about to march into a 'gimme' German Hol from Bel, but you
don't want your English 'ally' to suddenly get the bright idea to scoop up
Bel for himself from the North Sea, either with the fleet, or perhaps with
an army convoyed from somewhere else, so you request he make the convoy.
This way, you will take Hol, if England is indeed trustworthy, and if not,
you are not moved out of position.




> My particular ax to grind is not this particular rule, but that
> adjudicators (both human and automated) behave predictably. Beyond
> adherence to the rules, that means explicit definition of as many of
> the corner cases and ambiguities as possible.

No argument there.



Ideally, those
> definitions should be aligned with the expectations of a clear majority
> of players. In the absence of a clear majority opinion on a rule - due
> to fractitiousness, ambiguity or complexity - then an interpretation
> should be made according to the perceived intent of the author of the
> orders as communicated in the order set with the understanding that
> order are *usually* written with the intent that they succeed.

If a clear majority are wrong, or if the rule is stated explicitly (as
opposed to programmed in a computerized adjudicator) so everyone knows what
the outcome will be, are two obvious exceptions. Also, the presumption of
an order or order set being valid is not stated in the rules, though it is a
popular adjudication choice.



> There's no case to be made that "the Army can use the convoy route if
> he or she indicated 'via convoy'" does not mean "must" because the
> 1999/2000 rules were not written to be read that precisely. They were
> written to make the game more accessible to new players.

Yup. That is indeed one of the problems with the rules.


Accordingly,
> an interpretation of the rules cannot be made any more successful by a
> deeper analysis, but by a more ready acceptance of how a casual reader
> would interpret it... or put another way, "What interpretation of the
> rules is least likely to cause a new player to say, 'Now that's &@^%ed
> up,' and leave the table?"

That is one way to approach it.