Play Diplomacy Online (beta test)

> Misunderstanding would be
> understandable and it may not be a distinction worth making, but if it
> reflects different history source information than I cited above, I'd
> love to read it. (If it's personal memories of you or someone you know,
> I want to conduct an electronic interview.)


I played diplomacy using the judges 10 years ago, when I was 14, at
which time it didn't occur to me there could be a distinction. So yes,
they are personal memories. I will adapt the About page with your
input, which is greatly appreciated.


>
> What's the state of adjudicator testing? Have you been using the
> DATC test cases (http://web.inter.nl.net/users/L.B.Kruijswijk/)?No
> fault if you haven't, but it might be helpful if you wish a systematic
> rather than (or in addition to) a stress testing approach.

I wasn't aware of these test cases, but I will certainly take a look
at this with the judge programmer (I have to be honest, I didn't do it
myself).


> but it is surprising to get kicked out to
> a start page when you are expecting the registration page.

The gateway page was being used as a teaser at first, but we are
probably going to keep it. So you are right: we will need to change
the links to the login and registration page

Thanks for your comments! I will send you an e-mail with my contact
details (the one I posted this with is invalid now).

Jan
 >> Stay informed about: Play Diplomacy Online (beta test) 

Usually I play on the judges, but I've got to say this website is
really easy to use; I've just started my first game on it, and it
looks great. I like the graphical interface and the messages from
players are kept in a very nice and simple form, which is good.

One feature that would really make the most of the graphical interface
is the possibility of taking the current board position and
adjudicating a 'practice' set of orders yourself, just to see what
would happen. In such a case some arrows that indicate what orders
you have just inputted (as in floc.net) may be useful. No idea if
this is easy or difficult to program....

Good contribution to the hobby!
Pete.
 >> Stay informed about: Play Diplomacy Online (beta test) 

Entering a position and testing adjudicator is available in the
www.stabbeurfou.org site. Use the gadgets.
Direct link below :
http://www.stabbeurfou.org/Interrogation.php

And also in the diplomatie-online french website, and not in the
18centres french website (or not to my knowledge).

The http://www.stabbeurfou.org/ has adjudication engine been fully
tested withe the DATC test cases (amongst others).
Compliant to Hasbro 2000 rules, to former French Descartes 1994 rules.
See within rules page.

Test reports are of course available on site, see below :
http://www.stabbeurfou.org/docs/regles/en/Test_report.htm

Deviations from DATC are only due to different interpretation of the
rules (land route first for convoy to adjacent place for instance).

There are to my knowledge at least two more serious automated
websites : dipbounced and dpjudge. I have not seen any test report of
their adjudicator.... phpdiplo for instance is not cited since I
understood phpdiplo does not implement the full diplomacy rules. And
there are probably more automated website around.

My personnal comment : people tend to be more interested in the
quality of the maps then the adjudication....

On Dec 21, 5:09 pm, j....TakeThisOut@de.volder.be wrote:
> >  Misunderstanding would be
> > understandable and it may not be a distinction worth making, but if it
> > reflects different history source information than I cited above, I'd
> > love to read it. (If it's personal memories of you or someone you know,
> > I want to conduct an electronic interview.)
>
> I played diplomacy using the judges 10 years ago, when I was 14, at
> which time it didn't occur to me there could be a distinction. So yes,
> they are personal memories. I will adapt the About page with your
> input, which is greatly appreciated.
>
>
>
> > What's the state of adjudicator testing? Have you been using the
> > DATC test cases (http://web.inter.nl.net/users/L.B.Kruijswijk/)?No
> > fault if you haven't, but it might be helpful if you wish a systematic
> > rather than (or in addition to) a stress testing approach.
>
> I wasn't aware of these test cases, but I will certainly take a look
> at this with the judge programmer (I have to be honest, I didn't do it
> myself).
>
> > but it is surprising to get kicked out to
> > a start page when you are expecting the registration page.
>
> The gateway page was being used as a teaser at first, but we are
> probably going to keep it. So you are right: we will need to change
> the links to the login and registration page
>
> Thanks for your comments! I will send you an e-mail with my contact
> details (the one I posted this with is invalid now).
>
> Jan
 >> Stay informed about: Play Diplomacy Online (beta test) 

On Sat, 29 Dec 2007 09:39:50 -0800 (PST)
"jeremie.lefrancois@gmail.com" <jeremie.lefrancois.TakeThisOut@gmail.com> wrote:

> Entering a position and testing adjudicator is available in the
> www.stabbeurfou.org site. Use the gadgets.
> Direct link below :
> http://www.stabbeurfou.org/Interrogation.php
>
> And also in the diplomatie-online french website, and not in the
> 18centres french website (or not to my knowledge).
>
> The http://www.stabbeurfou.org/ has adjudication engine been fully
> tested withe the DATC test cases (amongst others).
> Compliant to Hasbro 2000 rules, to former French Descartes 1994 rules.
> See within rules page.

Jérémie, is it your engine running behind playdiplomacy.com? If so, I'm
somewhat relieved that there hasn't been an enormous duplication of
effort. I know you spent about 10 years working on yours.

> Test reports are of course available on site, see below :
> http://www.stabbeurfou.org/docs/regles/en/Test_report.htm
>
> Deviations from DATC are only due to different interpretation of the
> rules (land route first for convoy to adjacent place for instance).
>
> There are to my knowledge at least two more serious automated
> websites : dipbounced and dpjudge. I have not seen any test report of
> their adjudicator.... phpdiplo for instance is not cited since I
> understood phpdiplo does not implement the full diplomacy rules. And
> there are probably more automated website around.

DP Judge, like nJudge, requires convoy routes to be specified. It's a
fairly serious flaw in the shared ancestry of both programs that I think
should be addressed in nJudge 2.0 if such a creature is ever born. DP
Judge is even less likely than nJudge to get such a serious update to
the core code.

> My personnal comment : people tend to be more interested in the
> quality of the maps then the adjudication....

That is until there is a problem with the adjudication. The also care
little about security... until they think someone is cheating. You can
always fix a map. You can fix most security problems, although you may
have to flush the user database. When an adjudicator is broken, however,
it's broken for the life of the program. At least that's what it seems
like. :-/

Anyhow, look at jDip if you want to see some good maps and RealPolitik
if you want some maps to compare to good maps.

Chris
 >> Stay informed about: Play Diplomacy Online (beta test) 

No, my engine only runs on stabbeurfou.

It did not take 10 years to build it but I have worked on it by
episodes, starting in 1992, 15 years ago then.
The only way to have a flawless adjudicator I think is to use the datc
algorithm (from my experience).

The strength of the stabbeurfou adjudicator stems in the error and
warning messages when orders are submitted - I think.


On Dec 30, 5:49 pm, Chris Babcock <cbabc....TakeThisOut@cox.net> wrote:
> On Sat, 29 Dec 2007 09:39:50 -0800 (PST)
>
> "jeremie.lefranc...@gmail.com" <jeremie.lefranc....TakeThisOut@gmail.com> wrote:
> > Entering a position and testing adjudicator is available in the
> >www.stabbeurfou.orgsite. Use the gadgets.
> > Direct link below :
> >http://www.stabbeurfou.org/Interrogation.php
>
> > And also in the diplomatie-online french website, and not in the
> > 18centres french website (or not to my knowledge).
>
> > Thehttp://www.stabbeurfou.org/has adjudication engine been fully
> > tested withe the DATC test cases (amongst others).
> > Compliant to Hasbro 2000 rules, to former French Descartes 1994 rules.
> > See within rules page.
>
> Jérémie, is it your engine running behind playdiplomacy.com? If so, I'm
> somewhat relieved that there hasn't been an enormous duplication of
> effort. I know you spent about 10 years working on yours.
>
> > Test reports are of course available on site, see below :
> >http://www.stabbeurfou.org/docs/regles/en/Test_report.htm
>
> > Deviations from DATC are only due to different interpretation of the
> > rules (land route first for convoy to adjacent place for instance).
>
> > There are to my knowledge at least two more serious automated
> > websites : dipbounced and dpjudge. I have not seen any test report of
> > their adjudicator.... phpdiplo for instance is not cited since I
> > understood phpdiplo does not implement the full diplomacy rules. And
> > there are probably more automated website around.
>
> DP Judge, like nJudge, requires convoy routes to be specified. It's a
> fairly serious flaw in the shared ancestry of both programs that I think
> should be addressed in nJudge 2.0 if such a creature is ever born. DP
> Judge is even less likely than nJudge to get such a serious update to
> the core code.
>
> > My personnal comment : people tend to be more interested in the
> > quality of the maps then the adjudication....
>
> That is until there is a problem with the adjudication. The also care
> little about security... until they think someone is cheating. You can
> always fix a map. You can fix most security problems, although you may
> have to flush the user database. When an adjudicator is broken, however,
> it's broken for the life of the program. At least that's what it seems
> like. :-/
>
> Anyhow, look at jDip if you want to see some good maps and RealPolitik
> if you want some maps to compare to good maps.
>
> Chris
 >> Stay informed about: Play Diplomacy Online (beta test) 

As writer of the DATC I can say, that is not entirely true.
DipTool uses a different algorithm and also passes all
test cases.

I am great supporter of adjudicators that has a generic
solving mechanism, rather than specific such as the
DPTG and other older adjudicators. Those specific
algorithms are very difficult to get bug free. Currently,
no such adjudicator exists that is flawless (the DPTG has
3 hard adjudicator bugs).

However, the DATC algorithm, is not the only possibility.

The problem is in situations on the board, that result in
a cyclic dependency of orders to be resolved. Those
situations can be divided in two groups:

- A cyclic dependency, with 0 or more than 1 resolution.
- A cyclic dependency, but still with one resolution.

The first group are circular movements or convoy paradoxes.
They must be resolved by a special rule, the circular movement
rule or a convoy paradox solving rule.

However, if we would only have the first group, then it would
not be very difficult for the adjudicator. We just write a recursive
procedure, that applies the circular movement rule or a
convoy paradox solving rule, when it detects a lock up.

The problem is the second group. Those situation are comparable
with the first group, but with some additional orders, which
result that there is only one resultion. Examples of that are
a circular movement with one of the units with a support or
one of the units bouncing on an other unit. At that moment there
is only one valid adjudication for the circular movement (before
considering the circular movement rule). Similar with a convoy
paradox, additional orders can interfere with the paradox such that
there is only one resolution.

Still, the second group, consists of cyclic dependent orders.
The tackle this, the DATC algorithm can solve those situation
by making decisions on partial information.

For instance, in case of a circular movement. If a unit in
the movement gets a support or bounces, then the DATC
algorithm can solve this order, without first solving the order
of the unit on the destination area.

The ability to resolve on partial information, is sufficient of
resolving all situations of the second group under the
standard rules (it might not be true with extension rules).

However, an alternative for the second group of orders
is to 'try' a certain outcome. This is how DipTool works.
DipTool builds up a graph of the orders. The orders for
which all dependencies are solved (this includes orders
without dependency at all), can be resolved directly.
For a cyclic dependency, it tries a certain outcome and
checks whether the outcome is consistent. If so, it also
tries all other possibilities. If there is only one consistent
outcome, it will take that one, if not, the circular movement
rule or paradox rule is applied.

If only single circular dependencies can occur (what is
the case for the standard rules, but might not be the
case for extension rules), then it can even be programmed
in a recursive function (no graph logic necessary).
This simplifies the DipTool algorithm.

The recursive funtion will first do a 'guess' of the order
that must be resolved. Then it resolves first all orders on
which it depends, by calling itself recursively. The recursive
functions, also returns whether the result was dependent
(directly or indirectly) on a guess. If all the dependent
orders, where not dependent on a guess, then its own guess
is replaced by the real outcome. If it was dependent on
a guess. some additional logic is needed (left as exercise
for the reader).

This result in algorithm that is probably a little bit shorter
(and faster) than the algorithm in the DATC, because
the specific Diplomancy rules, only need to be programmed
once.

Regards,

Lucas

On Dec 31 2007, 11:17 am, "jeremie.lefranc...@gmail.com"
<jeremie.lefranc....TakeThisOut@gmail.com> wrote:
> No, my engine only runs on stabbeurfou.
>
> It did not take 10 years to build it but I have worked on it by
> episodes, starting in 1992, 15 years ago then.
> The only way to have a flawless adjudicator I think is to use the datc
> algorithm (from my experience).
>
> The strength of the stabbeurfou adjudicator stems in the error and
> warning messages when orders are submitted - I think.
>
> On Dec 30, 5:49 pm, Chris Babcock <cbabc....TakeThisOut@cox.net> wrote:
>
>
>
> > On Sat, 29 Dec 2007 09:39:50 -0800 (PST)
>
> > "jeremie.lefranc...@gmail.com" <jeremie.lefranc....TakeThisOut@gmail.com> wrote:
> > > Entering a position and testing adjudicator is available in the
> > >www.stabbeurfou.orgsite. Use the gadgets.
> > > Direct link below :
> > >http://www.stabbeurfou.org/Interrogation.php
>
> > > And also in the diplomatie-online french website, and not in the
> > > 18centres french website (or not to my knowledge).
>
> > > Thehttp://www.stabbeurfou.org/hasadjudication engine been fully
> > > tested withe the DATC test cases (amongst others).
> > > Compliant to Hasbro 2000 rules, to former French Descartes 1994 rules.
> > > See within rules page.
>
> > Jérémie, is it your engine running behind playdiplomacy.com? If so, I'm
> > somewhat relieved that there hasn't been an enormous duplication of
> > effort. I know you spent about 10 years working on yours.
>
> > > Test reports are of course available on site, see below :
> > >http://www.stabbeurfou.org/docs/regles/en/Test_report.htm
>
> > > Deviations from DATC are only due to different interpretation of the
> > > rules (land route first for convoy to adjacent place for instance).
>
> > > There are to my knowledge at least two more serious automated
> > > websites : dipbounced and dpjudge. I have not seen any test report of
> > > their adjudicator.... phpdiplo for instance is not cited since I
> > > understood phpdiplo does not implement the full diplomacy rules. And
> > > there are probably more automated website around.
>
> > DP Judge, like nJudge, requires convoy routes to be specified. It's a
> > fairly serious flaw in the shared ancestry of both programs that I think
> > should be addressed in nJudge 2.0 if such a creature is ever born. DP
> > Judge is even less likely than nJudge to get such a serious update to
> > the core code.
>
> > > My personnal comment : people tend to be more interested in the
> > > quality of the maps then the adjudication....
>
> > That is until there is a problem with the adjudication. The also care
> > little about security... until they think someone is cheating. You can
> > always fix a map. You can fix most security problems, although you may
> > have to flush the user database. When an adjudicator is broken, however,
> > it's broken for the life of the program. At least that's what it seems
> > like. :-/
>
> > Anyhow, look at jDip if you want to see some good maps and RealPolitik
> > if you want some maps to compare to good maps.
>
> > Chris- Hide quoted text -
>
> - Show quoted text -
 >> Stay informed about: Play Diplomacy Online (beta test) 

On Wed, 2 Jan 2008 05:21:42 -0800 (PST)
L.B.Kruijswijk.RemoveThis@inter.nl.net wrote:

> As writer of the DATC I can say, that is not entirely true.
> DipTool uses a different algorithm and also passes all
> test cases.
>
> I am great supporter of adjudicators that has a generic
> solving mechanism, rather than specific such as the
> DPTG and other older adjudicators. Those specific
> algorithms are very difficult to get bug free. Currently,
> no such adjudicator exists that is flawless (the DPTG has
> 3 hard adjudicator bugs).
>
> However, the DATC algorithm, is not the only possibility.

Actually, we were talking about the DATC test cases rather than the
adjudication algorithm. However an adjudicator is written, it needs to
be tested systematically. Testing against the DATC cases seem to be a
suitable alternative to testing all possible positions. As far as I
know, it is the only tool of its kind in this regard.

>% Snip of good description of adjudicator issues %<

In most adjudication situations, speed is not a serious issue. A helper
program like DipTool or a server like we've been discussing needs to be
reliable and predictable - reliable in the sense that it won't crash or
get locked in infinite loops and predictable in that it produces the
expected outcomes. Diplomacy judges do have the tendency to stack
move adjudications for the same time of day and tools users expect
local calculations quicker than a website load, but the algorithms
employed so far have been adequate for these tasks.

The main application that calls for faster adjudication is Bot writing.
Albert is the top bot right now and probably the only 'smart' bot, in
that it actually attempts to calculate possible outcomes, as opposed to
simply relying on a system of relative weighting. Default settings are
for 100 adjudications per move, but users often change that to 50
because waiting the full 5 minutes for the move deadline to expire is
"too slow".

I like the interest shown in developing alternatives to the existing
base of servers. This shows an unexpected depth to the community. A
number of these projects are released under a GPL or similar license. I
hope that the programmers begin to leverage the potential of open
source by setting up source repositories and encouraging others to make
contributions to their code. Sourceforge and Freshmeat are both good
general purpose source hosting options. If anyone would like a narrower
audience for their development proposals, I'll have a CVS source
repository running around June.

Chris

PS - I nominate this thread for best thread jack of the year for 2007.
Please refer to the original post for a very promising Diplomacy game
server and contact me via email if you need any 2 center austria
replacements.
 >> Stay informed about: Play Diplomacy Online (beta test)