Actually, the Simplified Alliance Pointers,

E:9, F:9, G:12, I:12, A:12, R:12, T:9

should read as :

E:3, F:3, G:5, I:4, A:5, R:5, T:3 (num, of 'current effective neighbours').


3*3*5*4*5*3 = 5940 (a rather tiny num. of 'vectors')

using a Real variable from (-1 .. +1) as Trust_Indicator

(ex.: +0.91, 0.31, -0.99, -0.075, ...)


Event_Field_of_Trust-Factor_per_Number_of_Other_Players
using a Real variable from (-1 .. +1) as Trust_Indicator

6*6*6*6*6*6*6 = 279936 (2,8x10e5)
(less than a third of a million, rather small)


--> If it runs through only optimized solution surfaces
that are elements of 279936, a few times (3-9, let's say)
during the 'diplomatic phase', it should clock very quick
computer time, when considering that searching for best
Move_Plans according to Objectives, and 5940 pointers...


---


Mini_Pseudo_Code_For_DIP_AI;

0 = The_Game_Master_Bot_AI

Three (headed) modules of BOT =

1 - Bot_Player_Personality_AI . (The Head) User GUI & Stats & Web, etc...
2 - Bot_Strategic_Planning_AI . (right-brain) Strategy_Bot
3 - Bot_Diplomatic_Attaché_AI . (left-brain) Diplomacy_Chat_Bot


The_Game_Master_Bot_AI makes new Map_Postions_And_Config

At Start, each Country gets Positions, and
chooses (2 or 3) Move_Plans, and Sets_Objectives ,

Diplomacy_Chat_Bot issues a Greetings depending on Positions

1.) Countries submit them to their Diplomacy_Chat_Bot to 'negotiate' & evaluate :

Event_Field_of_Trust-Factor_per_Number_of_Other_Players ,

Then submits Field to Strategy_Bot; with Plans and Objectives ,

Strategy_Bot sends modifications to 1.)
along with Flags_in_Objectives ,

UNTIL ( Plans_Are_Good_To_Submit=TRUE OR TIME_Elapsed ) ,

Game_Master_Bot does moves ,

if (Any_Retreats = True) THEN Countries do Retreats

if (Fall_Moves_Done = True) THEN Countries do Adjustments

if (Game_Over = False) THEN Go Back to 1.)



Actually, the Simplified Alliance Pointers,

E:9, F:9, G:12, I:12, A:12, R:12, T:9

should read as :

E:3, F:3, G:5, I:4, A:5, R:5, T:3 (num, of 'current effective neighbours').


3*3*5*4*5*3 = 5940 (a rather tiny num. of 'vectors')

using a Real variable from (-1 .. +1) as Trust_Indicator

(ex.: +0.91, 0.31, -0.99, -0.075, ...)


Event_Field_of_Trust-Factor_per_Number_of_Other_Players
using a Real variable from (-1 .. +1) as Trust_Indicator

6*6*6*6*6*6*6 = 279936 (2,8x10e5)
(less than a third of a million, rather small)


--> If it runs through only optimized solution surfaces
that are elements of 279936, a few times (3-9, let's say)
during the 'diplomatic phase', it should clock very quick
computer time, when considering that searching for best
Move_Plans according to Objectives, and 5940 pointers...


---


Mini_Pseudo_Code_For_DIP_AI;

0 = The_Game_Master_Bot_AI

Three (headed) modules of BOT =

1 - Bot_Player_Personality_AI . (The Head) User GUI & Stats & Web, etc...
2 - Bot_Strategic_Planning_AI . (right-brain) Strategy_Bot
3 - Bot_Diplomatic_Attaché_AI . (left-brain) Diplomacy_Chat_Bot


The_Game_Master_Bot_AI makes new Map_Postions_And_Config

At Start, each Country gets Positions, and
chooses (2 or 3) Move_Plans, and Sets_Objectives ,

Diplomacy_Chat_Bot issues a Greetings depending on Positions

1.) Countries submit them to their Diplomacy_Chat_Bot to 'negotiate' & evaluate :

Event_Field_of_Trust-Factor_per_Number_of_Other_Players ,

Then submits Field to Strategy_Bot; with Plans and Objectives ,

Strategy_Bot sends modifications to 1.)
along with Flags_in_Objectives ,

UNTIL ( Plans_Are_Good_To_Submit=TRUE OR TIME_Elapsed ) ,

Game_Master_Bot does moves ,

if (Any_Retreats = True) THEN Countries do Retreats

if (Fall_Moves_Done = True) THEN Countries do Adjustments

if (Game_Over = False) THEN Go Back to 1.)



Actually, the Simplified Alliance Pointers,

E:9, F:9, G:12, I:12, A:12, R:12, T:9

should read as :

E:3, F:3, G:5, I:4, A:5, R:5, T:3 (num, of 'current effective neighbours').


3*3*5*4*5*3 = 5940 (a rather tiny num. of 'vectors')

using a Real variable from (-1 .. +1) as Trust_Indicator

(ex.: +0.91, 0.31, -0.99, -0.075, ...)


Event_Field_of_Trust-Factor_per_Number_of_Other_Players
using a Real variable from (-1 .. +1) as Trust_Indicator

6*6*6*6*6*6*6 = 279936 (2,8x10e5)
(less than a third of a million, rather small)


--> If it runs through only optimized solution surfaces
that are elements of 279936, a few times (3-9, let's say)
during the 'diplomatic phase', it should clock very quick
computer time, when considering that searching for best
Move_Plans according to Objectives, and 5940 pointers...


---


Mini_Pseudo_Code_For_DIP_AI;

0 = The_Game_Master_Bot_AI

Three (headed) modules of BOT =

1 - Bot_Player_Personality_AI . (The Head) User GUI & Stats & Web, etc...
2 - Bot_Strategic_Planning_AI . (right-brain) Strategy_Bot
3 - Bot_Diplomatic_Attaché_AI . (left-brain) Diplomacy_Chat_Bot


The_Game_Master_Bot_AI makes new Map_Postions_And_Config

At Start, each Country gets Positions, and
chooses (2 or 3) Move_Plans, and Sets_Objectives ,

Diplomacy_Chat_Bot issues a Greetings depending on Positions

1.) Countries submit them to their Diplomacy_Chat_Bot to 'negotiate' & evaluate :

Event_Field_of_Trust-Factor_per_Number_of_Other_Players ,

Then submits Field to Strategy_Bot; with Plans and Objectives ,

Strategy_Bot sends modifications to 1.)
along with Flags_in_Objectives ,

UNTIL ( Plans_Are_Good_To_Submit=TRUE OR TIME_Elapsed ) ,

Game_Master_Bot does moves ,

if (Any_Retreats = True) THEN Countries do Retreats

if (Fall_Moves_Done = True) THEN Countries do Adjustments

if (Game_Over = False) THEN Go Back to 1.)


---

This Thread is :

http://webdiplomacy.net/forum.php?viewthread=462962

Note: The number of 'current effective neighbours' (5940) varies during a game,
tending to get smaller.

LOL... yeah, flesh out some pseudo-code for that "Strategy_Bot"... you remind me of one of the business clients we deal with regularly. He's famous for saying, "All you have to do is add a few IF's, right?" We stopped answering years ago.

I'm aware this is a lot of Work!... Not saying I'll really-really do, it now toying with ideas,
& reading up some stuff... But the 'numbers' in this train of thought are 'within grasp'...

or was that a long Case statement ?

Use the Map with complete history of moves as 'arrows'...

doable in C, but how about LISP ?

Dude... This is.. crazy! >.<

I'm not saying this is a bad idea! Because I like it! But.. The idea of having this in the game is great, really.. But the idea of making this... ;O

Not sure it's everyone's cup of tea, but
'Bots (of quality) in the Real-World, and Diplomacy,
are of course inevitable !...

"Inevitable" perhaps in the most literal, long-term sense, but you're still dramatically underestimating the processing speed and storage resources that it will require. Your numbers also ignore the enormous complexity of the possible relationships between support and movement orders between different cooperating orders, and things like "intentionally failed" movements intended to cut support, and why those things happen, and what should be done to counter it. Hence my laughter at the hand-waving one-line note, "do grand strategy here". It isn't enough to just say, "What can I do?" because to play effectively, you must also make some educated guesses about what everyone else is going to do.

Look how much time and effort was required before Kruiswijk finally pulled together all the pieces necessary to produce a "perfect" adjudication algorithm: 10 years, by his own estimation. And that's just processing the rules correctly. The bot would need to incorporate this processing into basically every decision. Go write an adjudicator first. You'll be shocked at the extent of the interdependencies of even a relatively simple series of real-world orders.

Basically, since the rules are completely deterministic, Diplomacy theoretically boils down to a linear system (a collection of simultaneous linear equations, and worse, equations with linear dependencies). Look up Choelsky sparse matrix transposition to get an idea of the level of mathematics that would probably be required to actually evaluate a reasonable set of potential moves in some useful period of time. But that would be a Deep Blue level of automation, and I admit neither of us were really talking about solutions at that level (although it would be interesting -- and I rather suspect it would still be a failure against truly skilled Dip players).

The existing bots, of course, take all kinds of shortcuts and make a bunch of assumptions. That's perfectly reasonable, but it's also absolutely necessary if you want to finish a game within the span of a natural human lifetime. The bots concern themselves mainly with what supply centers can be reached and who is immediately adjacent to them, and they don't spend any time looking at regions they don't touch at all. They're easily boxed in and overwhelmed. You can plan several moves ahead. They can't. And because several moves involves at least tens of millions of possibilities using your numbers, or octillions of possibilities using the numbers of the folks who have studied the problem in depth, it seems unlikely that any bot will do much planning ahead any time in the near future.

Heck, chess bots don't plan ahead -- not even the Deep Blue variety -- and chess is child's play, in computing terms.

It's an intriguing problem, yes, but the hard problem isn't how to pseudo-code the high level process. The hard problem is behind those one-liners you're throwing out as if they're just an implementation detail. I've begun to suspect that players with a lot of face to face experience may have a better feel for the complexities of implementing Diplomacy. Online play is practically one step shy of gunboat, as far as I'm concerned, and I think it makes the problem "feel" a lot less complex than it is.

It's funny you'd mention LISP... I recently experimented with a Prolog backchain approach to adjudication, but my Prolog is rusty (Borland TurboProlog was the last time I used it seriously, like 1991 maybe) and it didn't go anywhere. For awhile I thought that might be a good way to write a player bot, but Prolog's goal-seeking methods really just magnify the problem of storage and processing time. Kruijswijk's recursive approach of unwinding the dependencies is about as elegant as it gets for adjudication, and I now believe the problem of how to actually play is a totally separate problem.

And I was hoping it would take 2-3 years for a 'great product'! ?Quìen Sabé?

Of course all this is not meant as only mere tossing ideas in the air...

Some 1-liners are great train_of_thought-catalysers, for myself at least!


"The existing bots, of course, take all kinds of shortcuts
and make a bunch of assumptions..."

--> EXACTLY ! & we CAN find even more cutting edge ways, I'm sure !


"The hard problem is behind those implementation details."

Simple game-paradigm specific breakdown of strategy dynamics.
The question is not thinking "in computing terms",
but in what we are looking at, the Game of Diplomacy,
how can we program efficient and specifyable 'bots',
in this specific 'direct (symbolic) application terms'.


Cholesky matrix transposition is nice, but it's a huge matrix!
Sure folding it in two is making a smaller form, but mind you
that we are talking about a very Sparse matrix...
It costs (1/3)x(N exp.3) FLOPs (that is, NxNxNx(1/3), where N is size of the matrix A
(And looks even longer when read in the Numerical Recipies book!)

For decent time limits on our current PCs,
the whole process has to be less than 1 TeraFLOPS (10 exp.12)
i.e.: less than 16.6 minutes...


There are a few steps for Strategy;

Discard 'bad' moves first! --> cuts away lots...

Cholesky matrix transposing ,
Or Perato Principle Optimization ,
Or MOSA: Multiple-Objective Simulated Annealing method ... ?

--> Swarm 'Vectoring of Units'

That's why I noticed the Swarm Theories as probably an
ideal approach to this Game, and they do use a MAP,
POSITIONS, etc.... as their basic mainstay cornerstone
concepts... and hugely simply representations and
information processing in a few basic lumps ... ;))

The Diplomat_Bot clarifies objectives per neighbour (one can
only move near them) , yes, also considering longer-distances
and in a time-frame of a recurring series of ;
Move-Retreats-Move-Retreats-Adjust-...

And, here's the Cruncher: the MAP OF DIPLOMACY has 75 regions.
'Everything that happens on it' is represented by '34' units (or less)
over time-steps that are 'moving on' the said MAP in vector-chains towards
and through neighbouring regions, according to the Diplomat_Bot's
Objectives_Direction trimmed towards some basic rules of 'better playing'
OR what the User specified as 'tweeks' on the following possible
Bot_Type_Variables = ( aggressivity, cooperation, slyness-stabiness, etc...? )

( with their move-type-Flags, and all other variables accesible
'as if' in same matrix but are relegated to secondary access ).


(7 basic) Move_Types = (Build, Move, Convoy, Support Move, Support Hold, Retreat, Disband

I've to delve into topic to get a good feel of it.

http://en.wikipedia.org/wiki/Particle_swarm_optimization

...
--> Very High-Press level
players with a lot of face to face may have better feel for
complexities of implementing Diplomacy.
(The addition mainly is tha you know who is talking to whom,
how long, certain reactions, etc..)

--> Low-medium-Press level
Online play is practically one step shy of gunboat...
it makes the problem "feel" a lot less complex than it is.
... Just like working side-by-side ...


!... dramatically underestimating the processing speed !

No, not at all... that's why we need a more efficient approach!
! Pushing 3 units forward can't be that hard ! ;)

" You can plan several moves ahead. They can't. ...
it seems unlikely that any bot will do much
planning ahead any time in the near future. "


Some points raised pertain to the Diplomat_Bot that regulates Objectives:

different cooperating orders
"intentionally failed" movements
(intended to cut support, or position for stab, or, ...)
and why those things happen,
and what should be done to counter it.
you must also make some educated guesses...


The Math of Adjudication by Lucas Kruijswijk :
http://www.diplom.org/Zine/S2009M/Kruijswijk/DipMath_Chp1.htm


I do have a preference for Lisp; it's more elegant, used,
it's older ( the 2nd. Computer Language ever created! ),
more options, & probably Has a Future!, it's also
a language of choice for processing 'language'...


In the Freebies : http://www.ufasoft.com/lisp/

Schoolbook ex.: "Practicle Common Lisp" : http://www.gigamonkeys.com/book/

! For a Video on Lisp: ! (Almost a sleeper but some Great Lines...)

http://video.google.com/videoplay?docid=448441135356213813

The Adjucator is a functional skeleton upon which one may add on 'modules' ...

The 'LISP' reference these days, I guess :

http://www.franz.com/products/allegrocl/

Here's a way to group provinces per area, so as to weight their relative importances with another structure. One could argue that Bud, Vie, Boh are the real Center, and that
Syria is quite secondary as a corner, etc...

geometric placing:


4 - Corners = ( NAO, StP, Syr, MAO )

23 - Outer_Circle = ( NWS, BAR, Nor, Fin, GBO, Liv, Mos, Sev, Arm, EME, ION, Tun, NAf, WME, Spa, Por, Gas, Bre, ENG, IRI, Lvp, Cly, Edi )

29 - Ring = ( NTH, Lon, Wal, Yor, Den, SKA, SWE, BAL, Pru, War, Ukr, Rum, BLA, Ank, Con, AEG, Gre, Alb, ADR, Apu, Nap, TYS, GOL, Mar, Bur, Par, Pic, Bel, Hol )

18 - Middle = ( Bud, Vie, Boh, Mun, Kie, Ber, Sil, Gal, Rum, Bul, Ser, Tri, Ser, Ven, Rom, Tus, Tyr, Pie )

Anyone thought of this Province in Circles Mapping ?

Would more than a handfull of people like to see Bots. in Diplo ?...

Thank You for Your Input :

- NinjaPirateRobotZombie (NPRZ)
- figlesquidge
- TheMasterGamer
- SpeakerToAliens
- Toby Bartels
- OMGNSO
- Chrispminis
- Nick Douglas
- Katsarephat
- Sayjo the Lion-Hearted
- Star Revil


!!!

ANSI Common Lisp Programming Language :

http://www.franz.com/products/allegrocl/

(There's the (unlimited-time) trial version, the " Express ".) !!!

Thank You for Your Input :

- NinjaPirateRobotZombie (NPRZ)
- figlesquidge
- TheMasterGamer
- SpeakerToAliens
- Toby Bartels
- OMGNSO
- Chrispminis
- Nick Douglas
- Katsarephat
- Sayjo the Lion-Hearted
- Star Revil
- IcyMind
...

...

"... humans also rely on rules of thumb that limit our thinking,

which is why I don't see why a bot could never be as good as a human..."

Do You Believe This ? (Y/N)

My current off-in-summer REHASH Of Previous posts on:

"...number of possible non-equivalent openings: 4,430,690,040,914,420"
"...don't think that it is computationally impossible to write a half decent Diplomacy bot.


==>> 'Main Bot & GUI' ... Bot_Player_Personality_AI

( The Bot Arena =
"Albert",
"Random Bot",
"Hold Bot",
"The Paladin Bot",
"Berzerker Bot",
"The Moving Wall",
"Whirlwind",
"Arrow", ... )



0 = Game_Master_Bot_AI_&_Adjucator

A Bot is a Three (headed) modules of sub_Bots =
1 - Bot_Player_Personality_AI . (The Head) User GUI & Stats & Judge & Web, etc...
2 - Bot_Strategic_Planning_AI . (right-brain) Strategy_Bot
3 - Bot_Diplomatic_Attaché_AI . (left-brain) Diplomacy_Chat_Bot

(The Head) User GUI & Stats & Judge & Web, etc.... . . . . . . . (! obviously REWRITE THIS !)

---
-> Mini_Pseudo_Code_Bot; <-
The_Game_Master_Bot_AI makes new Map_Postions_And_Config
At Start, each Country gets Positions, and chooses (2or3)Move_Plans, & Sets_Objectives,
Diplomacy_Chat_Bot issues a Greetings depending on Positions.
1.) Countries submit them to their Diplomacy_Chat_Bot to 'negotiate' & evaluate:
Event_Field_of_Trust-Factor_per_Number_of_Other_Players,
Then submits Field to Strategy_Bot; with Plans and Objectives,
Strategy_Bot sends modifications to 1.) along with Flags_in_Objectives,
UNTIL ( Plans_Are_Good_To_Submit=TRUE OR TIME_Elapsed ),
Game_Master_Bot does Case:
(Any_Retreats = True) THEN Countries do Retreats
(Fall_Moves_Done = True) THEN Countries do Adjustments
(Game_Over = False) THEN GoBack to 1.)
Do Meta_gaming_stuff and out_of_Game_stats_prompts_mails, etc...
---

AI Bots Run and make Decisions given on input from Strategy&Diplo._Bots

essential goal is eighteen supply centers ( A_Player_Won = TRUE if Points_Player(n) = 18 ),
secondary goals = ( don't get eliminated, stop emerging winner(s), obtain best draws ) ...
- ( each w/ diff. weights ex.( 1.0, 0.8, 0.67, ...) This gives 'character' to Bots.
- "That means that people could learn to get a feel for how the AI thinks."
--> we'd add another 'random scalar' to the 'final moves', how's that!

implement tactical event map of game.

Decide and focus on primary goals and objectives at Opening

Responses to Countries going CD, or surprises ...

(-Prepare mid-game tactics & ploys for ongoing meta'_Strategies-)
- The pleasure is creative
- working in alliance(s) side-by-side ...
- steal someone's ally away from them
- war of attrition
- teach your bots emotions, lying, and
- the "stab"
- different cooperating orders
- "intentionally failed" movements
- (intended to cut support, or position for stab, or, ...)
- and why those things happen
- what should be done to counter it.

A moves_per_country_Alliances schema:
Per Total Moves per Main Openings: (1.50..x10E10)
E:9, F:60, G:112, I:12, A:12, R:288, T:9 *= 15049359360 .
_per_Neighbour_Alliance: (1.51..x10E7)
E:9, F:9, G:12, I:12, A:12, R:12, T:9 . * = 15116544

Simplified One_percentile: E:3, F:3, G:4, I:3, A:4, R:4, T:4 *= 5184 (Only 5 Thousand...)
-> Every Country has 4 lumped alliances *= 16384 (4e7) ... (3 too little, 5-6 redundant) (realistic!)
(best, ex.: 3 main + 1(3-in-One) = 4 (2 ennemies, 1 ally, others lump into an alliance-group)


==>> Openings;

ENGLISH OPENINGS--(3*3=9) : (4=3+1(of4)) : ( 6 )
English openings-- categorized by orders of fleets:
Main: (3)
1)(a) Northern Opening: ( F Lon-NTH, F Edi-NWG, a-lvp-Yor ), Pro-French, (0.70),
1)(b) Northern Opening: ( F Lon-Nth, F Edi-Nwg, a-lvp-Edi ), Pro-French, (0.90), Anti-Russian, (0.95),
2) Southern Opening: ( F Lon-Eng, F Edi-Nth, a-Lvp-Wal ), Anti-French, (0.95),

Secondary: (4)
3) Western Opening: ( F Lon-Eng, F Edi-Cly, a-Lvp-Wal ), Anti-French, (0.99)
4)(a) The Splits Opening: ( F Lon-Eng, F Edi-Nwg, a-Lvp-Wal ), Anti-French, (0.75), Pro-German, (0.95)
4)(b) The Splits Opening: ( F Lon-Eng, F Edi-Nwg, a-Lvp-Edi ), Anti-French, (0.55), Pro-German, (0.95), Anti-Russian, (0.90)

-

AUSTRIAN OPENINGS--(3*4=12) : (2=1+1) : ( 31 )
categorized in following groups: (2)
Primary:
Balkan Gambit, (0.81), A Bud-Ser, F Tri-Alb, a-Vie-(Gal, Boh, Tyr, Bud, Tri), 5.

Secondary Openings, (19%),
Roadhog Openings/Rumanian Gambit, ( F Tri H, A Bud-Rum, a-Vie-(Gal, Boh, Bud) ), 3.
Houseboat Openings, ( F Tri H., A Bud-Ser, a-Vie-(Tri, Gal, Boh, Tyr, Bud) ), 5.
Hedgehog Openings, ( F Tri-Ven, A Bud-Ser, a-Vie-(Gal, Boh, Tyr, Bud) ), 4.
Blue Water Openings, ( F Tri-ADR, A Bud-(Ser, Tri), a-Vie-(Gal, Boh, Bud) ), 6.
Tyrolese Gambit Opening, ( A Bud-Tri, A Vie-Tyr, F Tri-(Ven, Alb, ADR) ), 3.
Galician Gambit Openings, ( A Bud-Gal, a-Vie-(Tri, Gal, Boh, Tyr, Bud) ), 5.

-

FRENCH OPENINGS--(3*3=9) : (25=12+12+1),in 3 choices from: (2 Primary or 3 secondary)) : (60)
The French openings are categorized by order given to Brest fleet:

Primary: (2)
Manche (or English Attack) Openings (12) (F Bre-ENG) A-Par-(Bur OR Gas OR Pic OR Bre), A-Mar-(Spa, Bur, Pie, (S)Par-Bur, Gas)
Atlantic Openings, (12), ( F Bre-MAO, A-Par-(Bur OR Gas OR Pic OR Bre), A-Mar-(Spa OR Pie Or (S)Par-Bur Or Gas)

Secondary: (3)
Biscay Openings (F Bre-Gas) (12) (F Bre-ENG) A-Par-(-Bur OR Gas OR Pic OR Bre), A-Mar-(-Spa OR Pie Or (S)Par-Bur Or Gas)
Norman Openings (F Bre-Pic) (12) (F Bre-ENG) A-Par-(-Bur OR Gas OR Pic OR Bre), A-Mar-(-Spa OR Pie Or (S)Par-Bur Or Gas)
Houseboat Openings (F Bre-H) (12) (F Bre-ENG) A-Par-(-Bur OR Gas OR Pic OR Bre), A-Mar-(-Spa OR Pie Or (S)Par-Bur Or Gas)

-

GERMAN OPENINGS--(3*4=12) : (4=3+(1of7)) : ( 112 )
The German openings are categorized as follows:
Primary: (3)
Blitzkrieg Openings ( A Ber-Kie, A Mun-Ruh, F Kie-(Den OR Hol OR HEL OR BAL) ), 4.
Anschluss Openings ( A Ber-Kie, F Kie-Den, A Mun-(Ruh, Bur, Tyr, Sil, Boh, Ber, Hold) ) 7.
Dutch Openings ( A Ber-Kie, F Kie-Hol, A Mun-(Ruh, Bur, Tyr, Sil, Ber, Boh, Hold) ) 7.

Secondary: based on order of Berlin army. (7)
Gambit Openings ( A Ber-Mun, F Kie-Hol, A Mun-(Ruh, Bur, Tyr, Sil, Kie, Boh) ), 5.
Rhineland Openings ( A Ber-Mun, F Kie-Den, A Mun-(Ruh, Bur, Tyr, Sil, Kie, Boh) ), 5.
Prussian Openings ( A Ber-Pru, F Kie-(Den, Hol, BAL, HEL), A Mun-(Ruh, Bur, Tyr, Sil, Kie, Boh, Hold) ), 28.
Silesian Openings ( A Ber-Sil, F Kie-(Den, Hol, BAL, HEL), A Mun-(Ruh, Bur, Tyr, Sil, Kie, Boh, Hold) ), 28.
Berlin Defense ( A Ber H., F Kie-(BAL Or HEL), A Mun-(Ruh, Bur, Tyr, Sil, Boh, Kie, Hold) ) 14. )
Dutch Boy Openings ( A Ber-Kie, F Kie-Hel, A Mun-(Ruh, Bur, Tyr, Sil, Ber, Boh, Hold) ) 7.
Western Barbarossa ( A Ber-Kie, F Kie-BAL, A Mun-(Ruh, Bur, Tyr, Sil, Ber, Boh, Hold) ) 7.

-

ITALIAN OPENINGS--(3*4=12) : ( 12 )
Italian openings are categorized as follows:

Lepanto Openings (A Rom-Apu/Nap, F Nap-ION) a-Ven(-pie OR Tri OR Tyr) )
The Western Lepanto (A Rom-Nap/Tus/H., F Nap-TYS) a-ven(Pie OR Tri OR Tyr) )
Austrian Attack Opening (A Rom-Ven, A Ven-Tri) F Nap-(TYS OR ION) )
Obriani Openings (A Rom-Ven, A Ven-Tyr, F Nap-(TYS OR ION) )
Alpine Chicken Openings (A Rom-Ven, A Ven-Pie) F Nap-(TYS OR ION) )

-

RUSSIAN OPENINGS--(3*4=12) : ( 288 )
Russia has more possible opening moves than any other power.
A Mos 6 , A War 6 , F Sev 4 , F StP 2

F StP/sc H openings are known as the Northern Houseboat Opening.
F StP/sc-Lvn is known as the Kronstadt Opening.
F StP/sc-Fin is known as the Lapland Lurch.

These openings are named after the order of A Moscow.
There are therefore six systems for openings involving F StP/sc-BOT or F StP/sc-Fin:
The Inertia System (A Mos H.),
the Livonia System (A Mos-Lvn),
the Northern System (A Mos-StP),
the Southern System (A Mos-Sev),
the Ukraine System (A Mos-Ukr), and
the Warsaw System (A Mos-War).

-

TURKISH OPENINGS--(3*2=6) : (6)

Bulgarian Opening ( A-Con-Bul, (0.998), F-Ank-Bla, (0.712), OR Arm, (0.282) )
A ( Smy-(Con, (0.492), Arm, (0.433), Hold, (0.01) )

-

==>> !! MAP !!

group provinces per area,
weight their relative importances with other structures
(? Bud, Vie, Boh are real Center, Syria quite secondary as a corner, etc...)
Use the Map with complete history of moves as tagged-vectors

A geometric placing:
4 - Corners = ( NAO, StP, Syr, MAO )
23 - Outer_Circle = ( NWS, BAR, Nor, Fin, GBO, Liv, Mos, Sev, Arm, EME, ION, Tun, NAf, WME, Spa, Por, Gas, Bre, ENG, IRI, Lvp, Cly, Edi )
29 - Ring = ( NTH, Lon, Wal, Yor, Den, SKA, SWE, BAL, Pru, War, Ukr, Rum, BLA, Ank, Con, AEG, Gre, Alb, ADR, Apu, Nap, TYS, GOL, Mar, Bur, Par, Pic, Bel, Hol )
18 - Middle = ( Bud, Vie, Boh, Mun, Kie, Ber, Sil, Gal, Rum, Bul, Ser, Tri, Ser, Ven, Rom, Tus, Tyr, Pie )

Create provinces_per_rings_around_each_province lists

( weights, distance, importance flags, "projected plans" tag, add diplomacy & communication knowledge, ... )



==>> 'Strategy Bot'

... unit locations, state of alliances -> only few good moves

Evaluates Current Map

establish most probable future moves_map

evaluates moves according to states of 4 'main' alliances (percentile) as supplied by Diplo_Bot

sort centers in order of priority

Discard 'bad' moves first! --> cuts away lots...

and also trimmed towards some basic rules of 'better playing'

OR what the User specified as 'tweeks'

(7-basic)_Move_Types = (Build, Move, Convoy, Support Move, Support Hold, Retreat, Disband)


??? finding tradeoff surfaces:

Cholesky matrix transposing ,
Or Perato Principle Optimization ,
Or MOSA: Multiple-Objective Simulated Annealing method ... ?
--> Swarm 'Vectoring of Units', consider per centers & groupings

keep track of potential force concentrations on each center

neutrals that are very close could then be marked

(weights, distance, importance flags, "projected plans" tag, add diplomacy & communication knowledge,...)

There could be the Genetic algorithms that the Strategic_Planning_AI
could be use; long 'genes' to make logical ops.& comparisons of choices...



==>> 'Diplomat Bot'

Use of Language for Press and Information Communication

Find intentions of the players

difficulty is in receiving, translating, understanding press & replying with any sort of coherence.

"The pleasure is creative. _Bot_ won't give me pleasure 'til it passes Turing Test."

The Diplomat_Bot clarifies objectives per neighbour (one can only move near them)

considering longer-distances and events

time-frame of a recurring series of Move-Retreats-Move-Retreats-Adjust-...

Event_Field_of_Trust-Factor_per_Number_of_Other_Players using
a Real variable from (-1..+1)=Trust_Indicator (ex.: +0.91, 0.31, -0.99, -0.075, ...)

Max_Num_of_Alliance_Chance = 6*6*6*6*6*6*6 =279936 (2,8x10e5) (less than a third of a million, small)

--> If it runs through only optimized solution surfaces that are elements of 279936,
a few times (3-9, let's say) during the 'diplomatic phase', it should clock very quick
computer time, when considering that searching for best Move_Plans according to Objectives, and 16384 pointers for alliances, and more for ...

(...even on a single processor, enough seconds (time)--> We can get results now, few hours or in minutes!)


"Sentient Intelligence and Programmed Intelligence...
The strategy Bot & "communication scribe" must have a conversation...
It comes up with a few plans, shows it to the 'babble fish' that gathers input from the 'outside' , then reconsiders plans... & might repeat the process a few times...

"players with a lot of face to face may have better feel for complexities of implementing Diplomacy." (The addition mainly is that you know who is talking to whom, how long, certain reactions, etc.., you know, 'metagaming'.) "


Update-Rehash of: http://webdiplomacy.net/forum.php?viewthread=462962

;;; Map Variables

Land_Codes = (Water, 0, Province, 1, Capital, 2),

Country_Codes = (En="England", Ru="Russia", Fr="France", Tk="Turkey", Ge="Germany", It="Italy", AH="Austria-Hungary"),

Country_Unit_Codes = (EA="English Army", EF="English Fleet", RA="Russian Army", RF="Russian Fleet", FA="French Army", FF="French Fleet", TA="Turk Army", TF="Turk Fleet", GA="German Army", GF="German Fleet", IA="Italian Army", IF="Italian Fleet", AA="Austrian Army", AT="Austrian Fleet"),
;;; + Unit#. ... (Ex.:EF1)


Map_Annular_Codes = (Corner=0, Exterior=1, Ring=2, Middle=3, Center=4),
;;; A geometric 'spiral-chain' system for the 74 provinces of Diplomacy. -- (Have another system to add?)

;;; 4
Corners = (NAO, StP, Syr, MAO),

;;; 11
Exterior = (NWS, BAR, Nor, Fin, Mos, Sev, Arm, EME, ION, Tun, NAf),

;;; 31
Ring = (ENG, Lon, Wal, IRI, Lvp, Cly, Edi, Yor, NTH, SKA, SWE, GOB, Liv, War, Ukr, Rum, BLA, Ank, Smy, Con, AEG, Gre, Alb, ADR, Apu, Nap, TYS, WME, Spa, Por, Gas, Bre),

;;; 20
Middle = (Pic, Bel, Hol, HEL, Den, BAL, Pru, Sil, Gal, Bud, Bul, Ser, Tri, Ven, Rom, Tus, GOL, Mar, Bur, Par),

;;; 8
Center = (Pie, Tyr, Mun, Ruh, Kie, Ber, Boh, Vie),


Map_Values = (
;;; (Code, Land, Annular, Weight, Neighbours, Owners, Full_Name. ,,, =-> What to add/change ??)
;;; (Abc, L, A, 0.nn%, ( , , , ) , (Who?) , Name)

(1, NAO, 0, 0, 0.68, (MAO, NWS, IRI, Cly), (0, En), "North-Atlantic Ocean"),
(2, NWS, 0, 1, 0.72, (NTH, Edi, Nor, NAO, BAR, Cly), (0, En), "Norwegian Sea"),
(3, BAR, 0, 1, 0.64, (Stp, Nor, NWS), (0, Ru), "Barents Sea"),
(4, Nor, 2, 1, 0.82, (Stp, Swe, NTH, NWS, Fin, BAR), (0, Ne), "Norway"),
(5, Fin, 1, 1, 0.71, (Stp, Swe, Nor, GOB), (0, Ru), "Finland"),
(6.a, StP, 2, 0, 0.99, (Mos, Fin, Liv, GOB, Nor, Bar), (RF1, Ru), "St-Petersburg"),
(6.F1, StpS, 2, 0, 0.99, (Fin, Liv, GOB ), (RF1, Ru), "St-Petersburg SouthCoast"),
(6.F2, StpN, 2, 0, 0.97, (Nor, Bar), (0, Ru), "St-Petersburg NorthCoast"),
(7, Mos, 2, 1, 0.96, (War, Sev, Stp, Ukr, Liv), (RA1, Ru), "Moscow"),
(8, Sev, 2, 1, 0.98, (Rum, BLA, Mos, Ukr, Arm), (RF2, Ru), "Sevastopol"),
(9, Arm, 1, 1, 0.68, (Ank, Sev, Smy, BLA, Syr), (0, Tk), "Armenia"),
(10, Syr, 1, 0, 0.58, (Smy, Arm, EME), (0, Ne), "Syria"),
(11, EME, 0, 1, 0.59, (Smy, ION, AEG, Syr), (0, Ne), "East Mediterranean"),
(12, ION, 0, 1, 0.75, (Nap, Gre, Tun, AEG, TYS, Apu, Alb, ADR, EME), (0, It), "Ionian Sea"),
(13, Tun, 2, 1, 0.81, (ION, WME, TYS, NAf), (0, Ne), "Tunis"),
(14, NAf, 1, 1, 0.58, (Tun, MAO, WME), (0, Ne), "North Africa"),
(15, MAO, 0, 0, 0.79, (Bre, Spa, WME, Por, Eng, SpaN, NAO, IRI, NAf, Gas), (0, Fr), "Mid-Atlantic Ocean"),
(16, ENG, 0, 2, 0.91, (Lon, Bre, Bel, NTH, MAO, Wal, Pic), (0, Ne), "English Channel"),
(17, Lon, 2, 2, 0.98, (NTH, ENG, Yor, Wal), (EF1, En), "London)",
(18, Wal, 1, 2, 0.72, (Lon, Lvp, ENG, IRI, Yor), (0, En), "Wales"),
(19, IRI, 0, 2, 0.69, (Lvp, ENG, MAO, Wal, NAO, Cly), (0, Ne), "Irish Sea"),
(20, Lvp, 2, 2, 0.92, (Edi, IRI, Yor, Wal, NAO, Cly), (EA1, En), "Liverpool"),
(21, Cly, 1, 2, 0.59, (Lvp, Edi, NAO, NWS), (0, EN), "Clyde"),
(22, Edi, 2, 2, 0.95, (NTH, Lvp, NWS, Yor, Cly), (EF2, En), "Edinburgh"),
(23, Yor, 1, 2, 0.74, (Lon, Edi, Lvp, NTH, Wal), (0, En), "Yorkshire"),
(24, NTH, 0, 2, 0,89, (Lon, Edi, Bel, Nor, Den, Hol, ENG, NWS, Yor, SKA, HEL), (0, En), "North Sea"),
(25, SKA, 0, 2, 0.68, (Den, Swe, Nor, NTH), (0, Ne), "Skagerrack"),
(26, Swe, 2, 2, 0.88, (Den, Nor, BAL, Fin, SKA, GBO), (0, Ne), "Sweden"),
(27, GOB, 0, 2, 0.68, (Stp, Swe, BAL, Liv, Fin), (0, Ru), "Golf of Bothnia"),
(28, Liv, 1, 2, 0.69, (War, Mos, Stp, BAL, Pru, GOB), (0, Ru), "Livonia"),
(29, War, 2, 2, 0.99, (Mos, Ukr, Gal, Sil, Liv, Pru), (RA2, Ru), "Warsaw"),
(30, Ukr, 1, 2, 0.86, (War, Mos, Sev, Rum, Gal), (0, Ru), "Ukraine"),
(31, Rum, 2, 2, 0.87, (Sev, Bud, BulE, Ser, BLA, Ukr), (0, Ne), "Rumania"),
(30, BLA, 0, 2, 0.88, (Ank, Sev, Con, Rum, BulE), (0, Ne), "Black Sea"),
(31, Ank, 2, 2, 0.97, (Con, Smy, BLA, Arm, Syr), (TF1, Tk), "Ankara"),
(32, Smy, 2, 2, 0.92, (Con, Ank, Arm, EME, Syr), (TA2, Tk), "Smyrna"),
(33, Con, 2, 2, 0.99, (Ank, Smy, Bul, BLA, AEG), (TA1, Tk), "Constantinople"),
(34, AEG, 0, 2, 0.79, (Con, Gre, Smy, BulS, ION, EME), (0, Tk), "Aegean Sea"),
(37, Gre, 2, 2, 0.87, (BulS, Ser, ION, AEG, Alb), (0, Ne), "Greece"),
(38, Alb, 1, 2, 0.78, (Tri, Gre, Ser, ION, ADR), (0, Au), "Albania"),
(39, ADR, 0, 2, 0.68, (Tri, Ven, ION, Alb, Apu), (0, Au), "Adriatic Sea"),
(40, Apu, 1, 2, 0.73, (Nap, Ven, Rom, ION, ADR), (0, It), "Apulia"),
(41, Nap, 2, 2, 0.98, (Rom, ION, TYS, Apu), (IF1, It), "Naples"),
(42, TYS, 0, 2, 0.74, (Nap, Rom, Tun, ION, WME, GLY, TUS), (0, It), "Tyrrhenian Sea"),
(43, WME, 0, 2, 0.69, (SpaS, Tun, MAO, TYS, GLY), (0, Ne), "Western Mediterranean"),
(44.a, Spa, 2, 2, 0.81, (Mar, Por, MAO, WME, Gas, GLY), (0, Ne), "Spain"),
(44.F1, SpaS, 2, 2, 0.83, (Mar, Por, MAO, WME, GLY), (0, Ne), "Spain SouthCoast"),
(44.F2, SpaN, 2, 2, 0.79, (Por, MAO, Gas), (0, Ne), "Spain NorthCoast"),
(45, Por, 2, 2, 0.77, (SpaS, SpaN, MAO), (0, Ne), "Portugal"),
(46, Gas, 1, 2, 0.55, (Bre, SpaN, Par, Mar, MAO, Bur), (0, Fr), "Gascony"),
(47, Bre, 2, 2, 0.95, (Par, MAO, ENG, Pic, Gas), (FF1, Fr), "Brest"),
(48, Pic, 1, 3, 0.73, (Par, Bre, Bel, Bur, NTH), (0, Fr), "Picardy"),
(49, Bel, 2, 3, 0.86, (Hol, NTH, Bur, Ruh, ENG, Pic), (0, Ne), "Belgium"),
(50, Hol, 2, 3, 0.82, (Kie, Bel, NTH, Ruh, HEL), (0, Ne), "Holland"),
(51, HEL, 0, 3, 0.73, (Kie, Hol, Den, NTH), (0, Ge), "Helgoland Bight"),
(52, Den, 2, 3, 0.81, (Kie, Swe, NTH, Bal, SKA, HEL), (0, Ne), "Denmark"),
(53, BAL, 0, 3, 0.70, (Kie, Den, Ber, GBO, Pru, Liv), (0, Ge), "Baltic Sea"),
(54, Pru, 1, 3, 0.67, (Ber, War, Liv, Sil, BAL), (0, Ge), "Prussia"),
(55, Sil, 1, 3, 0.71, (Ber, Kie, Bur, Tyr, Ruh, Sil), (0, Ge), "Silesia"),
(56, Gal, 1, 3, 0.78, (Bud, Vie, War, Rum, Ukr, Sil, Boh), (0, AH), "Galicia"),
(57, Bud, 2, 3, 0.98, (Ser, Tri, Rum, Vie, Gal), (AA1, AH), "BudaPest"),
(58.a, Bul, 2, 3, 0.83, (Con, Rum, Ser, Gre, AEG, BLA), (0, Ne), "Bulgaria"),
(58.F1, BulE, 2, 3, 0.81, (Con, Rum, Ser, BLA), (0, Ne), "Bulgaria EastCoast"),
(58.F2, BulS, 2, 3, 0.85, (Con, Gre, Ser, AEG), (0, Ne), "Bulgaria SouthCoast"),
(59, Ser, 2, 3, 0.88, (Bud, Tri, Gre, Rum, Bul, Alb), (0, Ne), "Serbia"),
(60, Tri, 2, 3, 0.98, (Bud, Vie, Ser, Ven, Tyr, Alb), (AF2, AH), "Trieste"),
(61, Ven, 2, 3, 0.98, (Rom, Tri, Tyr, Pie, Apu, Tus), (IA2, It), "Venice"),
(62, Rom, 2, 3, 0.96, (Ven, Nap, TYS, GLY, Apu, Tus), (IA1, It), "Rome"),
(63, Tus, 1, 3, 0.82, (Rom, Ven, Pie, TYS, GLY), (0, It), "Tuscany"),
(64, GOL, 0, 3, 0.75, (Mar, SpaS, WME, TYS, Pie, Tus), (0, Fr), "Golf of Lyons"),
(65, Mar, 2, 3, 0.98, (SpaS, Bur, Pie, GOL, Gas), (0, Fr), "Marseilles"),
(66, Bur, 1, 3, 0.89, (Par, Mun, Mar, Bel, Ruh, Pic, Gas), (0, Fr), "Burgundy"),
(67, Par, 2, 3, 0.99, (Bre, Bur, Pic, Gas), (0, Fr), "Paris"),
(68, Pie, 1, 4, 0.78, (Mar, Ven, Tyr, Tus, GLY), (0, It), "Piedmont"),
(69, Tyr, 1, 4, 0.79, (Ven, Tri, Mun, Vie, Pie, Boh), (0, It), "Tyrolia"),
(70, Mun, 2, 4, 0.99, (Kie, Ber, Bur, Ruh, Tyr, Sil, Boh), (GA1, Ge), "Munich"),
(71, Ruh, 1, 4, 0.79, (Mun, Kie, Bel, Hol, Bur), (0, Ge), "Ruhr"),
(72, Kie, 2, 4, 0.97, (Mun, Ber, Den, Hol, Ruh, Bal, HEL), (GF1, Ge), "Kiel"),
(73, Ber, 2, 4, 0.95, (Mun, Kie, Bal, Sil, Pru), (GA2, Ge), "Berlin"),
(74, Boh, 1, 4, 0.74, (Vie, Mun, Gal, Tyr, Sil), (0, AH), "Bohemia"),
(75, Vie, 2, 4, 0.99, (Bud, Tri, Gal, Tyr, Boh), (AA2, AH), "Vienna") );


;;; ------
;;; Notes!:

;;; Neighbours are ordered by 'proposed' importance, and can vary...)

;;; Switzerland, is impassable, thus merely excluded.

;;; For Fleets we use seperate list to verify:
;;; Stp = StpS + StpN, Bul = BulS + BulN, Spa = SpaS + SpaN,

;;; ( where '+' is an ADDL, (Add Logically & proportionaly))

;;; The Weights % do need some more fine tuning...

;;; Is the 75x75 Distance_To_Another_Province Matrix really usefull for a bot?
;;; In what cases do we check for actual distances? That bings up the topic of Paths!
;;; Which approach?...

!!! I hope all this will serve later. (I'm out.) ;-))...

Is there really 4.5 quadrillion ways to play the first move? Did I read that correctly? Is that just spring 1901 or is that the first year (including retreats and builds)?

For the first move, I'd guess.

If every piece has 10 legal moves (hold, move to each neighbouring square and all the legal support move and support hold moves) you get 10,000,000,000,000,000,000,000 moves.

That's pretty amazing.

Keep it simple !!!

School's Back,,, anyone care to take this 'challenge' up?

9,909,497,515,357,440,000,000 moves? Nearly 10 sextillion moves.

I'm not going to read all of this, but I have a degree in computer engineering with a speciality/focus in AI, and I'll drop my theory on bots in diplomacy:

You can do a variety of minor things in Diplomacy with bots. I'd suggest making 5 or so 'style' of bots (conservative, kamikaze, etc...) and which bot is chosen is random, but it would only be worthwhile in gunboat games or to take over a CD. Why? Because this is a game of diplomacy... speaking... interacting.
Making a bot that will interact will create one of two things:
1.) It will be so simple it will be figured out and players can 'game' the bot before the end of the first game they play with one
2.) It will be so complex that the code will be an unmaintainable mess, the bot won't behave correctly (because of the complexity), and still won't be nearly as fun to play as a regular player

I'd say a bot in anything but a gunboat game (and only as a teacher or when you can't get 7 players) is a waste of time in a game like diplomacy. Tactics and strategy is a very very small part of the game. The biggest part is the interaction, and there is no decent AI techniques for interaction, especially at the levels required for this game.

If you are going to make a gunboat bot, I'd suggest doing something like a chess bot... look ahead 3, 4, maybe even 5 moves. Of course, the randomness of who is allying and who isn't will make the calculations a complete mess, and it may take quite a while for the bot to determine what to do...