| Version | Date | Name | Change |
|---|---|---|---|
| 1.0 | 2001-August-31 | Lucas B. Kruijswijk | First public version |
| 1.1 | 2001-August-31 | Flooey X. McBob | Added Test Cases 2.C, 4.E (renumbered following) |
| 1.2 | 2001-September-3 | Lucas B. Kruijswijk | Added Test Cases 2.C, 4.A, 4.B, 4.C, 4.D (renumbered following) |
| 1.3 | 2001-November-18 | Lucas B. Kruijswijk | Major update. Processed review comments from Brian Roberts, Andrew Rose, Rick Desper, Simon Szykman and Millis L. Miller |
| 1.4 | 2001-November-24 | Lucas B. Kruijswijk | Added comments of Allan B. Calhamer on issue II.D |
| 2.0 | 2003-December-13 | Lucas B. Kruijswijk | Major update. Comments collected during year processed. Adjudication algorithm added. Chapters rearranged. Some small changes in preferences. Choices of issues now numbered instead of bulleted. |
| 2.1 | 2003-December-16 | Lucas B. Kruijswijk | Small textual changes after comments from David McCooey. Added unwanted support. |
| 2.2 | 2004-January-23 | Lucas B. Kruijswijk | Added note on DATC compliancy verification. Added test case 6.D.34, about support targeting own area. Small textual changes. |
| 2.3 | 2004-February-6 | Lucas B. Kruijswijk | Small textual changes after remarks from Christian Hagenah. Added version number. Changed license. Note on certification after change. Added issue 4.A.7 and test cases related to this issue. Added test 6.E.15, which the DPTG fails to adjudicate correctly. |
| 2.4 | 2004-February-10 | Lucas B. Kruijswijk | Corrected test cases 6.G.16, 6.G.17 and 6.G.18. |
Reviewed by:
You are free to copy the whole or parts of this document. If parts are copied in another document, you may use those parts as if they are written by yourself. However, if any changes are made to this document, make very clear that it is not the same document anymore.
Diplomacy is the Avalon Hill Game Company's trademark for its game of international intrigue, which game is copyright 1976 by Avalon Hill. Avalon Hill belongs to Hasbro.
1. INTRODUCTION
2. 2000 RULEBOOK/DATC COMPLIANCY
3. HISTORY OF RULES
4. DISPUTABLE ISSUES
A. CONVOY ISSUES
B. COASTAL ISSUES
C. UNIT DESIGNATION AND NATIONALITY ISSUES
D. TOO MANY AND TOO FEW ORDERS
E. MISCELLANEOUS ISSUES
5. THE PROCESS OF ADJUDICATION
A. OVERVIEW OF ADJUDICATION DECISIONS
B. PRECISE DESCRIPTION OF MAKING DECISIONS
C. FROM DECISIONS TO ALGORITHM
D. SIMPLIFICATIONS AND OPTIMIZATIONS
E. PERFORMANCE
6. TEST CASES
A. TEST CASES, BASIC CHECKS
B. TEST CASES, COASTAL ISSUES
C. TEST CASES, CIRCULAR MOVEMENT
D. TEST CASES, SUPPORTS AND DISLODGES
E. TEST CASES, HEAD TO HEAD BATTLES AND BELEAGUERED GARRISON
F. TEST CASES, CONVOYS
G. TEST CASES, CONVOYING TO ADJACENT PLACES
H. TEST CASES, RETREATING
I. TEST CASES, BUILDING
J. TEST CASES, CIVIL DISORDER AND DISBANDS
7 COLONIAL VARIANT
A. TEST CASES, HONG KONG OPTIONAL RULE
B. TEST CASES, TRANS SIBERIAN RAILROAD OPTIONAL RULE
C. TEST CASES, SUEZ CANAL OPTIONAL RULE
D. TEST CASES, TRANS-SIBERIAN RAILROAD AND SUEZ CANAL COMBINED ISSUES
8 ICE VARIANT
9 CONVOYING COASTAL AREA VARIANT
10 DIFFICULT PASSABLE BORDER VARIANT
11 BUILD IN ANY SUPPLY CENTER VARIANT
12 1898 VARIANT
1. INTRODUCTION
Writing an adjudicator computer program for the game Diplomacy (TM) is not an easy job. Many adjudicator programs contain several bugs on their first release and even after some years, when the most severe bugs were removed, adjudication errors could still be found for more complex situations. To achieve a high quality adjudicator, the programmer has to overcome the following difficulties:
This document is a guide in handling these problems. With this document it is possible to write an adjudicator that is correct for the first release.
The principle of this document is to give the reader information and not to tell the reader how to do things. Therefore, all information is presented as neutral as possible and given with arguments or with reference to the source.
The primary sources of this document are the official English rulebooks. The additional sources are the 1998 DPTG (Diplomacy Players Technical Guide), comments from mister Calhamer (the creator of the game), magazines from the publishers of the game, various articles, discussions and house rules found on the internet.
A scientific approach has been taken to deal with all this information. If there was evidence that a certain idea or opinion is supported by a significant part of the Diplomacy (TM) community, then it was added to this document. This makes this document the most elaborated and complete source on the rules of Diplomacy (TM). Even so, it was not possible to list everyone's individual opinion.
An overview of the official English rulebooks is given in chapter 3. These rulebooks have in certain cases a different ruling and even when the latest rulebook is taken, then still some rule issues exist. These issues are listed in chapter 4 with the alternatives to handle them. An analysis of the process of adjudication is in chapter 5. Writing an adjudicator program is a straight forward job with this analysis. Finally, an adjudicator program needs to be tested in a systematic way and chapter 6 contains an extensive list of test cases with the expected resolution.
In case the rules are not clear on a certain situation, then the alternatives are in the list of issues and in the test cases. To prevent that the reader becomes lost in lots of choices I (Lucas B. Kruijswijk) commented every rule issue. Those comments contain my preference how the rule should be interpreted. In this way the reader can start with these preferences and deviate where he or she wishes. To make a clear distinction between the text that is not disputed and my comments with my preferences, my comments are written in italics. Note however, that my preferences are based on an extensive study on the rules. Although you may dispute them, the choices are certainly very acceptable within the Diplomacy(TM) community.
With the elaborated information about the rules, this document can also be used by people that judge a game manually. The preferences can be used as default house rules in case the own house rules of the judge does not cover a certain topic.
The version of this document consists of two numbers. The first number will change and only change, when one or more preferences changes. The second one, is a sequence number.
In chapter 7 and further variants are handled.
Variants are games that follow the rules of Diplomacy with a
different map or with some additional rules. The chapters on variants
are not complete yet. For instance, Air Units and Machiavelli are not
handled yet.
2. 2000 RULEBOOK/DATC COMPLIANCY
Everyone is free to interpret the rules as he or she wants. But people want to know, which rules are used. Therefore, the term '2000 rulebook/DATC compliant' is introduced. An automated adjudicator may call itself '2000 rulebook/DATC compliant' when the following requirements are met:
The idea of the DATC is not to enforce one way of interpretation of the rules. Therefore, you may call an adjudicator '2000 rulebook/DATC compliant' as long as you list the deviations with the specified requirements. However, the adjudicator may not fail on any of the test cases (of course, the results must be according to chosen preferences). Deviations does not need to be listed, when they can be turned off or when it is an extension in a client program that does not change anything when it is not used (for instance choice b for 4.A.6, choice a for 4.E.4 and choice a for 4.E.5).
Developers are considered to be honorable
persons and may perform verification by their selves.
After an update of the DATC an adjudicator does not
need to be retested to keep the status of
'2000 rulebook/DATC compliant'.
3. HISTORY OF RULES
Allan Calhamer started the development of Diplomacy in 1953. However the first commercial version was not on the market before 1959. So, the first real rulebook is from 1959.
Games Research bought the rights of Diplomacy in 1960 and their first edition was in 1961. Except from some restyling and the copyright notice of Games Research, the rules of 1961 are not different from the 1959 rules. These rulebooks have many ambiguities and are rather useless. Dealing with the issues of these rulebooks is out of the scope of this document.
Most of the issues of the 1959 and 1961 rulebooks were identified and addressed in the 1971 rulebook. The 1971 rulebook is the first mature rulebook. Since this rulebook has been popular for a long time and maybe still people play according to these rules, the differences with newer rulebooks are all explained in this document.
In 1976 Avalon Hill bought the rights from Games Research. In their first edition in 1976, they kept the rules the same and they only put their own name on the rulebook. In this document, only to the first appearance of a set of rules is referred. That means that the 1971/1976 rulebooks are referred as the 1971 rulebook in this document.
In 1982 Avalon Hill made a new edition of the game and made some small, but significant changes to the rules. These changes include the multi-route convoy disruption rule (see issue 4.A.1), convoy disruption paradox rule (see issue 4.A.2) and some clarifications on convoying to adjacent places (see issue 4.A.3).
In 1992 Avalon Hill they introduced the 'Diplomacy de Luxe' edition. The rulebook was restyled, but the rules are identical to the 1982 rules. The rulebook does also contain a list of abbreviations for the provinces. Since, the rules did not change, the 1982/1992 rulebooks are referred as the 1982 rulebook in this document.
In 1998 Avalon Hill was bought by Hasbro and in 2000 the first Hasbro edition emerged. Note that you can see the year 2000 on the front page of the rulebook, however the copyright notice on the last page says 1999. Some people refer to this rules as the 1999 rules. In this document it is assumed that the front page is read more than the last page and therefore these rules are referred to as the 2000 rules. To make the rules more readable, the rules were completely rewritten. Edi Birsan was the main contributor to this rulebook. Except from restyling, there are also some real changes to the rules. The unpopular 1982 convoy, disruption paradox rule (see issue 4.A.2) was changed again. The issue on convoys to adjacent provinces was further clarified (see issue 4.A.3). Waiving builds is explicitly allowed (see issue 4.D.7) and you can very well interpret the rules in such way that they do not allow the refusal of support by ordering an illegal order (see issue 4.E.1). Also a significant change in the abbreviations of the provinces (note that the abbreviation of 'Denmark' is missing).
Unfortunately there are a few changes in the 2000 rulebook that made things worse than the 1982 rules. A rather minor issue is the removal of units in civil disorder (see issue 4.D.8). In the 1971 and the 1982 rulebooks this was already rather vague, but in the 2000 rules it is really a mess. A more serious error is that it was forgotten to say that a failed convoy due to dislodged convoying fleets, has no influence on the destination province. Page 12 of the 2000 rulebook:
Dislodgement of a fleet in a convoy causes the convoy to fail. If a Fleet ordered to convoy is dislodged during the turn, the Army to be convoyed remains in its original province.
While rule XII.3 of the 1971 and 1982 rules say:
DISRUPTING A CONVOY. If a fleet ordered to convoy is dislodged during the turn, the army to be convoyed remains in its original province and has no effect on the province to which it was ordered.
There are several reasons to assume that the 2000 rulebook was not intended like this and that this must be considered an error in the rules. First of all, Edi Birsan the main contributor to the 2000 rulebook, confirmed that this was not intended. Second, there is evidence in the rulebook itself that the rule should be read as the 1971/1982 rule. The whole page 16 would become unnecessary if convoying armies can still cut support when the convoy is dislodged (there are no paradoxes anymore). But especially the following phrase on that page is clear evidence:
Italy could argue that dislodgement of the Fleet disrupted the convoy so that the Army could not arrive in Naples to cut the support. (Italy could state the rule, "Dislodgement of a fleet in a convoy causes the convoy to fail.")
Since in the research for this document no evidence was found that there is a significant group of people that actually play that a disrupted convoy can still cut support (or bounce another unit), this is not listed as a disputable issue in chapter 4, but just treated as an error and that the 1971/1982 rule was meant.
Finally, there are numerous alternative rulebooks. For instance the rulebook of the variant Colonial Diplomacy. This rulebook is based on the 1971 rules (although the copyright is after 1982) and is discussed in chapter 7. Furthermore there are several translated rulebooks made for non-English versions of Diplomacy. These are not treated in this document. With the 2000 edition also an CD-ROM version was released by Hasbro/Avalon Hill/Microprose. This program has a tutorial for the rules. In the research of this document no additional information could be found about whether this version contains a paper rulebook and if it is any different from the rulebook supplied with the board game. So, obviously this version is not handled in this document and if it exists and it contains the 1999 or 2000 copyright notice, then we still refer to the rulebook of the board game when the 2000 rulebook is mentioned.
As last remark must be said that according to some reliable
sources, Allan Calhamer has the right to veto any changes to the rules.
4. DISPUTABLE ISSUES
When playing Diplomacy, there is one golden rule. Making orders public is irreversible. If during a turn someone discovers that the adjudication was wrong, this can be corrected and even when the orders are submitted, the orders can discarded and be rewritten. However, when the orders are made public, the players have shown their true intentions and it is not possible to hide them again.
This golden rule has two consequences. First, when the orders are made public, the adjudication of the previous turn, becomes final. Changing the adjudication of a previous turn, would lead to a big mess. Second, when the orders are made public, a decision has to be made on those orders. This may lead to a serious dispute between the players, when those orders are not clear or the rulebook is disputable on the situation. Therefore, the main focus is on issues related to the pieces on the board and interpreting the written orders (some of them are irrelevant for an automated adjudicator, but are listed nevertheless). So, the problem of assigning powers to the players is not discussed, since this can easily be handled by the players.
The geography of the map is not disputed and therefore not discussed. To surprise to some new players, Norway is connected to St Petersburg both for army as for fleet, but this is not disputed. Although also not disputed, on some maps it is not completely clear, that Liverpool is connected to North Atlantic Ocean and Clyde is not connected to Irish Sea.
The issues in this chapter are not a FAQ (Frequently Asked Question) list, but most frequently asked questions can be found in the test cases. For instance the coastal crawl in test case 6.B.13.
After extensive study of the rules and discussions on the
internet, my preferences are finally based on the following principles:
I do not consider any aesthethical arguments. To my opinion, a rule is
aesthethical when it is based on the points made above.
4.A. CONVOY ISSUES
4.A.1. MULTI-ROUTE CONVOY DISRUPTION
When a convoy has multiple routes, the question rises when the convoy is disrupted.
The following interpretations are possible:
The 1971 rule has as disadvantage that a player that has a fleet that is almost certainly dislodged, can give an "unwanted convoy" as in test case 6.F.10. The DPTG avoids this problem by skipping any foreign units from a convoy when there is a route without foreign units. However, this solves the problem only partially. When there is no route without foreign units and there are two routes with each a different foreign fleet, it can not be decided based on the orders whether one of the convoy orders is unwanted (see test case 6.F.11). Probably due to these issues the rule was changed in the 1982 rulebook.
I prefer to use the 1982 and 2000 rulebook (choice b) on this
issue, because it prevents the mentioned problems. So, a convoy is
disrupted when all routes are disrupted.
4.A.2. CONVOY DISRUPTION PARADOXES
A convoy disruption paradox is a situation with a possible disrupted convoy and for which the rules give no resolution or more than one resolution.
Some people argue that some situations are not convoy paradoxes, since the rules give a resolution for those situations. Hence, for a proper discussion on paradoxes, rule XII.5 of the 1971, 1976, 1982 and 1992 rulebooks and the rule on the top of page 16 of the 2000 rulebook, should not be taken into account in the above definition of a paradox.
In case of a paradox, the smallest subset of orders for which the paradox still exists, is the core of the paradox. In case there are several independent paradoxes on the board, then it is possible that there are two different subsets of orders with the same number of orders. In that case, just one can be taken to proceed. Since the paradoxes are independent it doesn't matter which one is handled first.
The following ways for handling convoy disruption paradoxes are possible:
Several people proposed alternatives for the paradox problem. However, only the suggestions above have significant support from the Diplomacy community. Some people use the principle of 'cause' in their alternative. Where 'cause' can be a direct cause, but also an indirect cause. It is rather unclear what the advantages are of introducing this new notion, while it introduces significant complexity for automated adjudication.
For making a preference it is necessary to decide what is important and
what is less important. I think that the adjudication result in case of
a paradox is the least important thing, since the paradoxes are a rare
phenomenon in actual play and any rule in case of a paradox does not
change the balance of the game. However, I think the following is
important:
Based on these guidelines, I prefer the Simon Szykman alternative
(choice d). It can be enough defended that this rule is compatible
with the 2000 rulebook. Furthermore the case which can be disputed on
the compatibility of the Szykman rule with the 2000 rulebook, is very
theoretically and will not appear in real play. I have the impression
that the 1982 rule has never been very popular, since
I did not find many house rules that follow the 1982 rule (although I did
find some adjudicators that follow the 1982 rule).
4.A.3. CONVOYING TO ADJACENT PLACE
It is allowed to convoy to an adjacent place. This can be used to swap two units (such as in test case 6.G.1). If the units are ordered in such way that a convoy to an adjacent place is possible, the question arises whether the convoy route must be considered or the land route.
It can not just be said that the land route is taken or the convoy route, comparable with the multiple convoy routes rule. The reason for this is that different rules apply in the different situations. If an army takes the land route and the unit on the target area moves in the opposite direction, then there is a head to head battle. While in case of the convoy route there is no head to head battle. If the attacked unit is dislodged by the head to head battle, the attacked unit has no effect on the area from which it is attacked, while in case of a convoy it still has effect (see test cases 6.G.9 and 6.G.10). It is also possible that the convoy route is only available when the land route is taken (see test case 6.G.11). Finally it has a relation with the issue of cutting support by a convoy to adjacent place (see 4.A.4) and the issue of retreat after dislodged by a convoy to adjacent place (see 4.A.5).
For these reasons it must be clear which route is taken at some point during adjudication.
The following interpretations are possible:
11. Q: We think there is a conflict in the Rulebook. These orders might not occur often, but what if they do?Given this information, the following rule can be defined: "The land route is taken except when the unit in the target area moves in opposite direction.". This prevents at least the type of kidnapping as in test case 6.G.3, but some unwanted convoys are still possible as in test case 6.G.2 and 6.G.4. An advantage of this rule is that issue 4.A.4 is not relevant anymore and the already theoretical issue 4.A.5 is even more theoretical.
ENG: F Edi-Nth, F Lon S F Edi-Nth.
FRA: F Nth C GERMAN A Hol-Bel.
GER: A Hol-Bel.
Doesn't the disruption of the convoy prevent the German army from moving?
A: No. The Rules (VII.1, third paragraph) clearly state that an army may move to an adjacent province if unopposed. In this situation, the convoy order is not relevant to the army's ability to move from Holland to Belgium. Even if the F Nth were German, so that it was clear that the German player intended (for some arcane reason) that the army should move by convoy, Rule VII.1 would still operate to allow A Hol-Bel to succeed if unopposed. Actually, the only time you might expect to see an army convoyed between adjacent provinces is when an otherwise illegal exchange of units is taking place (see Rule XIV.5, third paragraph). Consider this situation:
ENG: F Edi-Nth, F Lon S F Edi-Nth.
GER: F Bel-Hol, A Hol-Bel, F Nth C A Hol-Bel.
In this situation, the German F Nth is dislodged and the convoy is disrupted. The two German movement orders, both still legal, now boil down to a stand-off.
There are two notations to order an army that it should use a convoy. The "by Convoy" notation is the most common notation, however the "via Convoy" notation is used in the 2000 rulebook. Through this document the notation of the 2000 rulebook is used.
The 2000 rule solves indeed the problem of kidnapped
armies. However, the rule is maybe a little too much. It
introduces quite a lot of details. Furthermore, in the
research of this document a case was found were two players
(in a game played with the 2000 rules) wanted to swap,
but didn't give the orders properly. It must be realized
that a swap with a foreign fleet maybe only happens one out
of ten games. If a face to face game takes about six hours,
then you need sixty hours of play to have at least one
experience with the 'via Convoy' order. So, what is worse?
Take also into account that the kidnapping of armies is also
fun. I advice that if there is a new edition of the
rules that choice c must be considered (after analyzing
a few dozen of games on the consequences). For the moment
I prefer to follow the rules and that means choice d.
4.A.4. SUPPORT CUT ON ATTACK ON ITSELF VIA CONVOY
A move can not cut a support, if the support is a support of an attack on itself. But what if the move is via convoy (see test case 6.G.13)?
Note, if a convoy to an adjacent place can only take place when the unit moves in opposite direction (choice b an c for issue 4.A.3), then this issue is not relevant anymore. Then the moving unit will take the land route and the support will not cut.
The following two interpretations are possible:
The main issue is the interpretation of the word 'from' as used in the rulebook. If 'from' is interpreted as the starting position of the army being convoyed, then the support is not cut. However, if the attack is coming 'from' the body of water, then the support is cut.
This has been discussed on the newsgroup 'rec.games.diplomacy' and on the 21th November 2001 Randy Hudson and Mike Lease gave the following arguments why 'from' should be interpreted as the starting position of the army:
It's not phrased as a clarification of the "cutting support" rule (Rule X) in the 1976 rules. I've now printed out a copy of the 2000 rules from the Hasbro web site, and in the explanation, it offers the example:
France: A Tun-Nap, F Tyn C A Tun-Nap
Italy: F Ion-Tyn, F Nap S F Ion-Tyn
The explanatory text goes on to say that France could argue that support is cut (thus preventing the convoy from being disrupted), citing the rule, "Support is cut if the unit giving support is attacked from any province but the one where support is being given." IOW, since the army is coming from Tunis, it would normally be entitled to cut the support for F Ion-Tyn given by Naples, thereby preventing the convoy from being disrupted. This "new rule" (author's words) gives an exception to that rule, overriding the usual rule and eliminating the paradox. This rule WOULD NOT BE NECESSARY if the army were deemed to be coming from Tyn (the space being attacked). But the author says it IS NECESSARY ("...this rule takes precedence" [over the usual rule that would apply]), therefore, the army MUST be deemed to come from Tun, and this rule exists to provide an exception to avoid paradox. But it only applies to a situation in which a convoy would be disrupted if the support is effective, but not disrupted if the support is ineffective. Therefore, for all other convoys, the ordinary interpretation should be that the army is deemed to come from the province in which it began the turn, and thus an army which can reach a province either via land or via convoy cannot cut support for an action against the province in which it started the turn, whether or not it is convoyed to the supporting unit's province. QED.
Stephen Agar sent this problem to Allan Calhamer (the creator of Diplomacy). Before he got a response he followed up with the arguments of Randy Hudson and Mike Lease. On the 24th of November 2001, Allan Calhamer responded:
Good argument. I had always thought of the Army as coming from its province, not from the body of water, but your correspondent appears to show actual rule support for the proposition.
Best regards, Allan
So, according to the creator the attack is coming from the starting position of the army.
However, you can have also another view on this issue. If you look to support as something that is passing a border, then the support is not cut when the attack and support are clashing on the same border. In case of a convoy, there is no such clash, so it would be more logical to say that the support is cut (choice b). Taking this border point of view, it would be preferred that the choice is consistent with issues 4.A.5 and 4.A.7.
Of course, I follow the interpretation of the creator.
Therefore, I prefer that the support is not cut (choice a).
4.A.5. RETREAT WHEN DISLODGED BY CONVOY
In a very rare situation (see test case 6.H.11 and 6.H.12) a unit can be dislodged by a convoy from an adjacent place. Then the question arises whether the dislodged unit may retreat to the starting place of the convoying army.
The following two interpretations are possible:
First note that if issue 4.A.3 is played according to choice c, then this issue is not relevant anymore. If it is played according to choice b, then this is only an issue in the extremely theoretical situation that both units attempted to move by convoy (see test case 6.H.12).
All the rulebooks are ambiguous on this issue. They say first "A dislodged unit must retreat to an adjacent province that it could ordinarily move to if unopposed by other units.". According to this sentence the unit may retreat to the starting place of the attacker (choice b). However, in further explanation all the rulebooks say: "The unit can not retreat to the province from which the attacker came.". For the interpretation of the word 'from' see also issue 4.A.4. This sentence must be interpreted that the particular retreat is not allowed.
Since the rulebooks are ambiguous on this issue, I prefer
the option that seems most logical. Since the attacking unit did not
pass the border over which the retreating unit wants to go, it is
logical that this retreat is possible. So, I prefer choice b.
4.A.6. CONVOY PATH SPECIFICATION
Some popular automated judges require that a path is specified for the army that convoys. For instance:
England: F North Sea Convoys A Yorkshire - Belgium A Yorkshire - North Sea - Belgium
This way of writing convoys has never been part of the official rules and is also not according to the DPTG. According to some sources this is because the person that wrote the first automated judge, did not know how to write an algorithm that searches for a convoy path and that since then it has never been changed. This is a little bit strange story, since writing an algorithm that searches for a path is much easier then writing an adjudicator. On the other hand the quality of the first version of the adjudicator might not have been that good.
The advantage of path specification is that it might easier to code and that "unwanted" convoys are not possible. There are four types of "unwanted" convoys:
The first three unwanted convoys are not possible with the 2000 rulebook. So, the argument of preventing unwanted convoys is hardly valid.
This issue can be handled in the following ways:
If path specification is required in a face to face game, this
may lead to problems. Suppose that someone clearly intents to convoy,
but forgets the path. For instance, England orders York to Holland
with a fleet in the North Sea that is ordered to convoy. If path
specification is required, then the order from York to Holland without
path makes no sense. But, since we do a best effort to interpret the orders,
we may still decide that the army is convoyed. If we allow this, then
we get all issues related to multi-route convoying back. So, I do
not prefer that path specification is required in face to face games
(I have not seen any house rules for manually adjudicated games,
that require path specification). Since, I prefer that face to face
games are played as much as possible the same as postal and email
games, and since I prefer that is played according to rules
I prefer that path specifications are not required and ignored
(choice a). However, for compatibility reasons I can also live
when both are allowed (choice b) but only for automated judge
systems.
4.A.7. AVOIDING A HEAD TO HEAD BATTLE TO BOUNCE A UNIT
The rulebooks say that if a unit is dislodged, then it has no influence on the area where the attacker came from. Of course, this is only significant when the units move in opposite direction. However, when two units move in opposite direction, they do not necessary be engaged in a head to head battle. It is possible that one of the units convoys. The question arises, whether the dislodged unit can still bounce a third unit. See also test case 6.G.10, 6.G.14 and 6.G.15.
The following interpretations are possible:
Although choice a is more according to the rulebook, I think choice
b is how it is intended. Therefore I prefer choice b. This choice is also
consistent with choice b of issue 4.A.5, where we
really look what is passing the border, instead of looking where the unit
comes from.
4.B. COASTAL ISSUES
4.B.1 OMITTED COAST SPECIFICATION IN MOVE ORDER WHEN TWO COASTS ARE POSSIBLE
If a move order of a fleet to a multi-coast area does not contain a coast where the fleet can move to both coasts, different adjudicators may react differently.
The following interpretations are possible:
The disadvantage of the default coast is, that there is no standard for default coasts and there is no graphical notation to show what the default coast on the map is.
I prefer that the move fails (choice a), because of the disadvantages
of the default coast.
4.B.2 OMITTED COAST SPECIFICATION IN MOVE ORDER WHEN ONE COAST IS POSSIBLE
If a move order of a fleet to a multi-coast area does not contain a coast where the fleet can only move to one coast, different adjudicators may react differently.
The following interpretations are possible:
As I said before I do not like the idea of a default coast.
I also do not see the need for specifying a coast when only one
coast is possible. Therefore I prefer that a move is attempted to
the only possible coast (choice a).
4.B.3 MOVE ORDER TO IMPOSSIBLE COAST
If a move order of a fleet to a multi-coast area does contain a coast but the specified coast is not possible, different adjudicators may react differently.
The following interpretations are possible:
I do not like the idea of changing an unambiguous order. Therefore
I prefer that the move fails (choice b).
4.B.4. COAST SPECIFICATION IN SUPPORT ORDER
The rulebooks clearly specify that a fleet can support another fleet to a coast which it can not reach (a fleet in Greece can support a fleet from the Black Sea to the east coast of Bulgaria). However, it is less clear whether the coast must be specified in the support order. This is one of the most controversial issues. Although the situation is rare because of the limited number of places on the board where it can happen, it is not theoretically and it will happen once in a while.
The following interpretations are possible:
Stephen Agar wrote:
Allan
Could you spare the time to give your view on how to adjudicate the following situations - as you can see they all depend on how you treat the coast designations in each situation.
Turkey: F Con-Bul(ec)
Russia: A Rum S Turkish F Con-Bul(sc)
Austria: A Bul Holds
Turkey: F Con-Bul(ec)
Russia: A Rum S Turkish F Con-Bul
Austria: A Bul Holds
Turkey: F Con-Bul(ec); A Rum S F Con-Bul
Austria: A Bul Holds
Many thanks.
PS. The next issue of Armistice Day is in the post - hope you like the cover!
Regards
Stephen Agar
Dear Stephen: I would consider the supports in the first two cases no good, since I think allies should be in evident agreement on the moves. The support in the third case I would consider good.
Best regards, Allan
I do agree with the concept of mister Calhamer that players should
be in evident agreement on the moves. Therefore, I do not like choice e.
However, mister Calhamer is a conceptual thinker and not a rule thinker.
In the research for this document I found some evidence that automated
judges changed the adjudication in favor of choice d after some complains.
If the conceptual right choice c of mister Calhamer is compared with the
practical choice d, the following two situations must be considered. The
first situation with choice c is that a person gives a support to another
player, but that support fails because the coast was missing. The second
situation is with choice d that a person gives a support to another player
without coast specification and the other player betrays and moves to
the not expected coast. Both situations are undesirable and must be avoided
if possible. However, I consider the first situation far more likely to
happen then the second situation. Furthermore in the second situation
the support giving player had still the opportunity to avoid the unexpected
move by specifying the coast (this is not completely true since the other
player could pressure the player to give support without coast, but this
is not very likely since that would be very suspicious). Finally, choice
d is easier for new players, because the issue does not need to be mentioned
and the adjudication is with or without coast specification more or less
what they expect. For this reason I consider choice d the most supportive
for the game and that is why I prefer that coasts may omitted, but must
match when they are specified (choice d).
4.B.5. WRONG COAST OF ORDERED UNIT
How should an order directed to a fleet on the north coast be executed when the fleet is actually on the south coast?
The following interpretations are possible:
I prefer that such garbage in the orders is just ignored (choice b).
4.B.6 UNKNOWN COASTS OR IRRELEVANT COASTS
How should an order be executed that contains an unknown coast, such as Spain west coast or Brest east coast. And how should an order be executed when the coast is just irrelevant, such as an army movement to the north coast of Spain?
The following interpretations are possible:
I prefer that such garbage in the orders is just ignored (choice b).
4.B.7 COAST SPECIFICATION IN BUILD ORDER
When a fleet is build in multi-coast area (St Petersburg), the coast must be specified. When the player does not specify the coast, the situation can be handled in different ways.
The following interpretations are possible:
As said before, I do not like the idea of default coasts.
Therefore, I prefer that the build fails (choice a).
4.C. UNIT DESIGNATION AND NATIONALITY ISSUES
4.C.1 MISSING UNIT DESIGNATION
When the designation of the type of a unit (A or F) is omitted, the following could be decided:
Since the order is not ambiguous, I prefer that the order is still
valid (choice b).
4.C.2. WRONG UNIT DESIGNATION
When the designation of the type of a unit (A or F) does not match with the actual type of the unit, the following could be decided:
I think a best effort should be made to interpret the order,
therefore I prefer that the order is still valid (choice b). This means that the
unit type designation is in fact ignored. You can argue whether the
unit type designation has any use. However, if other parts of the order
are unclear, the unit type designation might help.
4.C.3. MISSING UNIT DESIGNATION IN BUILD ORDER
A player might forget to specify whether it wants to build an army or a fleet. Such situation can be handled in the following ways:
I prefer that a best effort is made. Therefore I prefer that the
order only fails when the area is coastal area and the order did not
contain a coast (choice c).
4.C.4. BUILDING A FLEET IN A LAND AREA
A player might try to build a fleet in area that can not contain fleets. Such situation can be handled in the following ways:
First of all, I do not consider such order ambiguous. Second,
if it has to be changed, there are alternatives. The player may want
a fleet on another place or the player may want to waive the
build instead. Since this can not be concluded based on the given
order, I prefer that the build fails (choice a).
4.C.5. MISSING NATIONALITY IN SUPPORT ORDER
When the designation of the nationality of supported or convoyed foreign unit is omitted, the following could be decided:
Note that there are adjudicators that can not parse the nationality in support orders.
Specifying the nationality in support orders is a rule that is often
ignored. I prefer that the order is just valid (choice b).
4.C.6. WRONG NATIONALITY IN SUPPORT ORDER
When the designation of the nationality of supported or convoyed foreign unit is incorrect, the following could be decided:
Note that there are adjudicators that can not parse the nationality in support orders.
I think a best effort should be made. Therefore, the order should just
be valid (the second option). This means that the
nationality is in fact ignored (choice b). You can argue whether the nationality
has any use. However, if other parts of the order are unclear,
the nationality might help.
4.D TOO MANY AND TOO FEW ORDERS
4.D.1 MULTIPLE ORDER SETS WITH DEFINED ORDER
If a game is played by email or by post, the game master (or automated judging system) may receive multiple order sets. Of course, if there is a clear what the latest order set is, then the latest order set replaces any previous order sets. However, if the latest order set is incomplete, it could be handled in the following two ways:
In principle this is something that should be handled
in the house rules. However, when it is not handled in the
house rules, then I think the interests of both players
and judge are best served when in principle a new set
replaces an old set and only earlier sets are taken into
account when it is clear from the order set (choice c).
4.D.2. MULTIPLE ORDER SETS WITH UNDEFINED ORDER
Also in face to face games strange things can happen. When two sets of orders are submitted in a face to face game, then it is probably not clear which order set is the latest one. The following could be decided:
I think a best attempt should be made to interpret
the orders. Therefore, I prefer that all orders sets
are combined (choice b).
4.D.3. MULTIPLE ORDERS TO THE SAME UNIT
It might also occur that a player orders a unit twice or even more, with at least two orders that are not the same (if the orders are the same, then it should just follow that order). This situation is not covered by any rulebook and only in house rules.
First of all, it should be decided what should be treated as order. This is discussed in issue 4.E.1. Any illegal order should be ignored. That means that if a unit gets a legal and an illegal order, then there are in fact no multiple orders and only the legal order should be handled. This may sound theoretically, but this is exactly what can happen when someone swaps two names in a support order.
The following interpretations are possible:
Note that this is only an issue when the two orders are in the same order set. In case of multiple order sets see issue 4.D.1 and 4.D.2.
A player can have many units (especially in Colonial). In a face to face
game it might not be that clear which order is first (if multiple papers
are used or when the player uses columns). Therefore, I prefer the last
interpretation (choice c). However, for an automatic adjudicator program an
other solution might fit better with the user interface.
4.D.4. TOO MANY BUILD ORDERS
It might occur that a player orders too many builds. The rulebooks do not give a solution to this specific situation.
The following interpretations are possible:
Note that this is not an issue when the build orders are in different order sets. In case of multiple order sets see issue 4.D.1 and 4.D.2.
I have taken part in a face to face game where this happened.
France ordered three builds in 1901 while he was allowed to build
only two units. If all builds were denied, then it would not only
ruin the game for the player of France, but also for all other
players. Therefore, I prefer that the first legal orders are used
(choice b). Although, in case of an automatic adjudicator, a other
solution might fit better with the user interface.
4.D.5. MULTIPLE BUILD ORDERS FOR ONE AREA
Each area can have only one unit. This makes an order for building both fleet and army in one and the same area illegal.
The following interpretations are possible:
Note that this is not an issue when the build orders are in different order sets. In case of multiple order sets see issue 4.D.1 and 4.D.2.
Comparable to too many build orders, I prefer that
adjustments orders are just executed one by one. So, I prefer
that the first build order is used (choice b).
4.D.6. TOO MANY DISBAND ORDERS
Comparable with the situation that a user orders too many build orders, a player can also order to many disband orders. The different rulebooks do not give any answer how to handle such situation.
The following interpretations are possible:
Note that this is not an issue when the disband orders are in different order sets. In case of multiple order sets see issue 4.D.1 and 4.D.2.
I prefer that it is handled similar to the situation where too many
build orders are given. So, I prefer that the first legal disband orders are
used (choice b). Although, in case of an automatic adjudicator, a other
solution might fit better with the user interface.
4.D.7. WAIVING BUILDS
The 1971 and 1982 rulebooks are ambiguous about waiving builds. Two interpretations are possible:
Note that if the game is played by email or by post, it may required that the build is explicitly waived. In this way the waiving of builds can be distinguished from no orders received. This is not an issue in face to face games.
With the 2000 rulebook this is not really an issue anymore.
Allowing to waive builds is just a perfect rule. Therefore, I
prefer that builds can be waived (choice a).
4.D.8. REMOVING A UNIT IN CIVIL DISORDER
The rulebook specifies which unit must be removed when a country in civil loses a supply center. However, the way it is described allows multiple interpretations.
In the 1971 and 1982 rulebooks, the distance of each unit to the home supply centers must be calculated. The distance must be calculated by "the shortest available route, including convoys". This allows multiple interpretations. Must fleets be available for convoying? And must convoys be calculated as one move?
The 2000 rulebook is even more vague about this. The distance must be calculated to the "country". The term 'home supply centers' is not used. Taking the word 'country' literally means that also some provinces without supply center must be taken into account (for example, Apulia as part of Italy). Furthermore, information how the distance should be calculated is also less clear then the in the 1971 and 1982 rulebooks (the word 'convoy' is not used).
I interpret the 2000 rule just as the 1971 and 1982 rule. There was no reason to change this rule and the change in formulation was probably an attempt to simplify the formulation, with the consequence that the rule became just more vague. Taking country literally instead of home supply centers has some severe drawbacks. First it means that in some cases an unit has to be removed in a home supply center. Second, on some alternative maps the ownership of areas is not drawn. And finally, for automatic adjudicators it would require that they know the ownership of every area. This is currently not implemented and it will probably never be implemented for just the civil disorder rule. I consider the 2000 rules on this issue such a mess, that it is defendable decision to just take the rule from the 1971 or 1982 rulebook.
The following interpretations of 'moves' are possible when calculating the distance:
First it must be realized that it is sufficient to have an arbitrary result as long as it is a precise defined result. I think that choice a, b and c, in this respect, do not really qualify for this purpose. I think that the intent of the rule was choice d (as the DPTG interpretation) and therefore I prefer choice d. That means that the distance of fleets are calculated by only taking regularly moves, while armies may also move as a fleet.
There is also a complete different way of removing a unit (this is used in some house rules, but is not supported by any rulebook):
I do not prefer this alternative, because of the problems to apply this rule in face to face games.
If two distances are the same, then the alphabetical order of the areas becomes significant. This raises the issue of the language that should be used. For instance, a dispute may arise when a non-English board is used, but English versions of conference maps and rules were downloaded and printed from the internet.
I prefer that the English language is used by default. However, if the game is played face to face with one or more physical boards (conference maps and rulebooks not taken into account) and those boards are all in the same language, then I prefer that the language of those boards are used. In such case, a list of English names might not be available. If the game is not played face to face and played completely in an other language then English, then I still prefer that the English language is used. A computer program might have an user interface without any single English word, internally it still may use English names, for instance in a communication protocol. To avoid trouble here, the default language is English and the exception is only with face to face games. Furthermore, in a game that is not face to face, the availability of a list with English names is probably not an issue.
What to do with the alphabetical order when names contain special characters such as spaces, dots and hyphens?
This is probably never an issue. But if so, then I prefer that any special characters such as spaces, dots and hyphens are removed from the name, before determining the alphabetical order.
Finally, it should be mentioned that the full names should be used
in determining the alphabetical order and not the abbreviations.
4.D.9. RECEIVING HOLD SUPPORT IN CIVIL DISORDER
When a country is in civil disorder, all its units hold and do not support each other. But what should happen when an other country supports the hold of one of the units?
The following interpretations are possible:
I see no reason for introducing new issues for this rather insignificant
issue. Therefore, I prefer that the support succeeds (choice b). Also, in a face to face
game, a support to hold the armies, as an act of good will, might end the civil
disorder.
4.E. MISCELLANEOUS ISSUES
4.E.1 ILLEGAL ORDERS
"legal" is one of the most controversial terms in
the rules of Diplomacy. Of course, the same counts for
the opposite "illegal". This term has two aspects. First
its use and consequences and second the definition which
orders are legal and which are not. Both aspects can be
disputed. About the consequences the 2000 rulebook says
on page 3 (rather similar to the other rulebooks):
A legal order must be followed. An order written by mistake, if legal, must be followed. An "illegal" or ambiguous order or an order that is judged to be unsuccessful is not followed. A unit that is given an illegal order (or given no order) must stand in place. (The unit holds.)
In this document this is interpreted that to determine whether an order is legal or not is the first step in processing the orders. Any legal orders are added to the set of orders to be adjudicated, while the remaining is thrown away. That means that illegal orders do not have influence whatsoever.
This is considered a workable and sensible usage of the term 'legal', however some other documents do not use it this way. For instance, in the DPTG illegal orders may still have some effect (an illegal move can not be supported in a hold). It should be realized that with such usage the term 'legal' does not have much meaning other than that the order unconditionally fails.
Although it is now clear how the word 'legal' is used in this document, which orders are legal or not, must still be defined. The most significant question of this issue is, whether an impossible move (for instance A Moscow - York), can receive support to hold. Some argue that it is still a move and therefore the unit can not receive hold support. Other people say that it is an illegal order and those should be treated as a hold. This issue is not only about interpreting the orders, but it has a real consequence for the possibilities of the players. If impossible orders can receive hold support, then it means that a holding unit can not refuse hold support (by ordering an impossible move). This can lead to some interesting tactical situations. See for instance the Diplomatic Pouch, Winter 1995 Adjustments, Library of Diplomacy Tactics, Tactics Problems, Problem 3. However, it should be realized that the situation in the given problem is rather theoretical.
The legality of an order has also minor consequences. For instance, in case an order receives a legal and a illegal order. The illegal order must be ignored, so there is no problem by just accepting only the legal order (see also issues 4.D.3 and 4.D.4 and test case 6.D.28).
Finally, the legality of orders is very important in "no-press games", since illegal orders will be shown as hold orders to other players while legal attempts to move will be shown as they are ordered.
For listing the possible interpretations of 'legal', another term must be introduced. This is the concept of 'valid' orders. A valid order is an order for which a successful attempt can be executed, given all the other orders. More specific, non-matching support are invalid. Non-matching convoy orders or convoy orders that are not part of an ordered convoy route are also invalid. Even so, if a move order requires a convoy and no convoy path is ordered, then the move order is invalid. So, it is quite normal that during the game lot of orders turn out to be invalid. This terminology is consistent with the rulebooks. The 2000 rulebook uses the words 'valid' and 'invalid' first on page 7 in relation with the matching of orders.
Note, that in the definition of 'valid' the word 'attempt' is used. It is not necessary that the order actually succeeds or can succeed. If Portugal is occupied by an army and the Mid-Atlantic Ocean is empty, then a move order from Spain to Portugal can never succeed. However, that move order is perfectly valid.
The following interpretations of 'legality' are possible:
5. Q: What if you get these orders:
GER: A Mun-Mos, A Ruh S A Mun.
RUS: A Sil-Mun, A Boh S A Sil-Mun.
Does the Russian attack succeed?
A: Yes. Even though the German order was probably meant as a joke, it must be taken seriously, no matter how improbable. It's an order to move and a unit so ordered can't be supported in place -- so A Mun is dislodged.
Every so often a player will write a humorous or impossible order for a unit which is supposed to be holding. The question is, do you treat the order as an order to move? My inclination is to draw the line this way: an order to move to any space on the board, no matter how impossible, is an order to move within the meaning of Rule IX.6. This would include such orders as A Mun-Mos, A Mun-Switzerland, A Hol-Nth, F Pie-Ven, F StP(nc)-Mos, and so on.
On the other hand, if the unit is ordered to do some other thing, not involving a space on the board, I regard it as being ordered to hold, and it may be supported in place. Such orders might be: A Mun whistles "Dixie", A Mun-Oz, A Mun loot and pillage, A Mun-Mars (but not A Mun-Mar), and so on. See Spring 1914 in the Sample Game (Section IV) for another example.
If you wish to avoid such hair-splitting you can always prohibit nonsense orders, but that will take some of the fun out of the game.
Interpretation a has as disadvantage that almost every sentence that contains a hyphen must be treated as a legal order. For the implementation of a computer program this is a rather undesirable rule.
Interpretations a and b have as disadvantage that in fact three types of orders are distinguished. Legal orders that can be valid, illegal orders that are ignored completely and semi-legal orders that can't be valid but can still have some effect. This division in three types introduces an additional complexity in the settling of some disputes. For instance, in the DPTG, the group of semi-legal orders are called 'illegal'. This makes it unclear whether a unit that is ordered to move to the place it already stands in, can receive a hold support.
A disadvantage of interpretations a, b and c is that an user interface of a computer program may only allow the selection of orders that can be valid. If support can be refused by doing an impossible move, then such user interface should provide a feature to also enter an impossible move, because otherwise the player would be restricted in it is possibilities. This would be an awkward feature, for a rather insignificant issue.
The disadvantage of interpretation c and d is that it requires significantly additional coding when implementing a computer program for Diplomacy.
I consider it unimportant whether a unit can refuse
a support or not. The situations where this is significant are
extremely rare. Given the other arguments and the text of the
2000 rulebook I prefer option d. So, only orders that can
be valid in the current situation are legal.
4.E.2. POORLY WRITTEN ORDERS
Something that certainly will happen is that someone writes a poorly written order. The 1971 and 1982 rulebook say (page 3): "A badly written order, which nevertheless can have only one meaning, must be followed". The 2000 rulebook uses only a different term and formulation: "A poorly written order that has only one meaning must be followed".
First of all, it should be realized that some orders
are just ambiguous:
Example 1: A Bur - PirIn this example it is not clear whether the player wanted to go to Paris or Picardy. Therefore, the army in Burgundy should hold.
However, according to the rules, small spelling mistakes
must be accepted. If someone writes the following order:
Example 2: A Burgundy - ParizNo human judge would reject this order, because of the misspelling of 'Paris'.
On the other hand, unambiguous impossible orders are
never changed:
Example 3: Austria: Build F TyrAlthough 'Tyr' is close to 'Tri' and Austria maybe meant to build a fleet in Triest, this order must fail, because 'Tyr' is the correct abbreviation of 'Tyrolia'. The order to build in Tyrolia is not changed in a build in Triest. Of course, for new players some exceptions can be made.
Also, orders are never interpreted by looking to the orders
of another country:
Example 4: England: A Bre S A Bur - Par France A Bur - PirThe order of England can not clarify the order of France. Even for new players it is maybe better not to allow this.
Until now it was clear how the orders should be treated, however there is gray area where it is disputable whether an order should be accepted or rejected. One could choose one of the following principles when interpreting orders.
Example 5: F Rum - Bu.Since a fleet can not go to Budapest, this would be allowed. A move to the east coast of Bulgaria is attempted.
Example 6: A Gre S A Ser - Bu.Since Greece can not support a move to Budapest, this order must be interpreted as a move support from Serbia to Bulgaria.
Example 7: A Par S F Pir - Bre A Pir - Bre'Pir' can be interpreted as 'Picardy' or 'Paris', but since there is already an order for Paris, 'Pir' must be interpreted as 'Picardy'. So, the army in Picardy receives an order to move to Brest.
Example 8: A Ser - Bu A Gre S A Ser - BuThe order for the army in Greece must be interpreted as a move support from Serbia to Bulgaria, since a support to Budapest is not possible. Since, in this choice we assume that the player only wants to write matching orders, the order of Serbia is clarified by the order for Greece and must be interpreted as a move to Bulgaria.
I prefer that a best effort is made to interpret the order. So, I prefer the last interpretation (choice d).
Even with the above guidance, some orders are still disputable:
Example 9: F Rum - Bud(ec).Some people may reject this order, since 'Bud' is unambiguous and unambiguous orders are not changed. Other people would argue that the fact that a fleet can not go to Budapest, combined with the fact that the coast specification does only apply for Budapest, this order must be interpreted as a move to Bulgaria. The decision is for the judge or in case when there is no judge, in the mercy of the other players.
I think that allowing implicit orders encourages incomplete order sets and
that this will lead to more problems and errors. Therefore, I prefer that implicit
orders are not allowed (choice b).
4.E.4. PERPETUAL ORDERS
When a player wants to quit the game, he maybe wants to give 'Perpetual Orders'. That are orders that will be repeated every turn. For instance, when Italy has still two armies in Portugal and Spain, he may want to give the order that they will mutual support each other as long as it is possible. It can be decided that:
Although I do not really object against perpetual orders, I think allowing
perpetual orders should be decided before the game is started or all players
should agree during the game. If it was not explicitly allowed and a player
wants to give perpetual orders and another players opposes, then I prefer that
it is not allowed (choice b).
4.E.5. PROXY ORDERS
A 'Proxy Order' is an order that one does not order a specific unit by oneself, but that another specific player may give the actual order to the unit. I can be decided that:
A player that proposes to allow proxy orders, during the game,
may have a complex diplomatic plan. Therefore, I think proxy orders
should only be allowed when this has been decided before the game
started. Since, proxy orders are not according to mechanics of
order writing as described in the rulebook, they should not be
allowed by default (choice c).
4.E.6 FLYING DUTCHMAN
A 'Flying Dutchman' is a unit on the board that is illegal. This can be due to an adjudication error or cheating. Since it is not possible to replay previous rounds, the situation has to be corrected on the board. There are numerous ways to handle this situation (for instance, just play until the next adjustment turn).
Allan Calhamer wrote on this issue (in Diplomania, no. 12, August 1966, a copy can be found on 'www.diplomacy-archive.com'):
A variety of rough-and-ready tactics were developed at this time. One was the "Flying Dutchman", which consisted in playing with a piece to which you were not entitled. It was ruled that this practice was legal so long as it was a deception; i.e., any player had a right to demand restoration of the true position, but if moves had intervened, they could not be taken back. It was never clear what the rights were if the deception was discovered during a move.
I prefer that the unit is removed immediately according to the
following rules:
5. THE PROCESS OF ADJUDICATION
Writing a Diplomacy adjudicator program may look not more difficult than writing a program that checks the moves of a chess game. However, the contrary is true. A Diplomacy adjudicator that passes all test cases as described in this document contains many small and difficult details.
To help with writing an adjudicator program or just with the manual adjudication, the adjudication of Diplomacy is analyzed here. The first step in understanding the adjudication is to understand the principle that a set of orders leads to a set of decisions to be made. One order may lead to multiple decisions to made. For instance, when a unit is ordered to move, it can be decided that the move fails, but when the move uses a convoy, then it has also to be decided whether the unit has any influence on the area where it was ordered to move.
The second step is to understand that the decisions depend on each other. Certain decisions can only be made when other decisions are made first. For instance, when the units are ordered to follow each other in a move, then the decision of the unit moves at the end depends on the move decision of the unit at the front.
In the final step an algorithm is constructed. This algorithm is based on the decisions and respects their dependencies. In some other descriptions, such as the DPTG, the notion of "decision" is not described and the algorithm is described directly. This has as disadvantage that it gives no understanding of adjudication and why the algorithm is constructed as it is. Furthermore, with such description it is rather hard to look to alternative algorithms. We will see that there are at least two ways to tackle the problem of an adjudication program.
Note that in the adjudication description some sanity checks are left out.
For example a test whether a support order matches with the order of the unit
it supports. These are covered by the test cases and are not relevant for the
understanding of the process of adjudication.
5.A. OVERVIEW OF ADJUDICATION DECISIONS
There are eight different types of decisions. Each type of decisions is given a name, to refer easily to that decision type. To distinguish the decision type of the other text, the name is given in CAPITALS.
The first three decisions are directly related to the success or failure of an order:
Example: Italy: A Tyrolia - Trieste A Venice Supports A Tyrolia - Trieste Austria: F Trieste Hold
At the end of adjudication, the MOVE decision of Tyrolia is 'moves', the SUPPORT decision of Venice is 'given' and the DISLODGE decision of the army in Trieste is 'dislodged'.
When a unit tries to dislodge another unit, then the strength of the move must be calculated and if that strength is larger than the hold strength of the other unit, then the move of the unit succeeds. So, the following two decision types are needed:
Both ATTACK STRENGTH and HOLD STRENGTH are numerical decisions, because the result of the decision is a number.
In the previous example the ATTACK STRENGTH of the army in Tyrolia is two, while the HOLD STRENGTH of the army in Trieste is one.
A unit can not dislodge a unit of the same player. Also supports to a foreign unit can not be used to dislodge an own unit. However, these supports can be used to prevent that another unit enters the area. This strength can be greater than ATTACK STRENGTH and must also be calculated:
Example: Austria: A Vienna - Tyrolia A Tyrolia - Munich A Trieste Supports A Vienna - Tyrolia Germany: A Munich Supports A Venice - Tyrolia A Venice - Tyrolia
The ATTACK STRENGTH of the army in Vienna is zero, because it can not dislodge its own unit in Tyrolia. However, the PREVENT STRENGTH of the army in Vienna is two, which prevents that the move of the German army in Venice with an ATTACK STRENGTH of two succeeds.
When a unit is dislodged in a head to head battle, then the unit has no effect anymore on the area it was ordered to move. This means that the PREVENT STRENGTH is zero. However, it is still required to calculate the strength that prevents the opposite unit in the head to head battle to move. This value can be greater than ATTACK STRENGTH since all support has to be calculated. Therefore a separate numerical decision is necessary:
Example: France: A Belgium Supports A Burgundy - Ruhr A Holland Supports A Burgundy - Ruhr A Burgundy - Ruhr A Munich Supports A Ruhr - Burgundy A Marseilles - Burgundy Germany: A Ruhr - Burgundy
In this example the French army in Munich supports the move of the German army in Ruhr instead of the French army in Burgundy. This makes that the ATTACK STRENGTH, the PREVENT STRENGTH and the DEFEND STRENGTH of the German army in Ruhr are all different. The ATTACK STRENGTH is one, because the French support should not be counted for the attack. The PREVENT STRENGTH is zero, because it is dislodged by the French army in Burgundy and therefore it can not prevent the army in Marseilles to go to Burgundy. However, the DEFEND STRENGTH contains all supports and is therefore two. Still this DEFEND STRENGTH is insufficient in the head to head battle, since the French army in Burgundy has an ATTACK STRENGTH of three.
It is important to understand the difference between PREVENT STRENGTH and DEFEND STRENGTH. In some algorithms this is calculated by the same routine. The returned strength of such routine depends on the moment in the adjudication process. However, in this analysis decisions are defined independent from the moment in the adjudication process and therefore these two strengths needs to be distinguished.
Finally, when an army is ordered to move and the move will be convoyed, it has to be decided whether the convoy will succeed:
Example: England: A Yorkshire - Belgium F North Sea Convoys A Yorkshire - Belgium Germany: F Holland Supports F Denmark - North Sea F Denmark - North Sea
The fleet in the North Sea is dislodged, therefore the
PATH decision of the Yorkshire order is 'no path'. The PATH
decisions of non-convoying units are always 'path'. In this
case the PATH decisions of the moving fleet in Denmark is 'path'.
5.B. PRECISE DESCRIPTION OF MAKING DECISIONS
Before the description of how the different decisions must precisely be made, the notion of 'undecided' must be introduced. In the examples of the previous section the final result of decisions were given. But during the adjudication, decisions are made one by one and it must be possible to distinguish between decisions that are already made and decisions that are not made yet. This is simply achieved by introducing an 'undecided' status for the decisions that have two possible outcomes. So, a MOVE decision can have the status 'moves', 'fails' or 'undecided'. The decisions SUPPORT, DISLODGE and PATH decisions can also have the status 'undecided'. For the numerical decisions two values are administrated. The minimum values and the maximum values. The final result will be somewhere in the interval between the minimum and maximum value. During adjudication this interval will become smaller. The minimum values can only increase, while the maximum can only decrease. When the minimum and maximum value of a decision become equal, then the decision is fully made.
Initially all numerical decisions have a minimum value
of zero and a maximum value of infinity. The remaining
decisions are initially 'undecided'.
5.B.1. MOVE DECISION
This MOVE decision has only one 'fails' result. For proper
reporting of the adjudication result to the players, more
information might be appropriate. For instance, if a move fails
due to a move of another unit to the same area, this could be
reported as 'bounce'. Also the exact bouncing unit might be
listed in the report. This can be implemented by introducing
different 'fails' as result of the MOVE decision result. Or
this information could be part of the 'fails' result. Anyway,
these ways of failure should be treated the same in other
parts of the adjudication.
5.B.2. SUPPORT DECISION
Note that the PREVENT STRENGTH become zero when the unit is engaged in a head to head battle and the MOVE decision of the opposing unit has status 'moves'. If the condition that the unit is engaged in a head to head battle, is left out, then this may lead to a situation where two units end in the same area (see test cases 6.G.16, 6.G.17 and 6.G.18).
If issue 4.A.7 must be adjudicated
according to choice a, then it should not be checked
whether the unit is engaged in a head to head battle.
Instead the PREVENT STRENGTH is zero when the MOVE
decision of the same unit has status 'fails' and the
MOVE decision of the opposing unit has status 'moves'.
5.B.7. DEFEND STRENGTH DECISION
Note that calculating the PATH decision does not require
that a search algorithm is started twice. Probably a
depth-first search algorithm is the easiest way for
programming. In such algorithm a search function is
called recursively. Normally, such recursive function
would only have two possible outcomes, 'failure' or
'success'. In this case it has three possible outcomes.
The recursive function first checks whether it is on
the destination and if so, the it returns 'sustains'.
Then it checks whether it has already been on the area.
If so, then it returns immediately 'no path'. In all
other cases the result of all sub-paths (calculated
recursively) is combined with the current DISLODGE
status of the area.
5.B.9. CIRCULAR MOVEMENT AND PARADOXES
It is possible that during adjudication according to the given prescriptions of making decisions, a situation occurs that no decision can be made anymore while there are still decisions to make. At such moment the board contains a circular movement or convoy paradox. Those situations need special treatment.
The first step in resolving a circular movement or paradox is to determine which decisions are part of the circular movement or the paradox core. This is achieved by listing for each decision the decisions it directly depends on. Decisions that are fully decided are completely left out of this procedure. When these lists are made, these lists are extended with all the indirect decisions dependencies. This is continued until each decision has a list with all direct and indirect (with possible multiple indirections) dependencies. Every list forms a paradox, but the smallest dependency list is a paradox core and only that list should be considered. Note that there are probably more than one dependency lists with an equal and smallest number of dependencies. It is unimportant which one of these lists is selected.
In the second step the dependency list is analyzed. If the list of decisions contains an undecided MOVE decision of a unit that targets a fleet with a convoy order, then there is a convoy disruption paradox. If there is no such decision, then there is a circular movement. Note that if one of the moving units in a circular movement moved with certainty due to a support, then the adjudication of the circular movement would have been resolved in the normal process adjudication.
In the final step the circular movement or paradox is enforced by special rules and the normal decision making process can continue. In case of a circular movement, all the MOVE decisions of the dependency list are resolved to 'moves'. In case of convoy disruption paradox, a convoy paradox rule must be applied on the dependency list. Note that the MOVE decision of the army that convoys is not in the dependency list, since for the paradox only the cutting of support is essential. Therefore only the ATTACK STRENGTH decision of the army that convoys appears in the dependency list. This is important when applying the Szykman rule or the 'All Hold' rule.
When the Szykman rule is applied, all ATTACK STRENGTH decisions in the dependency list are set to zero for both minimum as maximum. The corresponding MOVE decision is set to failed and the corresponding PREVENT STRENGTH is also to zero for both minimum as maximum.
If you interpret the 2000 rulebook in such that in some very rare cases the attacked unit is dislodged by the convoying army (see discussion in issue 4.A.2, and test case 6.F.17), then first the dependency list must be searched for a SUPPORT decision of a support order of an attack on a convoying fleet that convoys an army to the area of the supporting unit. That SUPPORT decision must be set to 'given'. If no such decision could be found, then the 2000 rulebook has no resolution and a fallback rule must be used such as the Szykman rule or the 'All Hold' rule.
Variant rules may introduce new kind of paradoxes
(see test cases 9.E, 9.F
and 9.G). If those rules are included,
then it is dangerous to conclude that there is a
circular movement in case the dependency list does
not contain a convoy disruption paradox. It is
better to check if the moving units of all the
MOVE decisions of the dependency list are part
of one single circular movement. If so, these
units advance, but if not then a fallback paradox
rule must be applied. In this fallback scenario
all the MOVE decisions of the dependency list
resolve in 'fails' and all the SUPPORT decisions
of the list resolve in 'cut'.
5.B.10. HEAD TO HEAD BATTLES DEPENDENT ON ADJUDICATION
In most interpretations of the rules, it can directly determined whether two units that move in opposite direction will fight a head to head battle or try to swap (by using a convoy). This can be decided before any adjudication. However, if issue 4.A.3 is interpreted according to choice c, then the convoy route to the adjacent place must be adjudicated before it can be decided whether the moves will result in a head to head battle.
The adjudication of this has some additional complexities. It can be implemented by introducing a new decision 'HEAD TO HEAD BATTLE'. This decision should have as result whether two units that move in opposite direction, will fight a head to head battle or not. The MOVE decision, the ATTACK STRENGTH decision and the PREVENT STRENGTH decision must be changed such that they use the HEAD TO HEAD BATTLE decision.
For the implementation of the HEAD TO HEAD BATTLE decision, the
'CONVOY PATH' decision must be introduced. When there is a successful
convoy path for at least one of the opposing units, then there is
no head to head battle, otherwise the units will fight a head to
head battle. The CONVOY PATH decision is similar to the PATH
decision, but will only result in 'path' when the unit can move
to the destination by using at least one convoying fleet. To
avoid duplication of code, the implementation of the PATH
decision can be simplified by using the CONVOY PATH decision.
5.B.11. RETREAT
To which areas a unit may retreat, must be calculated after adjudicating the main phase, because it depends on the orders given in the main phase. It is insufficient to just mark the areas to which a retreated is possible, since there are situation where one unit may retreat to a certain, while another may not (see test case 6.H.10 ). So, for every dislodged unit a list of areas must be created to which it may retreat.
How a retreat must be calculated is based on the principle
as worded in the rulebook: "A dislodged unit must retreat to
an adjacent province that it could ordinarily move to if
unopposed by other units." That means that the logic of a
retreat is similar to the MOVE decision. A unit may retreat
to an area if the unit may move to the area directly (so,
without convoy), the PREVENT STRENGTH of all units that
attempted to move to the area is zero, the HOLD STRENGTH of
the area is zero and there is no unit that moved successfully
in the opposite direction. All these calculations must be
made as part of the adjudication of the main phase.
5.C. FROM DECISIONS TO ALGORITHM
In the final step the principle of decisions is transformed into an algorithm. Whatever for algorithm is made, the decisions will always be part of the algorithm one way or another. This is for the simple reason that the decisions are directly based on the rules. However, there are two fundamental different ways to deal with the dependencies of the decisions:
In a sequence based algorithm, the program tackles the problem of dependencies by making the decisions in a hard programmed sequence. The sequence guarantees that when a decisions is made, any other dependent decisions are already decided. For instance, the algorithm will probably start by cutting supports. Then it will try to resolve moves that will dislodge a convoying fleet and finally it will resolve the moves. If a unit moves to an area where a second unit is ordered to move, it will first resolve the move of the second unit. This can be implemented by a recursive function.
The advantage of the sequence based algorithm is that it does not need to have the administration of all the decisions. The decisions do not require an 'undecided' state. A support decision is 'given' until it is decided to be 'cut'. The sequence in which the decisions are made, will guarantee that this will lead to the correct results. It is not too difficult to make a sequence based algorithm that handles the most common situations correctly. However, for a perfect adjudicator that passes all test cases in this document and that handles second order paradoxes according to the Szykman rule, a sequence based algorithm becomes very complex. The DPTG is an example of a sequence based algorithm and it shows the difficulties of handling convoy paradoxes in a sequence based algorithm.
The alternative is the decision based algorithm. A decision based algorithm starts with setting up the administration of the decisions based on the orders. Initially all decisions are undecided. Then the algorithm tries to make decisions one by one without paying attention to the sequence. If a decision can not be made due to a dependency on another decision, then the algorithm just goes to the next decision. When all decisions are handled once this way, it just starts all over again. It will repeat this until all decisions are made. It is also possible that there are still decisions to be made, but none of them can be decided. The algorithm detects it, resolved the circular movement or applies a paradox rule on the remaining decisions and restarts the decision process.
The key principle in a decision based algorithm, is that when a decision is made, then that decision is final. This principle is necessary because of the undefined sequence in which decisions are made. If a decision would be changed, then a second decision that is based on that decision would have a result dependent if the decision would be made before or after the change. So, in such case the result becomes dependent on the (undefined) sequence and therefore there is no guarantee that the result will be correct.
The advantage of a decision based algorithm above a sequence based algorithm