Diplomacy Zine -- EP #212 Chapter Seven From: Eric_S_Klien@cup.portal.com Date: Sat, 10 Nov 1990 00:27:49 +0000 Issue #212 of ELECTRONIC PROTOCOL: **************************************************************************** "Read my lips, no new taxes!" **************************************************************************** Chapter One contains: BAGHDAD, AUSTERLITZ, BLITZKRIEG, KING'S GAMBIT, PASSCHENDAELE, GET SOME, DRAGONS, BLACK OCTOBER And is published by uunet!cti1!rlister or rlister@cti.com/Russ Lister Chapter Two contains: REPUBLIC, BORODINO, JACAL, VERSAILLES, DRESDEN, KHAN And is published by sinhaa@mcmaster.ca/Anand Sinha Chapter Three contains: DAWN PATROL, BERLIN, EL ALAMEIN, SQUALANE, UNGAWE, BRUSILOV OFFENSIVE, CULLODEN, GANDALF'S REVENGE, GOODBYE BLUE SKY And is published by cwekx@htikub5.bitnet/Constantijn Wekx Chapter Four contains: OZARK, DEADLY DAGGERS, YORKTOWN, MONTREUIL-SUR-MER, FIRE WHEN READY, THUNDERDOME And is published by dm8sstaf@miamiu.bitnet/Douglas M. MacFarlane Chapter Five contains: DEF CON 5, BORDEL, ERIS, MASADA, YALTA And is published by jjcarette@watami.waterloo.edu/David Gibbs Chapter Six contains: TOKUGAWA, BERLIN WALL, HIROSHIMA, GENGHIS KHAN, SEA LION, VIOLENT PEACE And is published by ps9zrhmc@miamiu.bitnet/Peter Sweeney Chapter Seven contains: HELM'S DEEP, GROUND ZERO, GIBRALTAR, TIBERIUS, BETELGEUSE ------------- Chapter Seven ------------- Table of Contents: Thoughts on Programming a Diplomat By Michael R. Hall -- Part Two. And a bunch of letters. ---- Thoughts on Programming a Diplomat By Michael R. Hall Part Two So, finally we have several promising full candidate set of orders, the best of which is A Bul-Con, F Ank-Bla, A Smy-Arm. (I am pleased that it arrived first at one of the best possible sets of orders for Turkey. This was not rigged!) Since this is at the top of the agenda, it then looks at the enemy units that could have some effect on the outcome - F Sev, A Mos, F Tri, A Bud, F Nap. It tries possible the orders for these enemy units. It evaluates the possible final states, and gets some points knocked off for the possible Russian presence in Black Sea as well as the advances of Austria and Italy. Since this is the only set of orders thus far that have been tested against enemy orders, it is considered the current best set of submittable orders. If it has extra time for processing, then it would try the next item on the agenda, namely {A Con-Bul,F Ank-Bla,A Smy H}=1014. This is that special element that got put on the agenda when {A Con-Bul,F Ank-Bla}=1014 was exploded into its successors. It is tested against enemy orders, and it gets even more points knocked off for the possible Russian advance. Since this is not better than the current best set of submittable orders, it is discarded. If time remains, it would examine the state {A Con-Bul,F Ank-Bla, F Smy-Syr}=1014 against enemy orders, and presumably discard it. Next, it would come to {A Con-Bul, F Ank-Con}, and this represents a jump to a different part of the search tree. It would look at all the states generated by the orders for A Smy in this context. And so on until it ran out of time, at which point the best set of moves that have been tested against enemy moves would be submitted. Machine learning techniques will probably be necessary to lift the tactics of computer Diplomacy players up to a level competitive with novice human players. For example, base values of territories could be guessed at by humans and then optimized by machine learning techniques. Perhaps this will be the subject of another article. Well, I'm about done, except that I haven't mentioned *diplomacy* - the name of the game, after all. Danny Loeb has developed a protocol for communications between computer diplomats, so it is in theory possible to program diplomatic techniques. Most of the above still applies, except that the holdings of an alliance should be considered together in the evaluation of "altitude controlled", plus some catches need to be put in to prevent accidental stabs. The diplomat should probably work out two sets of orders each turn - one in the case that it remains an ally and one in the case that it stabs. If the predicted profits for stabbing exceed its estimated benefits for not stabbing, then it stabs. Similarly, you might want to have the system work out two sets of the opponent's orders (one he stays allied, one he stabs) for each set of yours, and evaluate the outcomes probabilistically, assuming a certain chance of a stab. This would allow the program to work with an ally while protecting itself without stabbing its ally. Strategic planning and diplomatic maneuvers will require a high-level "cognitive" component with a fair amount of smarts. Some other AI techniques other than best-first search will be necessary for this level of cognition. Production systems provide an easy way to toss a bunch of heuristic rules together to yield quasi-intelligent behavior. Truth maintenance systems supply a means of remembering deductions that production systems have made and maintaining a set of beliefs and disbeliefs. For example, if a diplomat is stabbed by a player it believed was its ally, it should remember to nullify anything (such as DMZ's) that even indirectly depended on the alliance (unless it can find some other justification for these things.) Here are some sources that may be of assistance to those interested in learning more about these AI techniques: * "Artificial Intelligence" by Elaine Rich (1983/1989) - the 1983 edition is a bit "dated" (being a minor revision of 70's editions) and it goes into a lot more detail on search than the newer texts, which generally concentrate on more recently developed AI techniques. I have not seen the 1989 edition. * "Artificial Intelligence Programming" by Charnkiak, Riesbeck, McDermott, and Meehan (2nd edition 1987) - this is a newer text that nevertheless gives a lot of nuts and bolts implementation details on best-first search, as well as production systems and truth maintenance systems. To get the most out of it, you will have to already either know Lisp or else learn Lisp by reading the first few chapters. * "Introduction to Artificial Intelligence" by Charnkiak and McDermott (1985) - this has a section on game trees, which may not be all that relavent to Diplomacy because of the simultaneous moves. In a later chapter it discusses how planning could be applied to the games of chess and Risk. * "Machine Learning: Applications in Expert Systems and Information Retrieval" by Forsyth and Rada (1986) - this is an introductory text on machine learning techniques. There are very few introductory texts for this area of AI. In summary, I believe that efficient search coupled with a good state evaluation function will allow a computer diplomat to devise decent tactics. While for humans, diplomacy is often more important than tactics, I think computer diplomats will be better off with good tactics before good diplomacy, because it is much harder for programs (written perhaps by different people) to communicate with each other (and then there's the problems of cooperating at strategic and tactical levels, not being too gullible, knowing when to stab, etc.) Good luck. ------------------------------------------------------------ Below is a selection of my discussion with Michael Hall on his article. Comments wholy incorporated into his article have been removed. 1) Comments by Daniel Loeb: Your article was very good. But here are a few more detailed comments. Diplomacy. However, we need not throw up our hands in dismay; the field of artificial intelligence (AI) has at least some of the answers. I think what you are talking about is really Game Theory and not AI, but that is just a matter of definition. The classic core of AI programs is search. Search is merely the I consider the core of AI to be pattern matching and learning. results. Another possibility is to assume that the opponent will do the orders that maximize the opponent's state evaluation, given your orders.) This is what is called MIN-MAX. That's the basic approach that I'm proposing. However, the Note that your program is going to be paranoid as all hell. But it sounds like it will be very game-smart. simultaneously - since each unit has at least four possible orders, there's over 4^(3*6+4)= 17,592,186,044,416 nodes in the full search tree. Thus, search needs to be done intelligently. Why don't you compute the exact number of moves. Actually I take that back. This would make a good contest..... (See Electronic Protocol for the contest...) Diplomat.) Second, as Danny Loeb suggested, we might view the game at more abstract levels, with territories grouped together into provinces, and perhaps provinces grouped together into larger regions, and so on. Thus, if the system for some reason decided Perhaps you should include my division of the world. I carefully thought it out so that the connectivity of the sections would be trivial. The world is cut up according to its accessibility by fleet and army. You will probably want to store "book openings" to give your program a little head start, but let's walk through how Turkey Question is how to chose openings. Do you have only one opening per country or do you chose them randomly. You have no game history on which to base potential alliances nor is anyone likely to declare war before 1901. In summary good article, do you mind if I send it to the DPP list as well. 2) Responses to my comments: >I think what you are talking about is really Game Theory and not AI, >but that is just a matter of definition. Everyone tries to take the stuff that is understood well and works out of AI. I'll mention both game theory and AI. >I consider the core of AI to be pattern matching and learning. Ah, but there is search in both pattern matching and learning. >Note that your program is going to be paranoid as all hell. But it >sounds like it will be very game-smart. I would imagine that it might be rather conservative, but a survivor and a long term threat to the other diplomats. The probabilistic approach that I mentioned would be better - if it could generate, say, 10 likely sets of moves of the opponents and then evaluate its own moves against them, then it could come up with the expected value of each of its sets of moves. This would allow it to come up with very robust sets of orders that work well almost no matter what the opponent does, but it would also allow it to sometimes take some risks. >Question is how to chose openings. Do you have only one opening per >country or do you chose them randomly. You have no game history on >which to base potential alliances nor is anyone likely to declare war >before 1901. I'm not going to address this in my article. A near-optimal approach would simply be to always use one of the Gamer's Guide openings - the same one every time would be fine, and it would probably make it easier to tune your program. For example, if England always opens anti-French, then you can tune the English diplomat to build some armies and get ashore in France and then attack Germany in the midgame. >In summary good article, do you mind if I send it to the DPP list as >well. Of course you may, but let me revise it first. I also have some thoughts on the higher level architecture. I'd like to have a production system filled with rules, plus a truth maintenance system to record the deductions. For example, the system might set up a DMZ with an ally, but then if the ally attacks the system, then it should nullify the alliance, and it should also remember that the DMZ was based on the existance of the alliance and so it should nullify the DMZ too, which could cause other deductions to get "undone" and so on. I'm probably going to program all this in LISP. This architectural stuff won't go into the article, though, because the article is intended to be much more of a "this is how things *should* be done" article than a "this is how things *could* be done" article. 4) My reply to his reply to my comments: I'm taking out the Syr/Tun example and replacing it with a tree filled with the letters A-J, so that I can explain breadth-first versus depth-first with an example. I might make an indented tree to show visually what's going on with the example of Turkey's opening moves. I'm taking out the hold orders, because if a piece is not given an order, then it is assumed to hold. The indentation idea is good. But an abstract example with A-J's is too theoretical. Why not discuss a simple game like Hackenbush (see "Winning Ways") or tic-tac-toe. Instead of showing what is going on with Turkey, try another example (not necessarily Spring 1901) inwhich a convoy is at least considered. After reading my strategy, you may realize that I consider when to convoy a difficult point in the strategy. I'll also give a reference to an AI book for those who want more information. The best book I know of for explaining search is Elaine Rich's "Artificial Intelligence", but it's a bit dated. Do you know of any more current sources that take the time to go through search in detail? There just was an article in (the French edition of) Scientific American on Chess programs, and that pretty much applies here. Most of what you say applies to 2-player games as well (and those are well understood). You should clearly emphasise the parts (stabbing and negotiation) which are new. How well does a diplomat need to know the rules of diplomacy? Writting a perfect adjudicator is a challenge in its own right (although nothing amazing theoretically). In my opinion, a Diplomat need only be aware of how supports help units succeed in what they do. The rules on cutting support and so on may lie above the level of detail we'll be working on. >I consider the core of AI to be pattern matching and learning. Ah, but there is search in both pattern matching and learning. There is also a search-feature in any editor, but doesn't make emacs an artificial intelligence program in my opinion. Actually, I hate arguing definitions. Let's leave it at that then. I would imagine that it might be rather conservative, but a survivor and a long term threat to the other diplomats. The probabilistic approach that I mentioned would be better - if it could generate, say, 10 likely sets of moves of the opponents and then evaluate its own moves against them, then it could come up with the expected value of each of its sets of moves. This would allow it to come up with very robust sets of orders that work well almost no matter what the opponent does, but it would also allow it to sometimes take some risks. It sounds to me as if you are describing yourself and not your program. Actually, the main differences is that your program will have a horizon of LESS than 1 turn look ahead (ignoring retreats and builds). And secondly, you personally engage in mock-wars. These wars might work well on computers since they might conclude two countries are at war as a result of any conflict (or more conservatively any exchange of supply centers). I think a program should generate a list of promises made by (and to) each other diplomat. (These promises would also include rumors.) There also should be a list of "default" promises for anybody who says he is an ally but doesn't give details. And even a set of default promises for people who don't declare war. Each set of promises would include an idea of how much they should be lowered on the "ally-enemy" scale for breaking it. In addition, diplomats should be raised for saying nice things (this raise would then be overcompensated by a lowering if these were just sweet nothings). Even saying mean things before attacking should in some way mitigate the penalty for attacking. This set of incentives can be used even before Spring 1901 to help determine moves. >Question is how to chose openings. Do you have only one opening per >country or do you chose them randomly. You have no game history on >which to base potential alliances nor is anyone likely to declare war >before 1901. I'm not going to address this in my article. A near-optimal approach would simply be to always use one of the Gamer's Guide openings - the same one every time would be fine, and it would probably make it easier to tune your program. For example, if England always opens anti-French, then you can tune the English diplomat to build some armies and get ashore in France and then attack Germany in the midgame. I agree that your way is a good way to start. Saves time for the programmer. But the finished product should be able to attack any of its neighbors. Otherwise, it will be outguessed even in Spring 1901. You ignore another way to save yourself some time (while developing your first version). Write a program which only knows the openings for one country. (Make sure nobody else is writting a similar program for the same country though!) >In summary good article, do you mind if I send it to the DPP list as >well. Of course you may, but let me revise it first. When I send it to DPP, I will include our discussion as an appendix since it might contain useful ideas that couldn't fit into the paper. I also have some thoughts on the higher level architecture. I'd like to have a production system filled with rules, plus a truth maintenance system to record the deductions. For example, the system might set up a DMZ with an ally, but then if the ally attacks the system, then it should nullify the alliance, and it should also remember that the DMZ was based on the existance of the alliance and so it should nullify the DMZ too, which could cause other deductions to get "undone" and so on. This is pretty much what I was talking about above. However, you should take into account the severity of any infraction. In Tango as France, I once attacked Italy for making his Winter 1901 build in Naples instead of in Venice as required (as a tiny clause in a big F-I-T treaty which he never read carefully). Normally, a diplomat should ignore indiscretions like that, but in Tango I was in a really unique situation where Germany had build so many fleets that I knew that England couldn't afford to tackle me before dealing with him, yet the two of them were in a major war with Russia. This gave me the freedom to hold Italy to the letter of the law. (This is my sole email win.) If my situation had not been so perfect, I would never have thought of attacking Italy in 1902! Thus, the DMZ agreement would not have been nullified despite Italy's technical violation. I'm probably going to program all this in LISP. This As long as you want a LISP like language why not program it in SASL. Advantages: 1) you get to communicate with my program through "ports" instead of std_in and std_out. 2) you can run programs in parallel. For example, you could run a) a master coordinating program which consults with the others and with my program, b) a program for each country which calculates its best moves AS A FUNCTION of the current best move output by each of the other such programs, c) a program for each other country which handles negotiations with that country, and also keeps track of all promises, and so on. 3) you can utilize certain of my functions (although not all, since that would be cheating in certain cases). architectural stuff won't go into the article, though, because the article is intended to be much more of a "this is how things *should* be done" article than a "this is how things *could* be done" article. Well, do write down your ideas on that matter as well for publication uniquely in DPP. Actually, our discussion might serve that purpose fairly well. Yours, Daniel Loeb "Combinatorics: The Thought that Counts" Here is a letter from loeb@geocub.greco-prog.fr/Daniel Loeb: I am ready to start new diplomacy games. They will be run off of the Diplomacy Adjudicator (judge@milton.u.washington.edu), and I will be the GM. The names are BEREZINA, CANNES, DIEN, and EYLAU to start us off. As usual I am running NoNMR rules and complying in all ways with your house rules as I modified them. The variants available are: standard, pure, youngstown, loeb9 Games played on any other map can be set up on demand. Players should be instructed to: 1) contact me personally 2) send the following messages to judge@milton.u.washington.edu HELP GET HOUSE.RULES GET FORM 3) the form should be filled in and returned to judge 4) the following final command should be sent to judge SIGNON ?gamename password Players will be informed of their country assignment after the final player has signed on. Observes can reduce all the above to: 1) send the command OBSERVE gamename password Yours, Daniel Loeb "Combinatorics: The Thought that Counts" Here is a letter from loeb@geocub.greco-prog.fr/Daniel Loeb in response to the following letter from EP #204: Well, there is a Diplomacy variant that I would like to test: all "neutral" supply centers are one-army or one-fleet Minor Powers, each owned by a separate player. Do you think it's feasible? I'd need 19 players! I don't think I have the time to run it right now, but maybe in a month or so I can get my schedule under control. -- Floyd McWilliams 919 Hillsdale Blvd. (415) 358-7335 The response: I have just one comment about your variant. The minor countries will certainly make the game more interesting for the 7 real players, but they would be extremely tedious to play themselves. Starting with just say Bulgaria I'm likely to be eliminated right away, and even if I am not... my builds are limitted to one unit per year. Thus, it is impossible for me to win before 1918! By 1918 I am sure everyone will have lost interest in the game. This sort of variant has been thought of before, but not successfully. Keep trying. The basic idea is good. Yours, Daniel Loeb "Combinatorics: The Thought that Counts" Floyd McWilliams/fmcwilli@oracle.com responded: Since a minor power starts with one-third as many supply centers as a major power, it should need one-third as many supply centers for a victory, i.e., six. This leads to the difficulty of simultaneous victories, though. Perhaps there should be separate victory categories: Major Power Victories and Minor Power Victories. Thus, a major power's win is not "tainted" by a simultaneous minor power win. This gives major powers and minor powers an incentive to cooperate, since they can help each other win. And if more than one minor power can claim a victory, that's fine too. Eric, feel free to publish our discussion if you wish. Loeb@geocub.greco-prog.fr/Daniel Loeb responded: Dear Floyd, I have asked Eric Klien to start signups for a game similar to the one you suggested, but with the following changes: 1) In order to maintain Game Balance, there are 34 players. All running a separate supply center. 2) In ordinary diplomacy, it is possible to attain maximum size by Winter 1905 (1904 for Russia). However, if you play with only one home supply center then it takes until Winter 1917 to do so! In order to maintain interest in the game, you need the possibility of fast growth. (Also the player in Serbia shouldn't be limitted to armies for the whole game!) My suggested rule change is that players can build in any of their empty supply center (regardless of whether it is "home") You suggested change to the victory condition is unclear and in my opinion unnecessary. Since a minor power starts with one-third as many supply centers as a major power, it should need one-third as many supply centers for a victory, i.e., six. This leads to the difficulty of simultaneous victories, though. Perhaps there should be separate victory categories: Major Power Victories and Minor Power Victories. Thus, a major power's win is not "tainted" by a simultaneous minor power win. This gives major powers and minor powers an incentive to cooperate, since they can help each other win. And if more than one minor power can claim a victory, that's fine too. Does that mean that the game ends when a small power reaches 6! If I am England with 17 supply center, I am not going to be happy if the game ends because Greece owns (Ser,Bul,Con,Ank,Smy)! In diplomacy, only one player can win. The interest in the game is the paradox that you can't win without cooperating with people while at the same time the only people with whom you can cooperate are out to defeat you. By changing the victory condition, in the way you propose you are undermining the very spirit of the game. Incidentally, Ken Lowe says it should be possible to have the Diplomacy Adjudicator assist the play of such a game. I would be willing to be game master, but since this is really your idea, you can be listed game master by the Diplomacy Adjudicator (judge@milton.u.washington.edu) and I will be a player. Yours, Daniel Loeb "Combinatorics: The Thought that Counts" Here is a letter from loeb@geocub.greco-prog.fr/Daniel Loeb: 1) Eric, Michael Hall is sending you an article (currently under preperation) on diplomacy automatons (diplomats) relevant to the DPP project. His estimate of the number of possible openings is not exact, but would make an interesting contest. Here therefore is the contest: How many different sets of openings are the in Spring 1901? Remember to take into account all the orders for all 22 units. You may not leave any units unordered. Moreover, any supports ordered may not be void. That is to say, we only count combinations inwhich a unit makes the move that it was supported to make, or remains in place as it was supported to do. 2) Eric, Here is Jamie's solution to an old problem. Date: Wed, 10 Oct 90 12:24:06 EDT From: Jamie <PL436000@brownvm.brown.edu> Another nice problem! I have worked out my solution. Do you have a solution? My answer is that the map is NOT 3-colorable, even if one separates the problem for armies and for fleets. This much was easy. 3) Eric, Here is a problem by Jamie, and my solution: I offer a subproblem (or two, depending on how you look at it). What are the smallest neighborhoods which cannot be 3-colored for armies? For fleets? I believe I have a complete solution to the subproblem. I'll send it along later, but I'm wondering what you have worked out. Jamie ANSWER: In either case you must have at least one space which is completely surrounded, since otherwise the map contains no simple loops of size greater than 3, and is therefore 3 colorable. Therefore, we must look for spaces off the edge of the map which contain a very small odd number of neighbors. In either case, coastal areas need not be considered. For fleets: there are several seas with 5 neighbors but none with 3. (ADR) ALB APU VEN TRI ION (BOT) SWE FIN STP LVN BAL (IRI) NAO LVP WAL ENG MAO For armies, there are several inland provinces with 5 neighbors but none with 3. (BOH) MUN SIL GAL VIE TYR (BUD) VIE GAL RUM SER TRI (MOS) STP LVN WAR UKR SEV (RUH) BUR BEL HOL KIE MUN (UKR) RUM GAL WAR MOS SEV (VIE) TYR BOH GAL BUD TRI Together each of these sets of six spaces forms an answer to the problem. Yours, Daniel Loeb "Combinatorics: The Thought that Counts" Here is a letter from loeb@geocub.greco-prog.fr/Daniel Loeb: To: ken@milton.u.washington.edu Cc: eric_s_klien@cup.portal.com Subject: 34 player variant Reply-To: loeb@geocub.greco-prog.fr Given that your program refers to countries by first letter, I hesitate to ask you, but do you think it would be possible to write a 34 player variant. (Code letters could be ABCDEFGHIJKLNPQRSTUVWXYZ0123456789.) In Electronic Protocol, someone proposed a variant in which all the neutral SC are played. His rules are lacking in 2 respects. 1) Game-balance. But this is easily solved by chopping up the major powers as well. 2) Growth. In Diplomacy, a country can grow very quickly. Building 3 units per year, one can reach maximize size by the end of 1905. (1904 for Russia). In this game, it would take until the end of 1917! Instead of leaving the game as it is, I propose a few possibilities: a - eliminate the rule restricting players to build in their HOME supply centers. Allow builds in any empty supply center they own. b - allow players to select (secretly?) before Spring 1901 a certain number of SC's to be their home. c - say that the first so-many SC's capture by a player can serve as his home. To maximize playability, game should begin with a build phase in which each player places one unit (of his choice) on the board. Yours, Daniel Loeb "Combinatorics: The Thought that Counts" Here is a letter from loeb@geocub.greco-prog.fr/Daniel Loeb: Eric, 1) Ken tells me that since developing the NoNMR option he has started some NoNMR games (which I have nothing whatsoever to do with), perhaps you would like to list his games. This is what I told Ken 1) I think Eric would be interested in any game which is NoNMR now, and in addition had no (NO ORDER RECEIVED)s during the game history. Please send him the name, player list, and HISTORY of all such games. Ask him to place those games in your zine. He (and the chapter publisher) should then signon as OBSERVER (or MASTER). At that point, you would then be able to benefit from Eric's substantial backup player list. In addition, the records of those games would join those of MANY other email diplomacy games in his archives. Also, would it be ok, if Ken uses my survey to find out what the players in his games feel about DA? 2) In the instructions I told you to send all future players (for the DA). I would like to make corrections: > Players should be instructed to: > 1) contact me personally > 2) send the following message to judge@milton.u.washington.edu > GET HOUSE.RULES in order to receive the D.A. variant of the E.P. house rules. 3) send the following message to the same address HELP You should then receive a file of helpful information. This file will include a form which you should fill in and return to the judge > 4) the following final command should be sent to judge > SIGNON ?gamename password Finally, you should note that the use of passwords by "observers" is optional (but recommended). 3) Ken Lowe says it would be possible to have D.A. run the 34-player variant based on the minor power idea. The variant is played on the standard map with the following rule changes: a) every SC is owned by a different player b) the game begins with WINTER 1900 inwhich the players each build their one unit c) players are allowed to build in any of their empty supply center during Winter turns. (Instead of just home SCs). Could you please start taking sign ups for this variant. And please really PUSH this variant. Without your ACTIVE support, I'll NEVER find 34 players. Place an advertisement in E.P. for me, and put this game at the BEGINNING (or at least near it) of your list of games in formation along with the short explanation above. I anticipate that a game of this size might constitute a chapter in its own right. 4) Here are my games on the DA in Washington which are in formation. You are listed as an observer in all these games. Name Rules Players-Missing Berezina standard 7 Cannes pure 7 Dien youngstown 10 Eylau loeb9 9 ??? => Fontenoy 34 players 34 Please send me the necessary players. Also, note that you can dribble the players in. There is no need to wait for the 7th player to send the other six. Meanwhile the 6 can sign on and figure out how D.A. works. (Countries are NOT assigned until the 7th player signs on.) Yours, Daniel Loeb "Combinatorics: The Thought that Counts" Here is a letter from Ken Lowe who has programmed his machine to use my no NMR rules: From: Ken Lowe <ken@dewey.cac.washington.edu> Message-Id: <9010170021.AA27502@dewey.cac.washington.edu> To: loeb@geocub.greco-prog.fr Subject: Re: D.A. Cc: eric_s_klien@cup.portal.com > Finally, please tell Eric if you happen to run any other NoNMR games > (with or without Masters). He might be willing to include them in his > zine. There are several. I've never been too excited about watching play-by-play diplomacy games. It's much more useful to get an annotated post-game summary of what happened and who stabbed whom. Of course that summary (particularly the annotation part) is not something that my adjudicator does, but players and/or observers certainly have the option of generating the summary and/or the play-by-play and feeding it into EP. > Players should be instructed to: > 1) contact me personally > 2) send the following messages to judge@milton.u.washington.edu > HELP > GET HOUSE.RULES > GET FORM > 3) the form should be filled in and returned to judge > 4) the following final command should be sent to judge > SIGNON ?gamename password The actual name for the house rules is "rules" without the leading "house.". Also, nothing following the "help" command in a single message is processed, although the form file is included in the response from "help" so that part wouldn't matter. I'm getting a number of requests sent to "judge-request@milton..." that contain two lines containing "syntax" and "rules". Whoever is telling people to send this should correct it to say send to "judge@milton..." and include the "get" verb on the "syntax" and "rules" requests. They should also be sending "get info" (identical to the response from "help") so they'll have complete information. Just in case this is coming from one of you. > Do you think it would be possible to write a 34 player variant? Sure. Option (a) to allow players to build in any owned center would be the easiest to implement. Do you have 34 people lined up for this chaos? > Ok, could you start a game on your DA for each variant currently > supported. Here they are. The password for your master positions is the same as it is in your observer position in game hideg. If you don't remember what that is let me know and I'll forward it to you. The password on observer positions is normally optional since a matching return address is sufficient. Game 'berezina' is waiting for 7 players to sign on. The parameters for 'berezina' are as follows: Move clock 1410 min 12.00 next 72.00 grace 168.00 delay 0.50 days --TWTF- Retreat clock -1 min 0.00 next 24.00 grace 168.00 delay 0.50 days --TWTF- Adjust clock -1 min 0.00 next 24.00 grace 168.00 delay 0.50 days --TWTF- Access: different-site, Level: any, Variant: standard, NoNMR. The following players are signed up for game 'berezina': Observer 66 ken@dewey.cac.washington.edu Observer eric_s_klien@cup.portal.com Master 21 loeb@geocub.greco-prog.fr Game 'cannes' is waiting for 7 players to sign on. The parameters for 'cannes' are as follows: Move clock 1410 min 12.00 next 72.00 grace 168.00 delay 0.50 days --TWTF- Retreat clock -1 min 0.00 next 24.00 grace 168.00 delay 0.50 days --TWTF- Adjust clock -1 min 0.00 next 24.00 grace 168.00 delay 0.50 days --TWTF- Access: different-site, Level: any, Variant: pure, NoNMR. The following players are signed up for game 'cannes': Observer 66 ken@dewey.cac.washington.edu Observer eric_s_klien@cup.portal.com Master 21 loeb@geocub.greco-prog.fr Game 'dien' is waiting for 10 players to sign on. The parameters for 'dien' are as follows: Move clock 1410 min 12.00 next 72.00 grace 168.00 delay 0.50 days --TWTF- Retreat clock -1 min 0.00 next 24.00 grace 168.00 delay 0.50 days --TWTF- Adjust clock -1 min 0.00 next 24.00 grace 168.00 delay 0.50 days --TWTF- Access: different-site, Level: any, Variant: youngstown, NoNMR. The following players are signed up for game 'dien': Observer 66 ken@dewey.cac.washington.edu Observer eric_s_klien@cup.portal.com Master 21 loeb@geocub.greco-prog.fr Game 'eylau' is waiting for 9 players to sign on. The parameters for 'eylau' are as follows: Move clock 1410 min 12.00 next 72.00 grace 168.00 delay 0.50 days --TWTF- Retreat clock -1 min 0.00 next 24.00 grace 168.00 delay 0.50 days --TWTF- Adjust clock -1 min 0.00 next 24.00 grace 168.00 delay 0.50 days --TWTF- Access: different-site, Level: any, Variant: loeb9, NoNMR. The following players are signed up for game 'eylau': Observer 66 ken@dewey.cac.washington.edu Observer eric_s_klien@cup.portal.com Master 21 loeb@geocub.greco-prog.fr -Ken Here is a letter from loeb@geocub.greco-prog.fr/Daniel Loeb: DIPLOMACY PROGRAMMING PROJECT UPDATE ----- OCTOBER 17, 1990 1) I have changed and added the following commands to the protocol: Message Syntax: LOD file Meaning: The game is being resumed. You have saved the results in the specified file. Note: All programs should expect to receive either HLO or LOD as one of their first inputs. However, in theory, another program could receive its HLO and then send you a message before you have received yours. Example: LOD GAME1.ENG.DUMP Reply with: YES or NOP depending on if you are ready to resume the game. Message Syntax: SVE file Meaning: Please save the current position in the specified file for use later with the LOD command. Note: 1) Unless followed by an OFF, the game will continue. 2) Your program should also keep track of all messages that were sent recently but not confirmed. Such messages may have to be resent when the game is resumed. Example: SVE GAME1.ENG.DUMP Reply with: YES if save was successful and NOP otherwise. Message Syntax: OFF Meaning: Your program is being asked to terminate. Perhaps the game is over (because of a draw), perhaps your program has gone CD, or perhaps the game has been saved and will be resumed later, or perhaps the game is being aborted. Don't reply Message Syntax: MSG label (msg) Meaning: Please consider the indicated message, and identify it for later reference by the indicated label. A unique label should be generated for each such occurance of the message. Reply as per the indicated message. Message Syntax: LAB label (string) Meaning: Please in the future accept the label as being synonymous with the following string. Reply: YES or NOP 2) Ken Lowe indicates that he has begun entering a simple Diplomat called Randy. (Let's see if you can guess its strategy.) He hopes to write a more complicated diplomat later. 3) Michael Hall is preparing a great article on the theory of Diplomat and will eventually write a Diplomat of his own. I will send you the article as soon as it is ready (along with some discussion between me and Michael Hall). Here is a letter from loeb@geocub.greco-prog.fr/Daniel Loeb: I am now organizing a few Electronic Protocol Diplomacy games on the Diplomacy Adjudicator in Washington (judge@milton.u.washington.edu). These games are being run with No Civil Disorder rules and with a (computer assisted) human GM. I signal out one of the games in particular: "FONTENOY" is being played with the CHAOS variant rules. Every supply center is played by a different player (and builds are not restricted to "home" centers). Since this game only gives you a single unit, it does not represent a big time commitment. On the other hand, I need 34 players, so please sign up for the game FONTENOY. (The other available games are regular diplomacy, youngstown diplomacy, pure diplomacy, and my 9 player variant. ) For more information, write to "judge" or contact me. People interested in RUNNING games with the aid of the judge are also wanted. Yours, Daniel Loeb "Combinatorics: The Thought that Counts" Here is a letter from Loeb responding to a Floyd McWilliams/fmcwilli@oracle.com comment: First of all, I'd like to remind you that the 34 player game is already taking signups under the game name FONTENOY with electronic address judge@milton.u.washington.edu. I'm listed as gamemaster, but if you want to run the game, then I would be interested in playing! So if England grows to 17 centers while letting Greece get 6, she's not doing her job and deserves to draw! Give me a break, England even with 17 supply centers will be in no position to DIRECTLY influence events in Asia Minor. England can therefore not help drawing under these circumstances. Why change the rules more than necessary? * If you're going to have 34 players, each with a supply center, why not try the following variation: Each player starts with a total of three "microarmies" and "microfleets". Up to three microforces (of the same power) can fit in a province; this lets players use 1/3 strength, 2/3 strength, or full strength forces at their discretion. This would require a real overhall of the Diplomacy Adjudicator to get it to work. And in any case, this rule is ambiguous, since you haven't stated whether you need a FULL fleet to capture a province. You haven't stated whether a microunit retreating to a province can prevent a full unit from retreating there. You haven't given the convoy rules. You haven't said what happens when 4 microunits are ordered to the same province. And so on.... I agree that you should be allowed to build on other supply centers that you have captured. If you agree with me there, then let's make that one change, and leave the rest of the rules fixed. ("If it ain't broke, don't fix it.") * If you keep egging me on, Daniel, I'll be forced to post my 1939 variation (complete with Tanks and Bombers!). I'd be glad to see it. And I bet Eric would be glad to print it. Yours, Daniel Loeb "Combinatorics: The Thought that Counts" Publisher comments: Quote was by George Bush ****************************************************************************** To join in the fun, send your name, home address, home and work phone numbers, and country preferences to Eric_S_Klien@cup.portal.com. ****************************************************************************** Up