Diplomacy Zine -- Chapter Four EP #141 From: Eric_S_Klien@cup.portal.com Date: Sun, 11 Feb 1990 23:25:04 +0000 Issue #141 of ELECTRONIC PROTOCOL: Chapter One contains: NAVARONE, BLITZKRIEG, OPERATION OVERLORD, GETTYSBURG, and HMS HOOD And is published by daybell@aludra.usc.edu/Donald Daybell Chapter Two contains: DRAGONSLAYER, BISMARK, COLD WAR, JACAL, MANHATTAN, and VERSAILLES And is published by tedward@cs.cornell.edu/Ted Fischer Chapter Three contains: TANNENBURG, ENTENTE, MULHOUSE, DAWN PATROL, SNIKKEL-2, and BERLIN And is published by cwekx@htikub5.bitnet/Constantijn Wekx (Yep! We're changing publishers again!) ------------- Chapter Four ------------- Spring '05 of the game RATATOSK (BNC number 1989IJ) (GM is jall@diku.dk/Mogens Jallberg) Due Feb 8 Spring '05 of the game BUSHIDO (BNC number 1989IN) (GM is ronin@cory.berkeley.edu/Sam Parazette) GM just came back from Hawaii, results should be published soon. Spring '05 of the game HUGO (BNC number 1989IO) (GM is willis@trwind.ind.trw.com/Willis Marti) Due Feb 11 Winter '03 of the game JUGGERNAUGHT (BNC number 1989IR) (GM is rdesper@eagle.wesleyan.edu/Rick Desper) Due Feb 13 Winter '03 of the game TOKUGAWA (BNC number 1989IS) (GM is rc0o@andrew.cmu.edu/Bob Cochran) Due Feb 13 Fall '03 of the game PETAIN (BNC number 1989IT) (GM is ssmith@ms.uky.edu/Scott Smith) Just replaced a player, results will be published soon. Fall '03 of the game DUNKIRK (BNC number 1989KN) (GM is sjzwange@phoenix.princeton.edu/Steven Jacob Zwanger) Just replaced a player, results will be published soon. Fall '04 of the gunboat game VERDUN (MNC number 1989AZrb32) (GM is eric_s_klien@cup.portal.com/Eric Klien) Due Feb 11 Fall '01 of the game BIG WILLIE (MNC number not known) (Blind game, GM is okamoto@hp-ses.hp.com/Jeff Okamoto) Received. Summer '01 of the game FIRE WHEN READY (BNC number 1990B) (GM is rlg@ai.mit.edu/Bob Givan) Nothing happened. Spring '02 of the game DREADNOUGHT (BNC number not known) (GM is rbk@aiai.edinburgh.ac.uk/Richard Kirby) England ------- F Lon - Wal <failed> F NTS - Lon <failed> F NWS - NTS <failed> A Nor S GERMAN F Den - Swe Germany ------- F Den - Swe F Kie - Hol A Hol - Bel A Ruh S A Hol - Bel A Mun - Bur Russia ------ F Sev - BLA <failed> F Swe - BAL A Mos - Ukr A War - Gal <failed> A Stp - Liv Turkey ------ F BLA S A Ank - Arm <failed> F Con - AEG A Rum - Gal <failed> A Ank - Arm A Bul - Rum <failed> Austria ------- F Gre - Alb A Ser - Tri A Vie S A Ser - Tri A Bud S A Ser - Tri Italy ----- F Tun - WME F Nap - TYS A Ven - Tyr A Tri S A Ven - Tyr <failed> A Pie Holds France ------ F ECH C A Bel - Wal F Bre S F ECH F Mar Holds A Spa - Por A Bel - Wal <failed> ========================================================================== Retreats -------- Since the two armies that have to retreat only have one space to go to I have taken the liberty of moving them there. If you want them to disband instead then please notify me ASAP. Italy: A Tri RETREATS TO Ven France: A Bel RETREATS TO Pic ========================================================================== Dreadnought Summer 1902 Unit Positions England (Scardamaglia) F NWS F NTS F Lon A Nor Germany (Hallen) F Swe F Hol A Bel A Ruh A Bur Russia (Pease) F BAL F Sev A War A Liv A Ukr Turkey (Cochran) F BLA F AEG A Arm A Rum A Bul Austria (Mullins) F Alb A Bud A Tri A Vie Italy (Nelson) F WME F TYS A Pie A Tyr A Ven France (Belle) F Bre F ECH F Mar A Por A Pic ========================================================================== Press ----- The Czar News Bear April 8, 1902 The Czar announced today that Austria will be crushed by the end of the year, and that Austria would be divided up between Italy, Turkey, and The Great Bear. The Czar also announced that a new map coordinator has been found, replacing the last one who was beheaded for failing to move St. Petersburg forces into Sweden via transport on the map, thus when the Czar put his orders through like he usually does, he called off map positions and only later reali that St. Petersburg was cut off from the war. When talking about the English, who, he promised, would also be wiped out by the end of the year, he said only, "Scum who talk to my allies and put words into my mouth, shall die for their arrogance." We, the general peasant public, can only guess at what he is referring to. ------------------ Messages to other players: England: You really thought you could pull that desparation move off, didn't you Well, I'll tell you right now that I'm not going to pull out of Sweden. Instead, we are going to nail you bigtime at Norway, and then your piddling fleets will slowly be hunted down and sunk one by one by Germany, France, and myself. If you ever think about telling Germany that my forces were pulling out of Sweden to take care of the south, you should make sure that Germany doesn't get back to me about it. But, you wont have a chance again...in two moves you are history...and the london bridge shall come falling down.... Turkey: Not sure if you noticed, but Austria is slowly creeping up your backside with his fleet. I suspect this move he will move the Greece Fleet into the Aegean, and once there he can nail constantinople or Smyrna. I'm moving my fleet at Sevastpol into the Black Sea. It has sat there idle for too long. I suggest you move to constantinople with either your fleet or your army or both. Don't worry, our nonagression pact still stands, but I require safe passage through constantinople if I ever want to come into the Aegean. You now must make up your mind. After the crush of Austria we have to split it up, and I need control of at leasat one more of the black sea ports...so pick one. But first, I suggest you nail Serbia and then support my crush of Vienna when the time comes. Italy already has Trieste and has agreed to support my attack on Vienna. TO Italy::: You read my message to Austria...take out Budapest and then support my attack on Vienna. To Germany::: I have not entered your provinces nor shall I. To France: Destruction of England is almost complete...careful how you and Germany split up your parts. To Germany & France: I have Sweden and I wish Norway. To Austria: Do you still want to crush me? But, you're right, it has been fun. =========================================================================== DEADLINES et al. ---------------- I have been asked to put forward another vote for speeding the game up. The next deadline for Fall 1902 moves, and for the vote will be TUESDAY 13th FEB at 10pm EST Spring '01 of the classic game RED STORM RISING (MNC number not requested) (GM is jbillone@jarthur.claremont.edu/Jeremy Billones) Due Feb 11 Summer '01 of the game HOHENZOLLERN (BNC number not known) (GM is jog@hpda.hp.com/Rajeev Jog) Nothing happened. Winter '1913 of the game #40 COMRADES IN ARMS (MNC number not known) (GM is att!druwa!hardlj or hardlj@druwa.att.com/Larry Hardouin) Due Feb 20 Spring '01 of the game ?? (BNC number not requested) (GM is cap@htikub5.bitnet/Joris Pinkse) I'll need to get the info on this game. Summer '01 of the game SNIKKEL (BNC number not known) (GM is cap@htikub5.bitnet/Joris Pinkse) Nothing happened. Publisher comments: As you can see, quite a few games have gotten started. Luckily, I put one of them in chapter three which is under new ownership. And I moved another game from this zine into that chapter. It would be real nice if I could get a few more guest publishers so I could reduce the size of this chapter. Note that being a guest publisher is my most important position, so only take it if you can do a real good job! I will be getting a computer for home e-mailing in the very near future. Probably this week. (I am looking for a good deal as we speak.) This should help me stay up to date in answering my mail and publishing this zine. If anyone knows of any good deals on IBM compatibles let me know. I need a hard disk drive and 2400 modem in the setup. I would be interested in new as well as used computers. Here is an interesting conversation about Diplomacy and computers from rec.games.pbm: From kdb@chinet.chi.il.us/Karl Botts: In response to my email request for information about Diplomacy software, Eric Klien replied: (I hope he will forgive me for replying in an open letter, in hopes of stimulating a wider conversaion in rec.games.pbm) > Software for playing Diplomacy is only good for GMing games. > Computers make very poor opponents. COMPUTER DIPLOMACY by Avalon > Hill is lousy, a better game is JUDGE for IBM compatibles published > by someone in Canada. (I'll get you the info if you want it.) > > Again, computers are only good for GMing games, not as opponents. > > Eric Klien I was thinking more along the lines of something that would assist a human playing Diplomacy; help him to organize his information, manage his communications, analyze scenarios, and so forth. This would appear to involve many of the same capabilities as would be needed to GM a game; thus, I would be interested in what you know about such software. I'd like to gather some info about what has been done, and get some other people's thoughts about what should be done, as well as getting a few more games under my belt so I can form my own opinions. Perhaps an article and some discussion in your Diplomacy Zine would ensue. Eventually, I'd give some consideration to developing something. I think you mentioned somewhere that you haven't gotten around to playing Blind Diplomacy yet; you should. One of the reasons I have been thinking about such software is that I believe the Blind variation potentially adds an order of magnitude of complexity to the game, if only you could manage the complexity. What I mean is that, in regular Dip there is a single state of the game at any time (call it "reality"), and a set of variables describing the future behavior of the players. The future states of the game vary on these variables only (I make the implicit assumption that predicting the future states of the game would be helpful in winning.) But in Blind Dip there are an arbitrary number of present states of the game, dependent on your evaluation of the information available to you. The state is indeterminate even if you restrict yourself to what you know to be true, i.e., what you can "see", because this information is incomplete. But if you also consider other information -- what other players have told you _they_ can see -- the state is even less determinate, varying with the veracity of your informants. Furthermore, the variables describing the future behavior of the other players are also less constrained, because they now vary with the information available to the players. This in turn varies, in part, on something you can immediately control, that is, what you tell the other players. The effect of this control is indeterminate too, however; you can tell them whatever you like but you cannot make them believe you. We humans generally deal with such situations by ruthlessly pruning the tree of possibilities down to something we can manage, which is pretty close to a single trunk. In other words, we pick one, or a small set, of the possible states of the game, label it "reality", and then proceed essentially as though it were known to be true. We humans are not good at managing multiple alternate realities. So it occurs to me that some software that could store and manipulate versions of Diplomacy Reality would be especially helpful in playing Blind Dip. Here's a small example of a concrete area where it might be useful: One problem of information management that immediately arises in Blind Dip is that, following each move, you often report the results of your moves to one or several allies; it is possible that these reports will differ. In future negotiations with the various parties it is necessary to remember what the contents of your recent reports to them were, or estrangements may occur. This tends to necessitate much grepping through mail messages and scribbling on pieces of paper, a process which is tiresome and error prone. It would be useful to have a program that would generate a screenful or two of information summarizing recent position reports transmitted to another player, while negotiating with or considering policy about him. If I were to impose a little discipline on the way I compose such reports using a general text editor or emailer, I could readily write a shell script to dig this info out for me, sort it, format it, and whatever. Of course, this would be mininal functionality; I can imagine special report editors, databases and managers... The possibilities are many. I just got to wondering if such a system already existed for regular diplomacy, and if I might be able to extend it for use with Blind Dip. Have you heard of anything along these lines? -- Karl Botts kdb@chinet.chi.il.us ...{mcdchg|att|nucsrl|gargoyle}!chinet!kdb Work, M-F 10-6 CST more or less: 312-630-9880 Home: 312-493-8152 (Now cross-posted to rec.games.board, which is more appropriate than .pbm) From bellcore!nvuxh!hall/Michael Hall: In article <1990Jan22.062456.1421@chinet.chi.il.us>, kdb@chinet.chi.il.us (Karl Botts) writes: > > I was thinking more along the lines of something that would assist a > human playing Diplomacy; help him to organize his information, manage > his communications, analyze scenarios, and so forth. [Munch.] > We humans generally deal with such situations by ruthlessly pruning the > tree of possibilities down to something we can manage, which is pretty > close to a single trunk. In other words, we pick one, or a small set, > of the possible states of the game, label it "reality", and then proceed > essentially as though it were known to be true. We humans are not good > at managing multiple alternate realities. I was considering writing a Diplomacy tactical advisor that would exhaustively (or heuristically) search through the space of possible moves and advise you on sets of moves considered most profitable for every country. The computer would not need to take into account alliances, because by ignoring alliances it will note potential stabs. You could play what-if scenarios to get a better grasp of the current situation. Unfortunately, the space of possible moves is so vast that I have (at least for now) given up on this idea. It is much worse than chess. In chess just one piece of one player moves at a time, whereas in Diplomacy all pieces of all players move simultaneously. (I have written an extremely fast Blackjack analyzer, but again, in Blackjack moves are sequential, not simultaneous.) 34 pieces to the power of (~5 possible moves + ~5 possible defense supports + ~5 possible attack supports) is the kind of combinatorics we're talking about for just one turn. Of course, you could try many things to reduce the search space. Humans can play Diplomacy, after all, and they don't even think that the tactics are the hardest part. What I usually do is segment a small portion of the board to deal with at a given time. Sometimes a situation on one side of the board will affect moves on the other side of the board, but not all that often. Similar techniques could be exploited to let the computer compute suggested orders in less than one week CPU time. Also, you either need a very good static state evaluation function to evaluate how favorable board positions are to each country or else you need to look very many turns ahead and then use a simple state evaluation like "number of supply centers". Otherwise, not much will prevent Turkey from trying to open with A Smy-Syr! I think a good approach would be to let the user specify a subset of all pieces and a subset of all moves for those pieces. The computer can then start trying to come up with good sets of orders for any given country. I think Mr. Botts had grander visions than a mere tactical advisor, but I think a tactical advisor is on the outter edge of what would be practical to implement. -- Michael R. Hall |"We have seen pictures [of Mars] where there are hall@nvuxh.cc.bellcore.COM|canals, we believe, and water. If there is water, bellcore!nvuxh!hall |there is oxygen. If oxygen...we can breathe."-Quayle ----- From kevin@math.lsa.umich.edu/Kevin Coombes: In article <485@nvuxh.UUCP> hall@nvuxh.UUCP (Michael R Hall) writes: > In article <1990Jan22.062456.1421@chinet.chi.il.us>, >kdb@chinet.chi.il.us (Karl Botts) writes: > > I was thinking more along the lines of something that would assist a > > human playing Diplomacy; help him to organize his information, manage > > his communications, analyze scenarios, and so forth. > > I was considering writing a Diplomacy tactical advisor that would > exhaustively (or heuristically) search through the space of possible > moves and advise you on sets of moves considered most profitable > for every country. The computer would not need to take into account > alliances, because by ignoring alliances it will note potential > stabs. You could play what-if scenarios to get a better grasp of the > current situation. > > Unfortunately, the space of possible moves is so vast that I have > (at least for now) given up on this idea. It is much worse than > chess. In chess just one piece of one player moves at a time, > whereas in Diplomacy all pieces of all players move simultaneously. > ... > 34 pieces to the power of (~5 possible moves + ~5 possible defense supports > + ~5 possible attack supports) is the kind of combinatorics we're > talking about for just one turn. Of course, you could try many > ... > Also, you either need a very good static state evaluation function > to evaluate how favorable board positions are to each country or > else you need to look very many turns ahead and then use a simple > state evaluation like "number of supply centers". Otherwise, not > much will prevent Turkey from trying to open with A Smy-Syr! > ... I've also thought about this some. I haven't had the time to implement anything, but I'm not as pessimistic as Mr. Hall. I think you have to define the problem in the correct terms. A ``tactical advisor'' should only try to find the best moves for one country. This reduces the number of pieces you're looking at considerably. You should also concentrate your efforts on writing a good state evaluation function, since a search through several game turns is prohibitively expensive. It might be possible to write such a function which only looks at the possible moves in the current game turn. An advantage of this approach is that it also limits the effective size of the board to the set of spaces which are at most two moves distant from any of your currently occupied spaces or currently controlled supply centers. Here are some random thoughts that might go into an attempt to write such an evaluation function. 1. Each space on the board should have some intrinsic value which depends on its status as a supply center, its mobility (separated into land connections, sea connections and coastal connections), and its access to other supply centers. 2. This intrinsic value should be modified by a number of factors, such as: who controls it, who occupies it, who is adjacent to it, and what season it is (mobility being more important in the spring, and control of supply centers in the fall). 3. The numbers generated by the considerations of 1 and 2 should be used to get a ranked list of the spaces which it is most important to occupy or to prevent others from occupying. You can then search through the set of possible moves and evaluate the resulting positions with respect to how many of the high priority goals are achieved. This approach will prevent Turkey from opening A Smy-Syr; it will, however, strongly encourage England to open F Lon-ECh, F Edi-Nth. I haven't been able to decide in general whether it is more likely to lead to aggressive or defensive play. If you don't make provisions to modify the values based on estimates of the strengths of alliances, then I suspect it will lead to aggressive openings followed by passive stalemates. Kevin Coombes <kevin@math.lsa.umich.edu> From kdb@chinet.chi.il.us/Karl Botts: >> I was thinking more along the lines of something that would assist a >> human playing Diplomacy; help him to organize his information, manage >> his communications, analyze scenarios, and so forth. >I think Mr. Botts had grander visions than a mere tactical advisor, >but I think a tactical advisor is on the outter edge of what would >be practical to implement. Well, actually my visons were considerably more pedestrian, at least for the beginning. I would just like to have a better way of organizing and presenting the information to myself; I would do the tactical analysis. For instance, I would like to be able to store a tree of maps, very much like an SCCS or RCS tree of source code versions. When I got new info, such as the results of a recent move, or some claims by an ally of mine about what he could see (again, I am thinking mainly about blind Dip at the moment), or perhaps just some guesses I had made, I would make a new version of the map. I would need to differentiate reliable branches -- i.e., those based only on move results -- from unreliable branches, i.e., those contaminated by rumors or guesses. There are lots of little tools like this imaginable. As I noted, the requirements are not a great deal different from the kinds of things that would be helpful to a GM, expecially if he was GMing several games. This was why I thought that something like it might exist. I would expect to do the text processing with the shell and awk and such as much as possible. You could almost do the whole thing with shell scripts, except that it would seem that somewhere along the line you are going to have to manage a fairly complex graph, with countries as vertices and legal moves as edges, and info atttached to the vertices representing units; call such a graph a "Dip State". This would be more work to deal with using text processing tools than in C. It's not a big deal, though; it's not really a very big graph. I had some idea that if I had a general purpose tool that could manipulate such graphs -- store them in files, apply diffs to them, and such -- I could hook it up with some text processing and see what I could do. The main thing I need to build such a program is the data; the countries and the legal moves between them and so forth. I suppose I could produce it by pouring through the Dip manual, but I know there are some ambiguities that have been resolved by tradition, and besides it would be laborious and error prone -- just a crummy data entry job. It occurred to me that GMs must have such data already, or Eric_S_Klien at least, and I was hoping somebody would mail it to me... Karl Botts kdb@chinet.chi.il.us, ...{mcdchg|att|nucsrl|gargoyle}!chinet!kdb From utility@quiche.cs.mcgill.ca/Ron: In article <485@nvuxh.UUCP> hall@nvuxh.UUCP (Michael R Hall) writes: >34 pieces to the power of (~5 possible moves + ~5 possible defense supports >+ ~5 possible attack supports) is the kind of combinatorics we're >talking about for just one turn. I should think that each piece averages no more than 8 orders per move. To start with, if it IS supporting your own piece, that restricts the other piece substantially, and often times there are few supports available. > Of course, you could try many >things to reduce the search space. Humans can play Diplomacy, after >all, and they don't even think that the tactics are the hardest >part. What I usually do is segment a small portion of the board to >deal with at a given time. Sometimes a situation on one side of the >board will affect moves on the other side of the board, but not all >that often. Similar techniques could be exploited to let the >computer compute suggested orders in less than one week CPU time. There exists an extension of the alpha-beta algorithm for look-ahead in the search space (I intend to publish this in future) for games like diplomacy. Of course, the analysis of the bargaining game needs heavy work. As for divide and conquer, I do believe that this is the only way to go in computer diplomacy play. Might as well abstract out to resource allocation in regions since diplomacy tactics are essentially regular enough that people can meaningfully speak of allocation of units to strategic tasks. >Also, you either need a very good static state evaluation function >to evaluate how favorable board positions are to each country or >else you need to look very many turns ahead and then use a simple >state evaluation like "number of supply centers". Otherwise, not >much will prevent Turkey from trying to open with A Smy-Syr! Smy-Syr is easily enough prevented by rules about threats and defense. It doesn't defend any endangered SC's nor does it threaten any SC's nor does it increase tactical support. >I think Mr. Botts had grander visions than a mere tactical advisor, >but I think a tactical advisor is on the outter edge of what would >be practical to implement. I don't really agree. A tactical advisor is not too tough to implement badly, and with a week of processing and some abstraction, 3 ply lookahead seems reasonable (of course, horizon effects would still be DEADLY). After this, the computer can probably engage in some kind of negotiation -- using game theoretic principles of arguing on the basis of its best understanding of the strength of a given player's cooperation (i.e. it presents arguments from rational alternatives for the player's cooperation). To the extent to which the computer believes the other player is rational and is strongly motivated to act according to its vision of rational play, it will trust that player. Others should be duped into doing what is worst without giving away its hand and making it clear that the computer is trying to dupe or WHAT tactics it has (e.g. don't write your obvious enemy Germany in a tense "which piece to support" situation with an offer for him to take e.g. Belgium, when this suggests too obviously that you will be defending that square and leaving Burgundy open). Ron Here is the latest stuff on my e-mail problems (which have pretty much faded away): Here is an update on the situation with Sun. First, about X.25 (Telenet and PCI). Sun sent someone to our site and attached a protocol analyzer to our machines. Lo and behold, it showed that the Sun was doing all sorts of bad things. Although we do not know why they are happening, the numerous protocol violations explain the mass disconnects and possibly the REJECTING 00 40 condition. Sun is now analyzing the information and trying to determine the cause. Next, about the rpc problems. We believe that we have generally fixed the rpc problems. We have a new version of some Sun software that fixes most of them, and have added a number of work- arounds to the Portal software to fix the rest. You may notice occasional delays as a result of the workarounds. If these get to be persistent, try logging out and back in. One unfortunate side effect of the rpc failure is that some people's subscriptions were lost. The rpc failures were in- correctly interpreted as an attempt to access a nonexistent conference, and the subscription to the conference was dropped. The rpc code would stay wedged during the entire subscription reading process, removing all of the conferences in the list. When we reported the problem with certain programs being run twice by the cron program, we were told that this was a known problem. Sun is working on a fix and should have it "real soon now." When pressed for details, they said that it might be as soon as a month. We modified the Usenet loader to detect this condition and work around it. So far, we have seen no other af- fects of the double execution. One more problem: Sun gave us a new mail handler (sendmail) that allegedly fixed a number of mail problems. In doing so, it created another one -- lost mail. Sendmail neither returned the letters nor delivered them, it just died. For several hours to- wards the beginning of the week, much of the network mail passing through Portal was lost. Sun has not yet responded to our bug report on this problem, but we have determined the situation that triggers the mass failure and reconfigured our mail system to avoid it. Note: this problem only affected network mail, not mail between Portal users. To conclude, we think that we have identified the major problems in the new Sun software. Sun is aware of the problems (and their awareness reinforced by us every day!) and working on solutions. Whenever possible, we are installing workarounds to avoid their bugs. We will post another report when we have more information to report. Thanks for your patience. ************************************************************************ I need more scribes! I need more scribes! I need more scribes! I need ************************************************************************ I am enjoying moderating this zine, keep that mail coming! Eric Klien Up