Diplomacy Magazine -- Chapter Two From: loeb@geocub.UUCP (Daniel LOEB) Date: Wed, 19 Feb 1992 12:32:37 +0000 Issue #274 of Chapter Two of the Electronic Protocol By Daniel Loeb (loeb@geocub.greco-prog.fr) February 19, 1992 Distributed via: rec.games.pbm rec.games.board dipl-l ------------------------------------------------------------- Electronic Protocol Games played on the Diplomacy Adjudicator ------------------------------------------------------------- **** TABLE OF CONTENTS **** PART ONE - Opinions, Letters, and Editorials: Introduction Play against the computer Victory list PART TWO - Summary of all moderated games on the Judge: List of game openings List of EP games on the DA For more info List of Gamemasters PART THREE - Diplomacy Programming Project ***** PART ONE ***** **** INTRODUCTION **** This issue of EP CHAPTER TWO is mostly concerned with the DIPLOMACY PROGRAMMING PROJECT. This is a project designed to study the negotiation, and to design an automated diplomat. Two of my master's students work actively on the project, and many people over the network correspond via email. The next two articles suggest some easy ways you can help participate, and Part Three is a copy of the last DPP update. For more general information, please contact me. **** VICTORY LIST **** The following is a list of (standard or gunboat or 1898 variant) games on the JUDGE that have finished with a 1 player win. I have the set of positions for each of these games and am currently analysing them so that our computer diplomat will know the true importance of each province. JUDGE Austria: Ajax, Jutland2#112 2 wins England: Xray 1 win France: Agincourt(1898)#120 1 win Germany: Orange 1 win Italy: Munk 1 win Russia: Zulu#77 1 win Turkey: Dragons#75, Turbo 2 wins Since the above data is rather sparse, I'm looking to supplement the data with NON Judge games. The following for example is a list of all standard or gunboat games in ELECTRONIC PROTOCOL according to their finish. *** PLEASE CONTACT ME IF YOU HAVE THE POSITIONS FROM ANY OF THE GAMES LISTED BELOW AS SOLO WINS *** ELECTRONIC PROTOCOL SOLO WINS A DUNKIRK#31, DRAGONSLAYER#12, NICKEL#46 E JACAL#17 F NAVARONE#8, CAPTAIN-CAVEMAN#50, GETSOME#81 G MAELSTROM#4, DEFCON5#80, REPUBLIC#74 I D-DAY#7, VERDUN#19 R TANNENBURG#18, MANHATTAN#36, SOMME#101 T --- 2-WAY DRAWS AG ARNHEM#69 AI DRESDEN#49 AR PETAIN#30, WARTHOG#117 EF DOUGHBOY#13, RIYADH'S-RECKONING#107 EG BAGHDAD#78, CHECK#89 ET THUNDERDROME#76 FI RED-STORM#5, BISMARK#14 FG GETTYSBURG#11 FT SNIKKEL2#43 GR OZARK#44 GT PANZER#3 RT BUNKER-HILL#1 3-WAY DRAWS AEF CONSTITUTION#33 AEG COLD-WAR#16 AER HUGO#26, BLITZKREIG#9 AFG MONTREUIL-SUR-MER#59 ART HORSE#51 EFI VERSAILLES#27, YORKTOWN#54, APPLE#87 EFT SNIKKEL#42, AUSTERLITZ*71, WARP#127 EGT PEARL-HARBOR#6, SNIKKEL#42 ERT OPERATION-OVERLORD#10 FGT HOHENZOLLERN#39, ARCHANGEL#53 FRT COLIN-POWELL#113 GIR TRENCHFOOT#20 4-WAY DRAWS AEFI GALLIPOLI#2 AEFT JUGGERNAUT#28, KHAN#60 AEFR LEIPZIG*128 AEGI BERLIN#32 AEGR EL-ALAMEIN#41 AEGT MASADA#56 EFGT TEUNISGEK#106 FGIT HIROSHIMA#61 5-WAY DRAWS AEFIR DAWN-PATROL#23 AEFGT TOKUGAWA#29 EFGIT RATATOSK#24 7-WAY DRAWS AEFGIRT BUSHIDO#25, DREADNOUGHT#37 **** PLAY AGAINST THE COMPUTER **** We hope to begin testing the strategic part of our diplomat in the near future. Probably the beginning of March. For this purpose, we are accepting signups for a game inwhich the computer may play one or more positions. Please signup for DPPA. Rules: (1) I will be GM, and will run the DIPLOMAT. (2) In order to keep the game fair, since our current DIPLOMAT does not negotiate. This game will be NO PRESS - GUNBOAT. (3) If DPPA fills up, then more games will be started. (4) I reserve the right to upgrade the DIPLOMAT to a higher version as our program progresses. (5) I reserve the right to replace players who quit with DIPLOMAT. (6) If the DIPLOMAT is eliminated, I reserve the right to inform the players of that fact, and resign as gamemaster. The remaining players can continue as an unmoderated game, or can find a new gamemaster. (7) Copies of the DIPLOMAT will probably *NOT* be available to the players. **** REC.GAMES.BOARD **** There has been an objection to this chapter being posted on REC.GAMES.BOARD. If you have any opinion either way, then please write.... Complaint From: toms@ichips.intel.com I stopped reading rec.games.pbm becuase I didn't want to see weekly diplomacy status for a zillion games. I used to like in the mid 80's when people just talked about the games and openings, no game results.. I could never see why email wasn't a better method of getting the results out. Let's keep the results to one news group. The people who are interested can check in pbm. My response: <Actually, I'm not sending out weekly results, but rather regular 1-line updates about many game. Mostly what my zine includes is a discussion of the game diplomacy. If this is unwanted in RGB then I'll take it elsewhere....> ***** PART TWO ***** **** LIST OF GAME OPENINGS **** Moderated Replacement Openings. Khafji - Standard Gunboat - England (3/2) - One removal needed Unmoderated games opening include: berkeley, squid, empire, math, zen, stockhol, tuba Signups available for the following moderated games in formation: gallon, 7senuf, turin, normandy, dppa Please volunteer to be a GM. **** LIST OF EP GAMES ON THE DA **** --- EXPLANATION --- Here is an update on games played on Judge. Each game is represented by a line of data (followed possibly by a line of comments --- please send me comments I can use if you are a GM!). The games are sorted according the the variant rules which are used. The 1st column gives the name of the game including a "#" if the game is a "private" or "unlisted" game. The 2nd column gives its Electronic Protocol number if available. If the game is published in another EP chapter (other than number two), then that is indicated after a slash. The 3rd column gives the name of the GM (see list of GMs below). The other columns give updates on this game in chronological order with the most recent entry on the right. The updates for a game in formation indicate the number of players needed to start. For example, -5. The updates for a game in progress indicate the season (F for Fall or S for Spring), the last 2 digits of the year, and the phase (M for movement, B for builds, and R for retreats). All this is possibly followed by the indication of the number of replacement players (-1), temporary replacements (T1) needed, or needed later on (*1). I'm listing the status of each game not only for this week but for the last several weeks, so that you can see not only where the game is, but how fast it is moving. Please tell me if this extra information is useful to you. --- LIST --- Name EP# GM Oct4 Nov4 Dec5 Jan3 Feb4 Feb10 Feb19 ------- --- -- ----- ---- ---- ---- ---- ----- ----- STANDARD RULES cubit 158 jdr S01M S01M F01M dram 159 jdr S01M S01M S01M evolt jdr -7 -3 S01M fulong jdr S01M gallon jdr -3 marengo 129 scottb S03M S05M F06M F07M S09M F09M S10M Ita strong. Ger builds. Fra survives. Eng/Aus weaken. osijek 137 cebulad F02M S04R S05M S06M F06M F06B The Juggernaut has stumbled, what will the FEGI do now?!? paris 134 skiman -4 S02M F03M S04M S05M F05M S06M portnoy 125 koll02 F10M F12M F14M F15M S17M F17M F17M sparrow 133 casmacin S02R F03B S05R S06M S07M F07M F07R Italy trys to play Kingmaker, England pulls ahead of Turkey. # normandy loeb -2 -3 -3 -3 -3 -1 -1 All press will be written in French. Le jeu sera en francais. croatia 148 nick -2 F02M F02R F03M F03M S04M Germany Reprieved. EFG Alliance. Eng & Tur vs Russia. quebec 154 nick S01M S01M STANDARD GUNBOAT - Identities of players unknown. All messages are public. # khafji 138 skiman S03M F05M-1 F05B F06B F06B F06B-1 Fast pace. No press. Delays over a few hours unacceptable. boadicea 147 pl436000-6 F04M F08M S09M S11M S11R F11M conan pl436000 -5 -2 F02B F03M F03B dppa loeb -6 Play against a computer opponent (DPP version 1a). No press. YOUNGSTOWN RULES (10 players) Extended map including Asia and Africa. dien 124 jdr F07B F09M F10B-1 S11M F11M F11B S12M giggles dwiseman S03M F04B S06R S07R F08B S09M F09M Local ERIM game timor nick -10 -5 S01M LOEB9 RULES (9 players) Spain and Scandinavia are added as extra players. eylau 153 loeb -3 S01M F01B-1 S03M F03M F03M S04M Spain stabs England. Austria/Russia at war. GREAT BRITAIN RULES (7 players) Britain starts with 6 SC's but all armies! hastings 139 loeb S01M F01M F03B F04M-1 S05R F05B S06M Russia's only unit is DESTROYED while bouncing Turkish retreat CHAOS RULES (34 players) Regular map. Each SC is owned by a different player. fontenoy 114/4 jdr S05R-6 S06M F07R F08M F09M F09B 7 players left. 21SC B-S-3 alliance builds southern fleet iona jdr -20 -13 S01M F01M F01M Just started. 30 players share 34 supply centers. 1898 RULES (7 players) Regular map. Each player starts with only his capital. bataan 152 dmb -3 F99M F00B F01R S02M F02M CROWDED RULES (11 players) 4 more players added, leaving no neutrals. 7senuf jdr -5 -5 -6 -6 -6 -5 -4 emu mjmcleod S02M-4 S02R-1 F02M-2 F03M F03M-1 F03M MACHIAVELLI RULES - An economic variant of Diplomacy marketed by Avalon Hills vega cebula -7 -5 -4 -3 U54M F54M S55M Pope takes out Florence in 1st move but at what political cost? dagger jdr ??? S56R S57M U57R F57B-1 S58M S58M poverty andre -6 -6 -5 -3 S54M **** FOR MORE INFO **** 1) FOR A PARTICULAR GAME: For more detailed information about game's current status: Send "LIST <name-of-game>" to JUDGE@U.WASHINGTON.EDU For the game's history: Send "SUMMARY <name-of-game>" to the JUDGE For regular game updates: Send "OBSERVE <name-of-game> <password>" to the JUDGE For a copy of the variant rules: Send "GET INFO.<name-of-variant>" or "GET PRESS" or "GET GUNBOAT" For other information (including house rules): Contact the GM. (See list below) 2) FOR THE JUDGE IN GENERAL: For general information about the judge, Send "HELP" to the JUDGE For an update of the list of games given above: Send "LIST" to JUDGE@U.WASHINGTON.EDU For a more detailed list: Send "LIST FULL" to the JUDGE To be informed of all changes in the list of games, Send "OBSERVE CONTROL <password>" to the JUDGE. For a copy of the default house rules: Send "GET EP.HOUSE.RULES" to the JUDGE. For other information: Contact me (loeb@geocub.greco-prog.fr) or Ken (jdr@u.washington.edu) 3) FOR THE ELECTRONIC PROTOCOL IN GENERAL: For general information: Contact the editor Eric Klien (eric_s_klien@cup.portal.com) To solve Email problems, contact one of the email wizards: swb@ccwf.cc.utexas.edu, andre@hern.stonemarche.org eisen@cs.jhu.edu, eisen@jhuvms.bitnet, or wcw27974@uxa.cso.uiuc.edu For back issues: Contact the archivist, jlitvin@swtec1.intel.com/John Litvin **** LIST OF GAME MASTERS **** Brian Bacher bacherb@physics.orst.edu David M Bowen dmb@bigd.cray.com dmb@sequoia.cray.com Dave Cebula cebulad@physics.orst.edu Jamie Dreier pl436000@brownvm.brown.edu, pl436000@brownvm.bitnet Bryant Durrell durrell@umaxc.weeg.uiowa.edu Nicholas Fitzpatrick nick@sunburn.waterloo.edu Nawwar Kasrawi skiman@leland.stanford.edu Edward J Koll koll02@snybufva.bitnet Koll02@snybscva.bitnet Danny Loeb loeb@nestor.greco-prog.fr Ken Lowe jdr@u.washington.edu ken@milton.u.washington.edu Matt McLeod c9106225@wombat.newcastle.edu.au Michael Luft scottb@cs.utexas.edu Sean MacIntosh casmacin@atlas.cs.upei.ca John Aidan O'Regan J_ORegan%csvax1@iruccvax.UCC.IE Andre Verweij andre@duteina.tudelft.nl andre@hlniob.uucp ***** PART THREE ***** DIPLOMACY PROGRAMMING PROJECT UPDATE ---- FEBRUARY 12, 1991 Here is a status report of progress on the diplomacy programming project. (1) The DPP PROTOCOL has not gone any dramatic changes. However, Antoine Bertier (new member of DPP) has discovered a number of errors which have been corrected. (See dpp/protocol) (2) My students Constantin Stykov and Jean-Marc Aubert have begun programming a diplomat. CS's goal is a strategic program which given a position searches for the best move. For the moment, the program is written in series, and merely will maximizes the value of the resulting position for one player given the moves for the other players. However, eventually, I hope to have the program in parallel. Thus, each country (or alliance) would improve its moves based on the best moves currently being considered by the other countries. A stable situation then corresponds to a set of moves such that no alliance wants to change its moves unless another does. The program will take the following inputs: 1) A new postion each turn (which is stored as a JUDGE Gnnn file) 2) Various problems. Each problem consists of: - an id number - a list of alliances (current only one alliance can be considered at a time) - time allocated to this problem, or search depth - a list of moves that must be done - a list of moves that must not be done - a list of default moves Assuming the inputs are valid, the output then consists of a set of orders for each of the alliances being considered along with a numerical estimate of the value of these moves to each alliance. The series version of this program is being implemented in the following way. 1) Let D be the defaults indicated if any 2) Let the agenda A be empty 3) Let M be the set of moves which must be done 4) Complete M into a full set of moves MC by using defaults. 5) Calculate the position P resulting from MC (using the Diplomacy Adjudicator "dip -t") 6) Calculate the value of the position P 7) Insert the pair (M,P) into the agenda sorted by value of P. 8) Let (M,P) be the first line on the agenda 9) Remove (M,P) from the agenda 9) If M is a complete set of moves AND is the best so far then output and store (M,P). 10) If M is a complete set of moves, go to step 8 11) Otherwise, M is an incomplete set of moves. Calculate the set S of possible ways to complete M with an additional moves (which is not forbidden). 12) Eliminate from S those sets of moves already in the agenda 13) Insert each set of moves M from S into the agenda using steps 3-7. 14) Go to step 8 I would like to speak a little more about step 6. It computes for each player the set of provinces which he is closest to, and adds up the "value" of each of these province. Fractional values are given in the case of a tie. (In addition, a simple bonus is given for the number of supply centers controlled.) The base value of each province must be determined by experience. For example, in chess, control of the central squares is of greater importance than the sides. This is known through experience. I have demanded the records of standard (or standard gunboat) diplomacy games resulting in a solo win. Through a statistical analysis of these files, I am hoping to see which provinces are key to a win by each player. The program implementing step (6) is available for examination. It uses a fixed value of "100" for each province. Ask for ~staykov/dipex/exp. Here is some sample output for the position before Spring 1901: nation=0 valeur=849 austria nation=1 valeur=1210 england nation=2 valeur=1258 france nation=3 valeur=1143 germany nation=4 valeur=1208 italy nation=5 valeur=1132 russia nation=6 valeur=691 turkey A future version of step (6) will allow multiple ply "deep" searches. For example, SPRING moves can anticipate SUMMER. SUMMER can anticipate FALL. FALL can anticipate both AUTUMN and WINTER. AUTUMN can anticipate both WINTER and SPRING. And WINTER can anticipate SPRING. In order to do this (for searches of greater than 1 ply), subroutine (6) would call the program using the calculated position as input, and with a search depth of one ply less. (3) Now, let's turn to the other half of the project for which Jean-Marc Aubert is responsible. His program will use CS's program to run the negotiations. Each turn, he will send the movements from the previous turn to CS's program in order to determine the alliance structure in effect in the game. He will then inform CS's program of the new position, and begin negotiations. The cost of specific requests can be tested via the REQUIRED MOVE and the FORBIDDEN MOVE input on CS's program. Whereas, the seriousness of a rumor can be tested by setting these movements as the default and testing their stability. General requests to ally (possible against a given target) are judged using the ALLIANCE parameter. A table is used to keep track of agreements in effect. For each player, an estimate of his honesty is kept. This is done by storing the maximum advantage he passed up by not backstabbing, and the minimum advantage he was willing to backstab for. Eventually, we hope to be able to generate diplomatic messages as well. (4) Some other new interesting files are now availble. (See the file dpp/table.of.contents). For example, we have: (a) a number of tables which give the distance between any two provinces for a fleet or for an army (b) the source for the Diplomacy Adjudicator (c) the communication protocol in between CS and JMA's programs. (d) some clues as to the communication protocol used by Sarit Kraus's diplomat. (5) Many of you have said you are going to write a diplomat. Given the size of this project, it is normal that some of you get discouraged. But please tell me about any progress you do make. It is important for the group to have an idea about what is going on. If you are working by yourself and get stuck, why not teamup with another member of the DPP and write one diplomat together, or help out my students by making suggestions or writting a subroutine for them. (6) Hopefully, next update will not be so long from now, and it will contain more sample output from our diplomat. In the meantime, what do you all think of the not so far fetched idea of having JUDGE output along with each set of results: the conjectured alliances and the estimated strength of each player. Yours, Daniel Loeb, DPP | "Hello, I'm a signature | virus. Join in the fun, E-MAIL loeb@geocub.greco-prog.fr | and copy me into yours." HOME 150, cours Victor-Hugo; Appt D45; 33000 Bordeaux France WORK LABRI; Universite de Bordeaux I; 33405 Talence Cedex France PHONE(H)+(33) 56 31 48 26, (W)+(33) 56 84 69 05, (FAX) +(33) 56 84 66 69 note new Fax number ========// Up