LessWrong Diplomacy Game 2015

by Sherincall1 min read20th Jul 201528 comments


Personal Blog

Related: Diplomacy as a Game Theory Laboratory by Yvain.

I've been floating this idea around for a while, and there was enough interest to organize it.

Diplomacy is a board game of making and breaking alliances. It is a semi-iterative prisoner's dilemma with 7 prisoners. The rules are very simple, there is no luck factor and any tactical tricks can be learned quickly. You play as one of the great powers in pre-WW1 Europe, and your goal is to dominate over half of the board. To do this, you must negotiate alliances with the other players, and then stab them at the most opportune moment. But beware, if you are too stabby, no one will trust you. And if you are too trusting, you will get stabbed yourself.

If you have never played the game, don't worry. It is really quick to pick up. I explain the rules in detail here.

The game will (most likely) be played at webdiplomacy.net. You need an account, which requires a valid email. To play the game, you will need to spend at least 10 minutes every phase (3 days) to enter your orders. In the meantime, you will be negotiating with other players. That takes as much as you want it to, but I recommend setting away at least 30 minutes per day (in 5-minute quantums). A game usually lasts about 10 in-game years, which comes down to 30-something phases (60-90 days). A phase can progress early if everyone agrees. Likewise, the game can be paused indefinitely if everyone agrees (e.g. if a player will not have Internet access).

Joining a game is Serious Business, as missing a deadline can spoil it for the other 6 players. Please apply iff:

  1. You will be able to access the game for 10 minutes every 3 days (90% certainty required)
  2. If 1) changes, you will be able to let the others know at least 1 day in advance (95% certainty required)
  3. You will be able to spend an average of 30 minutes per day (standard normal distribution)
  4. You will not hold an out-of-game grudge against a player who stabbed you (adjusting for stabbyness in potential future games is okay)

If you still wish to play, please sign up in the comments. Please specify the earliest time it would suit you for the game to start. If we somehow get more than 7 players, we'll discuss our options (play a variant with more players, multiple games, etc).


See also: First game of LW Diplomacy


Well, the interest is there, so I've set up two games.

Game 1: http://webdiplomacy.net/board.php?gameID=164863  (started!)

Game 2: http://webdiplomacy.net/board.php?gameID=164912  (started! First phase will be extended to end on the 4th of August)

Password: clippy

Please note a couple important rules of the webdiplomacy.net website:


  1. You can only have one account. If you are caught with multiple accounts, they will all be banned.
  2. You may not blame your moves on the website bugs as a diplomacy tactic. This gives the site's mods extra work to do when someone actually reports the bug.
  3. Should go without saying, but you are not allowed to illegally access another player's account (i.e. hacking).



28 comments, sorted by Highlighting new comments since Today at 2:17 AM
New Comment

I'm in.

adjusting for stabbyness in potential future games is okay

I thought this was 'metagaming' and generally looked down on?

It wouldn't be in any way against the rules, but depending on how it is executed, it may be looked down on.

"Last game Sherincall stabbed everyone every chance he had. I'm not trusting him." would be pretty valid, the person obviously isn't trustworthy.

"Last game I trusted Sherincall and left all my centers open, and he stabbed me and soloed" would mean that you didn't properly understand the game - We are all playing to win, and everyone should choose to win if the option is available.

"Last game Sherincall stabbed me, so now I'm just going to make life miserable for him this game" would mean that you're not a particularly rational agent (a grudger bot?). That is looked down on because you will be a poor player if you stick to those principles. If you wont ally with me, I'll ally with person X and attack you. And then people will probably not want to play with you again.

[-][anonymous]6y 0

This would be game theoretically sound behavior if you planned to play many games in succession.

Slow reply, but two things:

  • Whether that is game theoretically sound highly depends on the other player's behavior. If it makes other players afraid to stab you for fear of retribution in the next game, then yes, it works. But I think that among experienced diplomacy players, it is more likely to get you excluded from the game entirely. Also people will be less likely to ally with you in the first place if you have a reputation if responding badly to stabs.

  • This kind of behavior is explicitly against the rules in most online diplomacy communities. Because it ruins the fun of the game. Diplomacy as a game works best when everybody is really playing to win with a cut-throat, no holds barred, sell your own mother if needed, kind of style. A player who is meta-gaming is not playing to win for that individual game, so it lowers the experience for that individual game.

(Of course tarnishing someone's reputation, deserved or undeserved, based on actual or made up previous games, is an entirely valid strategy. So long as you play each individual game with the goal of maximizing your performance in that individual game, keeping previous games in mind is perfectly fine.)

Seems like I have missed the fun. Maybe if someone drops out, or otherwise next time. I used to play a lot of diplomacy online, but I haven't in a while. It would be fun to play another game.

Two remarks though.

One: Like I said, I haven't played diplomacy online for a while, so this information might be outdated, but I think playdiplomacy.com is a much better platform than webdiplomacy.net. In my experience is a more mature site, with a more mature audience. Webdiplomacy also doesn't allow illegal orders, which in my opinion is completely against the spirit of the game (It's probably somewhat friendlier to new players, but at the cost of denying them the full experience of the game).

Two: Demanding a 90% reliability is WAY too low. An in-game year will require entering orders between 2 and 5 times, say 3 on average. That's 30 deadlines in a 10-year game. With 90% reliability each player will have only 5% chance of not missing any deadlines. Now keep in mind that there are 7 players, and that a single missed deadline by any of the players can potentially ruin a good game, and it becomes a certainty that your game WILL BE RUINED by this if you demand only 90% reliability.

A much higher level of commitment from all players is required to have a fun game.

Seems like I have missed the fun

Two slots left...

Demanding a 90% reliability is WAY too low.

I assumed it meant a 90% prob of making every deadline.

Maybe if someone drops out, or otherwise next time.

Well, the game is FCFS. I sent PMs to everyone who applied on the thread to give them a head start, but the password is public. There will also be a second game as soon as enough people express interest (I think 3 so far).

but I think playdiplomacy.com is a much better platform than webdiplomacy.net

Honestly, I chose webdiplomacy.net as I'm a regular there. I know it works, and the illegal orders protection is not that much of a bad thing really. The community is irrelevant in a private game, and I know I can easily get a hold of the mods if the need arises (e.g. if an unexpected pause is required)

Demanding a 90% reliability is WAY too low.

It is low, but not that low, because you are forgetting the 95% heads up. So, the player will miss every deadline with 0.1 * 0.05 == 0.005 (0.5%), which is okay. I think the 90% requirement is okay, but the 95% one should have been higher (and was originally)


This comment will explain the basic rules of the game. Reply to it with any questions relating to the rules themselves.

You can also read the rules on WikiBooks

The board initially looks like this.

You are randomly assigned a great European power. This is the only random element of the game. As you can see, the powers are not equal.

The map is divided into land and sea territories. These are traditionally abbreviated with 3 letters. The seas are uppercase (e.g. ION - Ionian Sea), while lands are written in lowercase (e.g. Bel - Belgium). Switzerland is unpassable. Some of the land territories contain Supply Centers, which are marked with a circle. There 36 of these. As soon as someone owns 18 (50%+1), they win the game. The SCs that are colored on the above map are a power's Home Supply Centers. These have a special meaning for the power they originally belong to (for others, they are the same as neutral SCs).

There are two types of units: armies and fleets. The fleets can only occupy sea and coastal territories. Armies can only occupy land territories. A power can posses as many units as they own supply centers. All units are of equal strength.

A game year is consists of five phases, some of which may be skipped if there is nothing to be done:

  • 1) Spring moves (AKA Diplomacy phase)
  • 2) Spring retreats
  • 3) Fall moves (AKA Diplomacy phase)
  • 4) Fall retreats
  • 5) Builds/Disbands

During a Moves/Diplomacy phase the players submit orders for every unit they own. Once all orders are submitted, they are evaluated simultaneously. There are 4 types of orders:

1) Hold - The unit does nothing, just holds its position.

This order is traditionally written as "A Ven Holds" (army in Venice holds) or "F BLA H" (fleet in Black Sea holds).

2) Move - The unit attempts to move to an adjacent territory.

This order is traditionally written as "A Ber-Mun" (army Berlin moves to Munich) or "F Lon-ENG" (fleet in London moves to the English Channel).

In a basic case where no other units attempt to move to a territory, the move succeeds, and in the next phase that unit is in the new territory. If a unit attempts to move to a territory occupied by another unit (friendly or not), that move will fail, unless the other unit moved out. If two units attempt to move to the same territory, neither move succeeds. This is known as a bounce. As an example, let's assume there are units in Greece (Gre), Serbia (Ser) and Romania (Rum), while Bulgaria (Bul) is vacant.

  • A) A Ser-Bul, A Rum-Bul, A Gre-Bul; No moves succeed, everything remains as is entering the next phase.
  • B) A Ser Holds, A Gre-Ser, A Rum-Bul; Greece army doesn't enter Serbia, Romanian army moves to Bulgaria.
  • C) A Ser-Bul, A Rum-Bul, A Gre-Ser; Serbia and Romania bounce in Bulgaria, meaning Serbia is not vacant. That means the Gre-Ser move fails as well, no one moves in the end.
  • D) A Ser-Gre, A Gre-Ser, A Rum-Ser; Serbia and Greece attack each other, nothing happens. Serbia is still occupied so Romania can't move into it.

Note: It is possible for two armies (e.g. in Gre and Rum) to defend three territories (e.g. Gre,Bul,Rum) by bouncing themselves.

Note: When it comes to coastal territories (i.e. a land bordering a sea) that can be occupied by both armies and fleets, same rules apply to both.

Now the question is how to defeat an enemy unit. The third order type is:

3) Support - The unit helps another unit move/hold.

This order is traditionally written as "A Ser S Gre-Bul" (Army in Serbia supports the move to Bulgaria from Greece) or "A Rum S Bul Hold" (Army in Romania supports the Bulgarian unit's hold).

A unit can offer support in any territory it could move to (e.g. Greece can't support Ser-Rum, because it cannot move to Rum. But it can support Rum-Ser). A support order is only valid if the supported unit does the specified action (e.g. Support Hold is only possible if the unit does not move. It fails even if the unit tried to move, but that move failed).

If an army is attacking a territory with support from another army (army owners are irrelevant for order resolution), we say that the attack has strength 2. If there are two moves to a territory, the one with a higher strength succeeds. If an army attacks a territory with strength 3, and the territory occupied by another army that is supported by only one other (strength 2), the defending army is dislodged and must retreat. If the attacks are of the same strength, the same rules apply as shown in the above examples.

Note: If there are two attacks of strength 2 on a territory occupied by an unsupported army, the two attacks bounce, and the occupying army remains there.

Instead of supporting hold of an army, you can counter a support by attacking the supporting army. If an army is attacked (even with strength of just 1), it must defend itself and cannot support another battle. Thus an army in Albania can help an attack on Bulgaria by hitting Serbia or Greece to "cut" that support, even though it cannot support directly. This is known as tapping.

4) Convoy - A fleet can convoy an army across sea territories.

This order is traditionally written as "F ENG C Lon-Bre" (Fleet in the English Channel convoys from London to Brest). When paired with "A Lon-Bre", it lets armies move across water. Armies can move across multiple water territories if there is a fleet in each one, and each fleet has the same order. The fleet owner is not relevant, only the order is.

A convoy can be disrupted in any fleet in the chain is dislodged (forced to retreat - convoys can't be "cut").

After the moves, comes the retreat phase. Any units that have been dislodged can retreat to any unoccupied territories adjacent to the territory they were dislodged from. If no such territories exist (or if the player so chooses instead) the units are disbanded.

Finally comes the builds phase. Any supply centers that have units occupying them change ownership to the player who's units those are. Unoccupied SCs keep their ownership from the last turn. So, to conquer a SC, you must hold it at the end of the Fall phase. The spring phase doesn't matter in this regard.

A player can have as many units as they own SCs. If they lost SCs this year, they choose which units to disband. If they gained SCs, they can build new units in one of their unoccupied home SCs. Fleets can, of course, only be built in coastal SCs.

Note: It is possible to have many SCs, but not have anywhere free to build if, for example, you had to defend your home centers, so they are still occupied. Or if you lost them.

Some special cases:

  • Bulgaria, Span and Saint Petersburg have distinct coasts (south and north). A fleet occupying one of those can only move to territories neighboring that coast. Thus, for example, a fleet in Spain's north coast can only move to MAO, Portugal or Gascony, but not to Western Mediterranean or Marseilles. If a fleet can move to either coast, it can give support in the entire territory. Only one coast can be occupied at the same time.
  • Constantinopole and Kiel have channels running through them, so they don't have coasts.
  • If two dislodged units retreat into a same area, both are disbanded.
  • Two units cannot exchange territories in one turn (A->B, B->A will bounce), but three can rotate (A->B, B->C, C->A). Fleets cannot exchange territories using coasts (it's the same territory).
  • You can retreat into an enemy SC and take control of it that way. Always know where the opponent can retreat when attacking them.

Like Chess, Diplomacy has some standard openings, for each country. You can read about most of them here. It might be worth to check out the ones related to your country.

[-][anonymous]6y 2

I'll repeat my suggestion from before, that you open up a password-protected WebDiplomacy game, and post the password here, to avoid issues where someone says they're in and then doesn't actually register. If one game fills up, open up a second, and so on.

(I also recommend turning on the Anonymous Players setting.)

I expect not to play, but may if there are only six other people interested.

I'll play.

I'm going hiking for a few days next week, but if we start in August and/or people are happy to pause, I'm in.

I estimate at least a week for it to start. Let's put the tentative date at 1st of August.

Germany is at risk of missing turn 1 in the first game; if you joined it, you should probably log in.

I have contacted the mods to pause the game and ping the player through email. If that doesn't work, we'll look for a replacement. It'd be a shame to waste 2 days of press between other players.

Missing your very first turn is a bad sign for future reliability. I don't know about webdiplomacy.net's etiquette, but on the site where I usually play such players are generally booted straight away, and a replacement is sought.

I'm willing to jump in if Germany doesn't show up. I'll be online again later tonight.

Please make an account and PM me the username, and we'll get the mods to do the swap.

The second game has started early. I contacted the mods to extend the first phase in case someone does not check before the 1st.

(I'm playing only in the first game, but keeping an eye on this one)

One player hasn't checked the second game since the start. If you have joined, please log in so the game can proceed.

[-][anonymous]6y 0

I'd play Diplomacy if there was a variant that could be completed in one sitting, online - say really fast turns.

[This comment is no longer endorsed by its author]Reply

I'm confused by the existing game: one player has 6 points, everyone else has 5, but the site says you start with 100.

(Also, it doesn't seem particularly anonymous for the points to be displayed.)

The worth in points means relative to the pot in that game, in terms of supply centers owned.

So we all paid 5 points to enter (thus people starting with 100 have 95 left), and the "worth" is calculated based on our relative power in the game. The sum will always be 35+/-1 for rounding. Currently Russia has 6 points because they have 4 SCs while everyone else has 3. Once we conquer all neutral SCs, the worth will pretty much be equal to SC count.

Now that you understand it, you can safely ignore it since those actual values mean nothing for a game like this (a winner-takes-all, and everyone shares equally in a draw).

I'm in! Diplomacy is great fun.