Categories
Uncategorized

Game Theory in the Secret Hitler Board Game: “Testing the Chancellor”

Secret Hitler is a deception game I enjoy immensely with my group of friends, and I particularly love it because it feels a little less random (and more rewarding) than other popular ones like Coup and Werewolf. Decision-making actually matters, and topics of this course including game theory, dominant strategy, and payoffs are highly relevant to the game, although payoff matrices are hard to make because much of the payoffs aren’t possible to quantify into exact numbers.

The rules are fairly long, so I’ll just summarize the ones relevant to this post (though the full rules are linked at the end). At the start of the game, each player is randomly assigned one of two possible secret identities: Liberal or Fascist. There is a slight majority of Liberals. Additionally, one player among the Fascists is “Hitler.” The Fascists know who each other are, but the Liberals don’t know anyone’s identity.

There is a deck of cards, and each card is Liberal or Fascist (all cards look the same when facedown), which represent “policies.”

At the start of every round, all players vote for a President and Chancellor. If the majority of players vote Yes, the vote succeeds. The President draws the top 3 cards from the deck, chooses one to discard, and passes the remaining 2 cards to the Chancellor. The Chancellor then picks 1 card to play on the board (discarding the other) – this is called “passing a policy” and is the end of the round. Although these decisions are not made simultaneously, there is no communication from the President or Chancellor until after the card is played.

The Liberals win if a certain number of Liberal cards are played, and the Fascists win if a certain number of Fascist cards are played. Fascists can also win if Hitler is elected Chancellor at any time after 3 Fascist cards have been played.

There are also special powers the current President gets if a Fascist card is played, when a certain number of Fascist cards are already on the board. This includes eliminating another player of the current President’s choice. The Liberals can also win if Hitler is eliminated from the game. 

General Liberal goals:

  1. Get Liberal cards played
  2. Help Liberals identify other Liberals (to elect Liberal Presidents and Chancellor, who would pass Liberal policies, and to avoid electing Hitler)
  3. Identify and eliminate Hitler

General Fascist goals:

  1. Get Fascist cards played
  2. Appear Liberal in order to get elected (since Liberals have a majority)

In this blog post, I’ll only be approaching game theory and payoff discussions from the perspective of Liberals players before any special presidential powers are available, for the most straightforward position possible.

If a Liberal President draws 3 Fascist (3F) or 3 Liberal (3L) cards, there’s no choice to be made.

If a Liberal President draws 2F1L, discarding a Fascist card is usually the dominant strategy, since you generally want there to be a chance for the Liberal card to get played by your Chancellor.

The controversial scenario appears when a Liberal President draws 2L1F. Both my friends and the unofficial strategy guide (linked at the end) recommend discarding a Liberal card and passing 1 Fascist and 1 Liberal card to the Chancellor, “testing” them with a choice. However, the linked sources claim that from a game theory perspective, the dominant strategy is to instead discard the Fascist card and hand 2 Liberal cards to your Chancellor. Here are the immediate payoffs for the Liberal President in this situation:

Strategy 1: Discard a Liberal card:

  • If the Chancellor plays Liberal
    • Goal 1: positive payoff
    • Goal 2: neutral payoff
  • If the Chancellor plays Fascist:
    • Goal 1: negative payoff
    • Goal 2: neutral payoff

Strategy 2: Discard the Fascist card:

  • Chancellor plays Liberal
    • Goal 1: positive payoff
    • Goal 2: neutral payoff

The argument made by these two sources is that the Chancellor playing the Liberal card, even if given a choice, doesn’t really mean they’re Liberal. Furthermore, if they play the Fascist card, the Chancellor will obviously claim that they were actually given 2 Fascist cards and did not have a choice (and claim you’re framing them). There’s no way to convince the rest of the players that you’re telling the truth and your Chancellor is lying. So, the Liberal President gets an equal or better payoff by discarding the Fascist card, regardless of the Chancellor’s choices (and in this case, denying the Chancellor any choice).

However, the payoffs get infinitely more complicated when not just considering immediate payoffs. For example, finding out your Chancellor is Fascist may give you a chance to eliminate them from the game in the future, or your vote could be the crucial tie-breaker in not re-electing that player (who you now know to be Fascist) to the position of President or Chancellor in the future. Furthermore, if you do have other players’ trust, you may be able to convince them that this Chancellor is indeed a Fascist, and you are the one telling the truth.

In real life, of course, players are not perfectly rational, and much of the fun of the game comes from heated discussions and vocal accusations. When I’m in this situation, as a Liberal president who draws 2L1F, I opt to discard the Fascist card and not give the Chancellor a choice. A large part of this is because I know that I’m not the most persuasive person. I would likely be put at a disadvantage by being in conflict with another player, so the safer option is just to guarantee that a Liberal card gets played and I don’t lose any perceived trustworthiness. Still, it’s somewhat satisfying to find out that at least from a simple standpoint, game theory seems to agree with my decision. 

Sources:

https://www.lesswrong.com/posts/bXoaRShdqbf2P9Ma9/always-discard-fascist-policies

Unofficial Strategy Guide:
https://secrethitler.tartanllama.xyz/#as-a-liberal-1

Official Secret Hitler rules: https://www.secrethitler.com/assets/Secret_Hitler_Rules.pdf

Categories
Uncategorized

Who is Game of Thrones really about?

I didn’t eagerly dive headfirst into the Game of Thrones fandom by choice. Rather, after several years, I eventually bent to the will of the surrounding culture and our algorithmic overlords (another topic relevant to this class!), reading the posts, memes, and videos seemingly everywhere on the internet discussing and analyzing the popular series. After beginning to watch and read the series myself, I too was highly anticipating the final season bringing answers to long-speculated fan theories and ominous prophecies that were presented earlier in the story.

With such a large cast and different plot threads, one of the frequent topics of discussion I constantly read about was how important certain characters were to the overall story. For one, being integral to the plot would very likely keep the character alive a while longer, at least until their purpose had been fulfilled. Another reason was to make predictions for who would ascend the Iron Throne as the ruler of the country and who would become “The Prince that was Promised,” the prophesied saviour to end the ice zombie apocalypse known as the Long Night. Some argued these roles would go to the true heroes of the story, while others argued being too important would actually invalidate their candidacy, due to the series’ penchant for breaking traditional fantasy tropes.

The Network of Thrones project has done an astounding job visualizing all character interactions from the 8 seasons as a weighted, undirected graph, using many of the concepts about graphs discussed in CSCC46. Each character is represented by a node (most are unlabeled for clarity), and the edges are interactions, which include simply being together, speaking together, being mentioned together, or mentioning each other. More interactions between two characters (including longer dialogue) affects the weight of their edge, represented by increased thickness on the graph. 

The colours denote the 7 communities of the graph, which roughly represent the different plot threads and settings.

Furthermore, five metrics of importance are analyzed for each character:

  • Degree measures the number of edges (ie. connections) that a character has
  • Weighted degree measures the weights of all edges connected to a character (ie. number of interactions)
  • Eigenvector centrality measures the number of connections that a character has to “important” characters.
  • Pagerank centrality measures the number of interactions that a character has with “important” characters. 
    • Represented by the size of the label
  • Betweenness represents how important a character is for connecting different groups, measured by how many short paths pass through that node.
    • Represented by the size of the node (can be a bit hard to see sometimes)

(spoilers ahead)

Across all metrics, Tyrion ranks in first place, meaning he’s been interacting with plenty of characters and was critical to the plot development. Despite his scenes in Season 8 being cringe-inducing and forced to many viewers, it seems likely that he would have always ended up playing a crucial role in choosing the future of Westeros. 

Jon and Daenerys come in second and third place, respectively, and they are considered by many viewers to be the main characters of the series. Not only are both the only survivors of the overthrown royal dynasty, but both go through easily identifiable Hero’s Journeys. Given both spent several seasons in isolated communities away from the main cast (the Night’s Watch and Essos), it makes sense that they score low in eigenvector centrality but high in betweenness.

From my own experience reading fan theories over the years, these three characters were by far the most popular candidates to be the ruler at the end of the series, and the latter two to be “The Prince that was Promised.” Bran and Arya, the characters who actually ended up fulfilling these two respective roles, are both considerably less important by measures of centrality, and although sometimes mentioned, they were far less expected by fans as well.

More than just being an impressive feat visualizing 47,168 interactions between 407 characters, this graph seems to provide a good representation of character importance that matches viewer perception quite well.

Source:

Beveridge, A. (2020, January 14). Seasons 1-8 | Network of Thrones. Retrieved October 23, 2020, from https://networkofthrones.wordpress.com/the-series/seasons-1-8/