Unsocial Deduction Games

Why I love Social Deduction Games

I am a huge fan of social deduction games: Avalon, Secret Hitler, SpyFall, Werewolf, you name it. I like to think of myself and my friends as honest people, but as soon as we start up a social deduction game, no can be trusted. The rush of anticipation, followed by gasps of astonishment and groans of betrayal make a social deduction game a fun and memorable experience.

In the past, gathering a group of friends to play any game was a tedious process: you’ve gotta find times everyone is free, get over to someone’s house or apartment, and set up the game. However, thanks to websites like netgames.io (that have seen a huge boost in popularity recently), the barrier to organizing a social deduction game night is minimal.

For as much as I love an afternoon of lies and deceit, I have noticed a few recurring problems in many of the social deduction games I have played.

Challenge 1: Balancing the Number of Players

Here is a graph of four of my favorite social deduction games using stats compiled from boardgamegeek.com — notice anything interesting? For one, they all have a “best” player count of five or more! Do you know how hard it is to find four other friends to sit in your tiny apartment and argue for an afternoon?

Even if a game can be played with three to four players, it almost always results in the “evil” team being forced to reveal too much information and make riskier plays. Some games try to avoid this issue by including special roles or modifying the rules for smaller group sizes. Despite these techniques, making a fun social deduction game that works well, especially for smaller groups, is difficult.

Challenge 2: When Games get Boring

Maybe you aren’t like me. Maybe you have tons of friends all knocking on your door at all hours, begging you to grace them with your presence and bust out the board games. In that case, Avalon having an optimal player count of seven to eight players is a complete non-issue, right?

Wrong.

With more players come other obstacles and pain points in every social deduction game. Playing a game with your ten best friends may seem great at first, but after a few minutes you begin to notice problems. Not everyone is participating as actively, and the conversations begin to get dominated by a boisterous few. Maybe someone is checking their phone, and they miss an important detail. Someone gets eliminated — what do they do now? Watch resignedly, helpless to aid their struggling teammates who have somehow all revealed their alignment and are now trying to win the game by loudly shouting baseless accusations?

Many of my friends who do not like social deduction games give this as their main criticism: social deduction games are boring. I don’t blame them. I too have sat through single turns of Secret Hitler that drag on for thirty minutes because someone is agonizing over which person to select as chancellor, then drag on for another forty minutes as someone argues why that person shouldn’t be chosen as chancellor. And sure, some games avoid this by pushing you to set a time limit on turns and rounds, but there is only so much that can be done to increase the speed of play.

Challenge 3: When Games get Too Complicated

“Wait, so now the president is choosing the next president, who will then choose a chancellor? And if we vote in favor of this chancellor, and the new president that was just chosen by the old president gets three fascist policies, then they can veto the cards and draw new cards? But if the new president can’t veto and we pass a fascist policy, the new president can execute someone? And none of this matters if the chancellor is Hitler, because then the fascists win automatically? But if we vote against this team then we will draw a policy card from the deck randomly, and that has a 1/3 chance of being a fascist policy? Is that right? Or am I about to throw away the whole game just because the rules are so convoluted that its become more about understanding random technicalities than enjoying a board game with friends?”

Challenge 4: Balancing “Social” and “Deduction”

My least favorite social deduction gaming experience happens when players are able to deduce the truth with so much certainty that bluffing is futile. When this happens, the game ends not with a big reveal but with a “Well, you nailed it. Good game, I guess.” To give a convoluted analogy: I want my social deduction games to play out like murder investigations on TV, not like murder investigations in real life. I want accusations, I want plot twists, and I want to be in my own apartment so that afterwards I can go to bed.

Most games combat this in a few ways. They add in randomness so that fully deducing everyone’s role is impossible. They introduce benefits to bluffing not just about your role but about what cards you have, what decisions you made, and who else is on your team. In my opinion, these tricks keep the game social. They allow you to win favor on charisma and confidence even when the odds of your story panning out are slim. Conversely, when there is too much randomness and too little information, the balance swings in the other direction and you feel like you are playing a game of luck. In other words, social deduction games are fun if they can promote deduction and strategy without eliminating social aspects.

I think a good measure of how well a game balances social and deduction is by examining how often computers beat people at the game. Computers are great at deduction, but not great at socializing. (Sorry, Twitter bots.) For example, grandmasters lose to the best computers at chess 100% of the time. It’s not close. Chess is pure strategy. Accusing the computer of being a dirty spy doesn’t improve your position, it only hurts the computers feelings. In Poker, a game that relies on some social bluffing but arguably a lot more strategy and deduction, computers win about ten times more than the average pro. And in a pure game of luck, say, flipping a coin, computers are only marginally better than humans.

So, how do social deduction games fare against computers? An AI for the game Avalon called DeepRole won about 60% of its games against online opponents, which is actually 12% higher than the human win rate. Computers can deduce and strategize enough to beat good players, but until they improve their social skills they won’t be mopping the floor with us. I like to think that social deduction games are one of the last remaining battlegrounds in science’s never-ending quest to make us all feel inferior to a box of wires.

An Idea for Improvement

With this list in mind, I set out to make my own social deduction game with the following requirements: any number of players can play and have fun, everyone is always actively participating, and the game balances “social” and “deduction” without being too complicated.

To be continued…

Credits: Monica Vyavahare (proofreading / inspiration) and Gerd Altmann (photos)