Summary: Play some guessing games using a prediction market, to get used to how those markets work.

Tags: Small to Medium, Repeatable

Purpose: Prediction markets are a relatively popular tool in rationalist spaces, but uncommon in the general population. This is an opportunity to practice with a simple, short term prediction market.

Materials: You need the prediction market board. Robin Hanson described how he built his- fifty or so plastic bins arranged in a grid- and I made mine out of a pizza box. At the time of this writing there also exists a digital version created by Hamish Todd at, which I would recommend if you have more than five or six attendees. The pizza box construction steps are detailed in the Building A Pizza Board section.

Announcement Text

Want to play some games with a prediction market twist?

In the early nineties, Robin Hanson developed a board game to represent a prediction market. Called "Murder She Bet" it took place alongside a murder mystery, with participants using monopoly money to predict whodunnit. A full description of the original can be found at Hanson's site. Well, I don't watch many murder mysteries, but I do like board games, so on this occasion we're going to misuse this mockery of financial instruments!

We'll have Two Truths and a Prediction Market, where participants can work out which of three statements about their fellows is a lie. We'll have speed games, where two players will play a one-on-one board game while the audience predicts who will win- but the players are allowed to look at the market. We'll have a short trivia game, and see if the market can hone in on the answers to oddball questions about the world. We may have other uses for it- if you show up with an idea for what to do with a paper-powered prediction market, I'm open to suggestions!


The three main things you need to do during this meetup are 1. Explain how the board works and give people starting money, 2. Set up each game you’ll play with the market, 3. Act as the bank: make change, pay out the winners, reset the board between markets.

1. How The Board Works: 

The basic rules:

The board is arranged into columns (top to bottom) and rows (side to side.) Each column is for a different outcome. Each row is for a different price. Each slot represents a particular price for a particular outcome.

A slot always contains either a payout card (“Pays 100 if __”) or some amount of play money. Whenever a player wants, they can put in the payout card for the slot’s column and take out the play money that was there (selling) or they can put in play money for the slot’s row and take out the payout card (buying.) Put another way, each slot is for making one very specific trade, in either direction. 

At any time, players can also trade 100 for one of each payout card, or vice versa. 

At the end of a market, one outcome will be correct. When that happens, everyone holding a payout card for that outcome gets 100, and all the cards go back to the board. 

Those are the mechanics. How does this come together and actually function?

The emergent results:

Imagine a market with only one player, plus the bank. If they’re sure that one outcome will happen and the others won’t, then they buy lots of the payout cards for that outcome. They also might buy one of each payout card from the bank, and sell the payouts for the outcomes they think won’t happen. The board will change to approximate the probability that lone player gives for each outcome; if they think the first outcome is 80% likely to happen, then buying it at 70% is a good deal. 

With more players, each of them buying and selling where they disagree with the market, the board ideally comes to a kind of consensus position of all the players. Does that actually happen? Play and find out!

I gave people 200 in fake money when they showed up, and let them roll over their money from game to game and round to round. Feel free to experiment with the right amount, or with resetting money at some point in the evening. (A few people ran out by the end of the night.)

2. Playing Games with the Market:

Any game with a set number of outcomes works decently. Here’s a list of what I’ve tried.

Two Truths and a Lie worked well and was a nice icebreaker. Have people take turns making three statements (generally about themselves, such as “I have never gotten a speeding ticket”) where one of those statements is a lie. Everyone else uses the market to predict which statement is the lie, with a break between each statement to use the market. The person making the statements isn’t allowed to bet.

Trivia worked well, though I apparently make my questions a bit too hard. (Or the player working an arbitrage strategy was very persistent.) To make trivia work with the market, you need to make it multiple choice. The usual trivia writing tips apply; cover a wide range of subjects, include neat details and context in the question itself, exact dates are harder than you think. 

One on One board games did not work very well. Firstly, even if the players can’t buy or sell on the market, you can run into cases where market participants start giving advice to the players which. Second, you need the market to have a decent understanding of the game to be able to follow along. Last, if the board game is very fast (3x3 Tak) then the market doesn’t have much time to bet, if the board game is very slow (Chess with no timer) then everyone else can get distracted and wander into their own conversations. That said, I think this would work much better for a game you all have a common interest in.

I still think Werewolf would be really fun with a prediction market. Sadly, I haven’t figured out an incentive system to keep the werewolves from outing themselves immediately in a staggeringly profitable display of insider trading. Coup apparently went very well, but that happened at another table and I wasn’t paying attention to the details.

3. Be the Bank and the Market

Making change and paying the winners is straightforward. Resetting the board requires setting the initial market odds; you want to make sure it’s high enough that buying one of each right as the game starts isn’t an obvious profit (say, by putting the first payout card for each of four options at 20, so someone can buy all for 80 and be assured a 100 payout) but I think the ideal starting position is just high enough that won’t happen.


You can of course follow Hanson's example and use a murder mystery movie instead of board games. Suggestions for what movies to use are appreciated, especially if they come with suggestions for what names to put as options! 

The digital version follows the Hanson Original in assuming one market throughout the evening. When I've done this, I've run multiple shorter markets letting players keep their earnings until the end of the night. I like the multiple short markets, since I think the rapid feedback helps learn faster. Still, both work!

One option is to put, at the end of the night, a prediction market setup for something which won't resolve that night but will have resolved before you meet again. Write down people's positions and announce the winner next time you meet. 


If the players are allowed to buy and sell at any time, then sitting next to the board becomes an advantage. You can mitigate this by having people go in turns, but this hasn’t been a particular problem when I’ve run it. I let people verbally tell me what they wanted to buy or sell if they were sitting out of reach and that worked too, though in a larger or more aggressive group I could see verbal calls getting mixed up. Using the digital version sidesteps this entirely.

Be aware that having a single winner at the end of the night warps market behavior more and more towards the end. In the stock market, there's no "end" of the game and so making new large risky bets doesn't make sense. In Murder She Bet and its variants the last ten or twenty minutes can encourage making large bets, even at negative expected value, because whoever is in second place tries to overtake whoever is in first.

Building The Pizza Board:

Start with a large pizza box, at least 18 inches. 

With a pencil and a ruler, evenly divide it into 10 rows and 4 columns. I made my rows an inch and a half apart. Score (cut lightly into) the outline of the slots. They only need to be about a quarter inch high, and a touch more than three inches wide. Then there's an inch of space between each slot and its neighbor.

Cut the slots open. Careful not to tear it.

I suggest affixing the thin strips you got from the slots as guide rails hanging vertically down between the slots, to prevent the payout cards and fake money from sliding around. I did this with some awkward tape work, if anyone comes up with a more elegant solution for this let me know!

Label the rows. I labeled mine 5, 10, 20, 30, 40, 50, 60, 70, 80, and 90.

Label the columns. I labeled mine A, B, C, and D on one board, and E, F, G, and H on the second board. That way I could use one label card to declare what outcome each letter stood for for each different game. The two sets of letters were so I would be able to combine the boards if I wanted eight options in a single market.

Make the payout cards. I used sharpie and index cards, writing “Pays 100 if __” on each, replacing the blank bit with a letter.

I used play money from amazon.

Credits: The original idea is courtesy of Robin Hanson. The digital version is courtesy of Hamish Todd. The pizza formerly in the pizza boxes was delicious.

New to LessWrong?

New Comment