A Jar has 1000 balls, X of which are Red. X is a randomly chosen number between 0 and 1000. Each of the N players is given a randomly selected number of balls. The total number of balls a given player is given is randomly determined.
Each player sees the colors of the balls he has been given, but not the color of the balls the other players have been given. After being given his balls each player lists a 90% confidence interval for how many red balls he thinks are in the jar. After listing this interval each player is shown the confidence interval of one or more of the other players. Each player then picks a new confidence interval, and then this process repeats.
That game could be played (albeit awkwardly) on a Less Wrong post.
The person running the game puts up a top-level post, writes a little program, and private messages everyone who wants to play with their result. Then each player posts a confidence interval as a comment, and away you go.
Here is a simple N player game that might work.
A Jar has 1000 balls, X of which are Red. X is a randomly chosen number between 0 and 1000. Each of the N players is given a randomly selected number of balls. The total number of balls a given player is given is randomly determined.
Each player sees the colors of the balls he has been given, but not the color of the balls the other players have been given. After being given his balls each player lists a 90% confidence interval for how many red balls he thinks are in the jar. After listing this interval each player is shown the confidence interval of one or more of the other players. Each player then picks a new confidence interval, and then this process repeats.
That game could be played (albeit awkwardly) on a Less Wrong post.
The person running the game puts up a top-level post, writes a little program, and private messages everyone who wants to play with their result. Then each player posts a confidence interval as a comment, and away you go.
Nice one, with many potential variants—both simpler and more complex ones.