Number of Agents

One of the really important points about the game is that these agents are trying to be in the minority. They want more agents to choose what they didn't choose.

For the basic Minority Game (MG), this problem is solved by simply using an odd number of agents. No matter what, there will be one side in the minority.

However, for this more interesting game by Lamper and Johnson, they use an additional criteria. The agents only play if the strategy has been a 'winner' in the recent past. So each agent keeps track of how its strategies have done over the last X steps (for L/J it is 100) and only plays if the strategy has been a winner more than t percent of the time (for L/J, 53%). So it is entirely possible that an agent will not participate on a given step. This will cause the total number of agents participating to fluctuate, and gives the possiblity of ties. Ties are resolved by a coin flip. Also, you can use the total number of agents participating as a volume proxy. If not many agents particpate, it would be low volume, and with many participating, high volume.

In the game I am creating, I might choose to use the difference in players as a magnitude in movement proxy as well. So if there is a large difference in the number of agents choosing one side over the other, this would correspond to a large movement in the market at that step.

I am just about finished describing the basics of the game. Soon I'll be able to start talking about the cool stuff!