Agent Based Modeling in Financial Markets

I've been looking at this, and decided to build a simplified model of the Minority Game in Excel.

How can it be more simple? Well, I only used all avaliable strategies and didn't assign agents more than one strategy. If you are not familar with the Minority Game:

Imagine a time series that consists of 1 and 0.
Agents make a decision based on the recent past of the time series about what the next step will be.
The agents in the minority are correct in this t+1 step.

Of course there are other details, like the number of agents, and the fact agents are able to choose from more than one strategy, but these are the essential elements of the game.

The interesting part is that this game generates a series that will duplicate itself over and over, given the same initial conditions. It isn't random. However, it looks and feels random, and holds up to most statistical tests of random-ness. Neat, eh?

If you alter the inital conditions, you get a new resulant series. I'll explain why this is useful soon.