Market theory dynamics

Here's an idea for a complexity theory thesis based on my The Game of Monetary Policy post (and by genetic algorithms).

Assumptions

1. Markets are created by willing buyers and willing sellers. The price of an asset is the price they are willing to trade at. The only cause of price dynamics is trading.

2. There are always irrational (at least the sense of in-model wealth maximisation) buyers and sellers.

3. There are always theory-driven buyers and sellers: they have an idea, and they trade based on that idea.

4. Successful ideas eventually become well known.

5. Unsuccessful ideas are eventually discarded by some (but possibly not all) market participants.

Thesis Plan

1. Literature search, all the usual guff

2. Select implementation technology: swarm perhaps?

3. Implement a model based on a 2 level simulation which I'll call simulation and meta-simulation. There is a population of buyers and sellers. Some are trade randomly (see 2.). Others trade with a rule-based theory. We start off with some distribution of assets among the traders, and trading is permitted for some number of rounds, say 75 (rough no. of business days in a quarter).

4. At the end of the simulation, the meta-simulation takes over. All traders publish their rules. All traders assess their rules based on their personal performance. Agents that have done well are less likely to change; ones that have done worse are more likely to change. Everyone who changes can select a published rule, with weighting driven by how successful that rule has been for others.

5. Repeat 3 and 4, ad lib, or at least through 100 meta simulations. See what the asset price dynamics looks like as a function of the simulation parameters. Can we recover the kind of fat tails and positive auto-correlation we see in the dynamics of real financial assets?

6. Review results, present conclusions.

There, the actual execution is entirely routine from here...