Agent-based market analysis [electricity supply industry]

The authors propose an extension of Monte Carlo computational methods to evaluate complex financial options and to explore market dynamics in complex markets that are not well described by standard techniques of financial analysis, such as markets with network or capacity constraints. Instead of the the diffusion equation that is normally used in financial analysis and its implied random walking price trajectories, replace them with independent computational agents, each with finite decision making power, memory and possibly learning capability. A cloud of agent-based ´smart financial particles´ is released in a simulation, and each particle describes a path into the future, taking in information from its environment (interacting financial time series of the underlying commodities) and making decisions based on that information, such as hedging, diversification and speculation. Even though the actual content of the agents´ portfolios will soon diverge, methods of VAr and NPV can be used to evaluate the performance of these smart particles and reduce them to a single number. This will allow the valuation of complex portfolios and contingent options in a consistent manner with current valuation techniques.