Profit opportunities, crash prediction and risk minimization in artificial and real-world markets

This paper reports on the use of multi-agent games to model financial markets. Our research employs multi-agent games to address three questions which are of great practical importance in quantitative finance: how profit opportunities may be identified, large price movements predicted, and inherent risk exposure minimized. The present paper focuses on the aspect of prediction. In particular, we report a technique for predicting future movements of financial time-series using multi-agent games. A third-party game is trained on a black-box time-series, and is then run into the future to extract next-step and multi-step predictions. Such predictions have potential use not only for speculative gain, but also as the basis for improved risk management and portfolio optimization strategies.