Interacting heterogeneous algo-traders : an extension of the Day and Huang model

Abstract

We develop a model with heterogeneous and socially interacting investors applying different technical trading rules (algorithms), by extending the seminal model of Day and Huang (1990). The original model consists of (sophisticated) -investors, (unsophisticated) -investors and a market maker. We have studied the nonlinearity features and described the dynamic behavior of the market. In the extended model, -investors are replaced by heterogeneous and socially integrated algo-traders. Through the communication process, each investor is able to obtain information about certain other investors and his characteristics (wealth, stress indicator and trading rule). If he finds a superior investor, he will adapt his or hers algorithm. Based on ten dissimilar technical trading rules we constructed some numerical experiments, and simulated the model. Then we evaluated the mean wealth and the long run price behavior. The combination of algo-traders and the sophisticated investors resulted in price fluctuates of different types. The volatility was typically highest at the beginning of the different price series, and in one of the series a stable 10-cycle appeared. This cycle seems consistent for some levels of the flocking coefficient in the bifurcation diagram that was generated for the original Day and Huang model. The main conclusion is that unsophisticated investors does not destabilize the market. Our extended model provides several starting points for future work.