The Study On Agent-Based Optimal Trading Strategies

In recent years experimental economics has achieved rapid development. It has a belief that agents in real economy only have bounded rationality rather than full rationality which is believed by the traditional economic theory. The global financial crisis in 2008 challenged the efficient market hypothesis and the representative theory greatly; meanwhile it gave chance of development of experimental economics.In this paper, we realize an agent-based artificial financial market with the help Visual C++, designing four rules of agents'price formation based on herding behavior, similarity heuristics and theory of cost of carry and introducing a mechanics of clearing after trading a certain period. Traders in our artificial financial market make decisions to maximize their return undertaking certain risk. Traders will form their expected fair price according to their own functions of price formation, then give call or put to the market, all of which will be recorded in the book. In our artificial financial market, we adopt the mechanics of continuous auction to match the deals according to "price priority and time priority". The model is able to reproduce fat tails and clustered volatility, the two most significant characteristics of a real financial market. According to the two stages ARMA (p, q) models'R square, we find the agent in our model possess some ability to learn. We design a type of traders who learn optimal strategies using genetic algorithm. However, this type of traders cannot overcome the market.In this paper we identify the agent who has realized the largest percentual gain and get all the operations of this trader. However, our simulation results prove a simple strategy that most traders use to obtain better return:buy and hold when the market is during the upward trend and clearing all the positions when the market is during the downward trend.