Journal on Policy & Complex Systems Vol. 2, Issue 2, Fall 2015 | Page 53

Journal on Policy and Complex Systems
trading volume and the transaction price of the stock . As a result , a trust metric may alleviate the impact of asymmetry in information by simply following market ’ s trends . Investors who adjust their trading rules to accommodate the current market trends can outperform the average investor and maximize their profits .
Although trust plays an important role in stock markets , learning is also an inseparable part of stock trading . It is an act of acquiring new or modifying existing knowledge , behaviors , or skills . Learning builds upon the previous knowledge over time . Similar to trust metrics , learning investment strategies involves mimicking the behavior of top market performers , copying their philosophy of stock trading . However , unlike trust metrics , learning is a continuous process . Investors are constantly aware of the latest market changes , and they possibly benefit from the current state of the market ( Linn & Tay , 2007 ). A learning curve captures the progress of learning over time . It is a graphical representation of the increase of learning with experience . The slope of learning reflects how aggressive ( eager , motivated , or capable ) the learner is in attempting to become a better performer over time . A typical learning curve is shown in Figure 1 . A more aggressive learner tends to copy more from the market best performers , and consequently has a much steeper learning slope . Less aggressive learners , however , tend to use best performers ’ behavior only as a minor correction to update their own trading rules .
Figure 1 . Learning curve .
Computers and sophisticated analytical techniques have offered an automated extraction of trading strategies , ( Teixeira & de Oliveira , 2010 ), although with limited success . In the past decade , complex adaptive systems ( CAS ) -inspired methods , primarily using agent-based modeling ( ABM ) techniques , have made advances in simulating traders ’ behaviors and capturing the intricacies of stock trading ( Kodia , Said , & Ghedira , 2010 ). This paper introduces an agent-based model for searching for the optimal balance between the level of trust and learning aggressiveness . The model is a derivative of the multisectors-trading model introduced by Su and Hadzikadic ( 2014 ). The system has been evaluated in the context of a financial services company stock performance in the period of 1987 – 2014 . The model significantly outperformed the buy-and-hold strategy on both S & P 500 and the bank ’ s stock .
50