Journal on Policy and Complex Systems
Based on the simulation results , the optimal levels of self-confidence and aggressiveness are 0.45 and 0.94 , respectively . These values indicate that investors should use ( assign weight ) 45 % of their lifelong accumulated decision-making wisdom and use 55 % of the information received form the latest market changes when making decisions on stock market investments in order to maximize their performance in the stock market at that time . The degree of aggressiveness is optimized at 0.94 , which is very close to the initial upper bound for the variable . This high value indicates that agents should learn as much as possible from the strategy sets of the local best performers . Because the effect of learning is permanent , agents with a faster learning speed are able to mimic all the best rules in a much shorter time .
Since the market in our simulation was mainly a bull market , the bull market-trading rule set is more representative in the simulation . However , in simulations with a longer time horizon , the bear market-trading rule set actually evolves overtime , and it emerges as one of the major components in the final trading rule set . The best decision rule set is described as follows :
However , these rules are adjusted according to the degree of trust in the latest market momentum , which is generated on a daily basis . The above rule , after being modified according to the market momentum information , was updated to assume its final form as follows ( this rule was actually used by an agent on December 31 , 2014 ):
The rationale for the best performer ’ s decision rules basically states that an investor should track the past performance of a particular stock for a long time , and then make a decision whether to get in the market or not . As for the exit strategy , if the trend of the stock does not follows the investor ’ s expectations , then the investor should clear all positions immediately .
7 . Discussions , Conclusion , and Future Work
This agent-based model allows investors to see the behind-the-scene actions
of agents , as well as to make long-range forecasts of the anticipated behavior of agents . The CAS stock-trading model provides a higher return on
62