This paper presents a multi-agent financial market simulation. The market is composed of traders who have different initial trading biases to take a specific action. Traders not only buy or sell an asset, but also cover their position in the following periods. Trading strategies are generated using stock price movements and other technical indicators. An XCS learning classifier system is used as an individual learning mechanism to implement the evolution of trader strategies. The results reveal that initial trader bias affects market price dynamics and evolutionary learning prevents the market from crashing, stabilizing the system. Covering mechanisms clearly illustrate the intermediate and minor trend following behaviors of traders. The results contribute to the understanding of potential deviations from efficient market equilibrium.