Standard financial theory believes that in the efficient market people cannot beat the market except by luck. However, in the real world, different investors use different lengths of past information to determine their investing process and gain profits. Therefore, this artical tries to investigate whether the trading behavior of long- and short-memory investors could affect the learning process of the market. A simple agent-based model is built to do the task. Although the technology of the model is still imperfectly, it seems to provide a reasonable sense to the actual world.