Stock investment has been a popular, convenient, but important means of portfolio management for the public. And yet it is often twisted as being highly speculative. The empirical study done by this article covers data for Banking and Insurance during the years of 1990 to 1995. In this article we attempt to arrive at rules for short term and long term of stock fluctuation from probabilistic points of view through various methods, namely kernel method, variable kernel method and maximum penalized likelihood method. In order to deduce better investment rules for the public investors, we find the probability densities of the stock fluctuation for 14 companies. Two major findings can be derived from this research. First, the probability densities of the stock fluctuation for short term are different in the situations of politics and economics; the probability densities of the stock fluctuation for long term is symmetric with narrow high peak. Second, we find that probability densities of the stock fluctuation for 14 companies are all symmetric with narrow high peak. We also give the suggestions for the risk aversors, the risk lovers and the long-run investors based on daily average return, risk and yearly dividend.