This paper develops mutual funds classification algorism under which higher degrees of performance persistence within the same fund class may prevail. Specifically, we hope to find a classification scheme that does not exhibit the problem of performance reversals. Our sample includes 3,981 equity funds from 1992 to 1997. Funds' standard deviation of monthly returns, average portfolio price/book ratio, average ROA, as well as stock holding ratios of major sectors are used in a factor analysis. The resulting four fund styles are stab le and they are better in terms of performance persistence comparing to the Morningstar classification. However, consistent with prior researches, there are still some significant reversals of past performance. After we add a variable representing the degree of a fund's momentum strategy to the factor analysis, the previous performance reversals are largely removed. This result shows that fund s dynamic investment strategies should also be included when determining funds' styles.