To solve existing problems for evaluation of distribution line loss and the related improvement measurement, this paper presents a probabilistic algorithm, which is used to evaluate the loss reduction strategy. In view of lack of real-time data about distribution network, firstly, the distribution loads are classified by their characteristics, and the same class loads have similar power curve. The loads obtained are assigned by power curve and energy supply. Then, all of load data are revised by matching power flow and data under typical data. Based on above-mentioned, the mathematic distribution character of line loss with load fluctuation is calculated by Monte Carlo simulation method, which is the basis for evaluation of loss reduction. The validity and efficiency of the proposed method is testified by real-time data of 10 kV distribution line at somewhere in Shandong Province.