SWAT model was utilized to simulate the results of Best Management Practices (BMPs) for non-point source (NPS) pollution in Dangjiangkou Reservoir basin. Cost, change of cultivated land area and reduction of non-point source pollution load of each BMP as well as the pollution reduction was taken into consideration when conducting the management practices optimization in different slope-degree zones. Multiple attribute decision making based on information entropy was applied to optimizing BMPs. Crop residual cover, contour ploughing and no-tillage were suggested to be conducted in region with slope degree less than 5°. In the zone with slope degree from 5° to 15°, terrace was demonstrated to be the most effective practice. This research also stated that hedgerows was firstly suggested to conducted in 15°~25° zone, while crop residual cover was another prior practice in this area. Conversion of cropland to forest had superiority as the best management practice in >25° area. This research provided decision making support for BMPS for NPS pollution reduction in Danjiangkou Reservoir basin.