Abstract: Depreciation is a significant cost associated with farm ownership. Remaining value functions established as ASABE standards have provided a popular means of estimating costs, but these functions are not responsive to the conditions of farm machinery and agricultural economy in China, the test forecast errors are not acceptable. In order to develop new remaining value functions for large and medium-sized tractors, 116 recent tractor auction sales data in Shanghai State Farm were collected, by different Box-Cox transformations and regression methods, six proposed models with explanatory independent variables of tractor age and average yearly use hours were analyzed and compared, and the double square root model was selected as best one finally. The real test data also show this model can obtain satisfactory forecast results, and tractor age is the main factor for determining tractor remaining value.