The invention relates to a cold-rolling mill motor loss torque coefficient test method. The method comprises the steps of acquiring design parameters of a rolling mill AC motor, and calculating the motor power factor multiplied by the efficiency value cos alpha*eta; determining the working conditions of a rolling mill before testing; applying rolling force at a given value; and obtaining the motor loss torque coefficient gL0-gL4 of the rolling mill through a regression formula. By actually measuring the parameters such as the rolling force, the roll speed, the motor voltage and the motor current and consideration to loss of the operation efficiency of a motor, the loss torque of the motor is calculated more accurately. The method is used for determining a theoretical regression model of the motor loss torque coefficient. The selected independent variable and polynomial structure are more reasonable. The deviation ratio of online prediction precision of the motor loss torque is guaranteed to be 10% or below. The cold-rolling mill motor loss torque coefficient test method can be extensively used for calculation of the loss torque of the motor of a single-rack or multi-rack cold-rolling mill and has a certain application and popularization value.